overclocking

[modern walrus]

Why do I have two of these & which one is right?

THey both say the same thing. The part not showing says Intersil ISL 69269

maybe because I have two power connections to the motherboard. One is typical 8+8, the other 4+4 (my PSU only has one 8+8 for CPU & the other is a dedicated Molex with converter--I should probably upgrade from 750w to 850w so it has two 8+8s)

sorry 4+4 not 8+8

VR VOUT = 1.305 (1.198 other one) POUT = 242w Input = 260w IOUT = 190 amps (max(

1 amp out of the other ISL69269 = useless

max temp = 88 (CB R23). Delta fans are very loud

[modern walrus]

ya I dunno I can survive CB with 52 all core, -1 AVX offset, & 49 ring ratio, but it maxed out at 91 degrees at about 1.310v under load

[steel zephyr]

My DDR4 5066 MHz G. SKill RAM keeps giving me errors at speeds above 4000 MHz using the XMP profile. TestMem5 is giving me hundreds of errors in a short time. I have the BIOS to optimized defaults and I still get errors above 4000 MHz. I'm not sure if my 10900K can't handle that speed or if its faulty memory. https://www.newegg.com/g-skill-16gb-288-pin-ddr4-sdram/p/N82E16820374177

Its on the ASUS ROG Maximus XIII Apex motherboard.

I checked the BIOS and its using the correct voltage and timing from the XMP profile.

[lavish tundra]

What CPU tho

[steel zephyr]

10900K

[modern walrus]

damn 5066mHz? choppin me up

[steel zephyr]

I know other people using the same 5066 MHz kit with a 10900K, so I figured I'd give it a try when it was on sale.

[lavish tundra]

10900k can have problems running above 4266 from der8auers testing IIRC

Your IMC is probably having issues

[steel zephyr]

I also tried manually setting the VCCSA and VCCIO manually to 1.2 then 1.25v and it didn't seem to make a difference. I guess the CPU can't handle it. I guess its just silicon lottery on if a 10900K can run it?

[lavish tundra]

From what der8auer tested all his 10900ks could do 4266MT/s

You must have a really bad bin if it struggles over 4000MT/s

[steel zephyr]

The Asus BIOS CPU rating thing actually rates my 10900K above average, but I guess that's just clock speed and voltage, not memory controller.

[lavish tundra]

I would wait for someone more knowledgeable than me to help th

someone like fitz or arshia or alatron

[steel zephyr]

I've been away form Intel overclocking for a few generations, so I'm not sure if I'm missing something. I've read other people's guides, but I haven't been successful in getting higher memory speeds yet.

[lavish tundra]

Ask alatron for help imo

[steel zephyr]

ok thnaks

[tall pelican]

Most 10900ks should be able to do 4400 dr, and like 4500 sr

Try like 1.3 io/sa

[steel zephyr]

The motherboard has fast boot enabled by default, which apparently doesn't retrain the RAM on reboots. I turned off fast boot and have better results now. So far DDR4 4400 MHz is stable and passes the tests in TestMem5 without issues. I'm testing DDR4 4600 MHz now.

[steel zephyr]

When I disable fast boot, 4600 MHz is stable without errors so far in TestMem5 with the extreme config. Trying 4666, 4700, and 4800, gives me lots of errors.

[granite pawn]

what do i do with it?

[short blade]

right click

run as admin

[orchid flame]

10900K can’t support that without proper tuning

[orchid flame]

Most are running that with a RKL chip

Since the RKL IMC does much better with high frequency DJR

You’re gonna need to tune IO and SA a lot

As well as properly set your Rtts

[orchid flame]

No

[orchid flame]

Those are nothing.

Try 1.35v IO 1.45v SA

RttNom off, RttWr 80, RttPark 240

[clever epoch]

We're never going to stack high performance logic without exotic cooling. Just not gonna happen.

AMD already has a patents on dummy silicon structures designed to improve heat transfer from underlying logic. It's possible we'll see those structures on the upcoming Ryzen vCache CPUs, although I'm not sure

[dull ginkgo]

I mean, I doubt it, it's just what, hbm2, not going to get that hot

[dull ginkgo]

but well, I believe there's a possibility if we can get some distance between the logic dies, of course that'll mean more latency and worse performance, but that's the next logical step imo, upwards

[steel zephyr]

Yes I did. I was wondering if my unique cooling would get his interest. That is my daily setup I used for the benchmark run and is why I used the quick disconnects.

[clever epoch]

HBM is memory, we've been stacking memory for years now. I'm saying that we'll never have logic stacked on top of (high performance) logic. Waaaay too much heat

[dull ginkgo]

I was replying to the v cache using dummy silicon structures for heat transfer, where I believe hbm will be used for v cache

[clever epoch]

Also, core-core communication isn't actually that important

[clever epoch]

That is not correct

It's not "memory" being stacked, it's SRAM

It's extra processor cache, basically

It's very much novel though

But it's another die, composed entirely of SRAM stacked on top of the existing SRAM on the CCD

[dull ginkgo]

Looks like I misread some things, although there is definitely some performance gain out of the extra l3 cache from what we've seen so far

[clever epoch]

And on top of the logic, to even out the height of the die, is the dummy silicon

[proven canopy]

Lol yeah, that vcache isn't anything close to HBM

[dull ginkgo]

hm

[clever epoch]

I believe the dummy silicon and SRAM are one silicon structure, I think it would be too complicated to stack 3 separate pieces of silicon on top of an existing die

Could be wrong in that regard, who knows

But yeah, it's not memory, it's extra cache. Turns a Zen3 CCD from having 32MB cache into 96MB

[dull ginkgo]

It's definitely possible from what TSMC is capable of, stacking multiple pieces of silicon, but who knows

[clever epoch]

It is absolutely possible, but there's no reason to add additional complexity. Fancy packaging has its own costs, and at this point the extra die area is probably negligible in terms of yields, especially considering most of it would be dummy silicon

But again, I could be wrong and they are actually stacking 3 different pieces on top

[clever epoch]

Actually I think that the dummy silicon is distinct from the cache chiplet. The dummy structures don't need to align that precisely, only the cache chiplet.

And the dummy structures would take at least as much space as the cache, so why waste half the wafer on dummy structures?

[grim shore]

Ok so I have a core i9 10850k on a rog strix z490 A, I believe and I want to overclock above 4.8 ghz it says in the bios that it turbos to 5.2 I managed it to get to 5.8 one time very unstable (froze right after) I was able to have it run good over the 4.8 but I was tweaking around and it hasn’t been up ever since. Any help would be great 👍

[tall pelican]

well for starters, you arent doing 5.8

[proven canopy]

5.8

[modern walrus]

5.8 is pretty heckin crazy

Got RAM chillin at 3400 16-19-19-38. Baby steps.

Change from stock: +200mHz/no change/+1/+1/no change

My motherboard has the options to increase VDDCR CPU current capability & VDDCR SOC current capability to 110, 120, or 130% just ran prime 95 & my pump is bored cuz thermals maxed like 67. Will bumping those both up allow me past 140 amps under load? Voltage was like 1.02 or something way low.

I know y'all are all about that RAM overclocking lately but this might be a decent thing to pin because it explains AMD BIOS terminology https://www.hisevilness.com/articles/technology/amd-overclocking-terminology-faq.html?showall=1

Maybe this one too https://www.hisevilness.com/articles/technology/ddr-ram-overclocking-terminology-faq.html

Well setting both to 130% (max) didn't raise EDC past 140A

Heck is this?

[zenith palm]

If you oc'd to 5.8 and it booted you're cpu already degraded my guy, the voltages you would have to use to get that to boot

[modern walrus]

Should I not exceed the motherboards PBO limits?

ie motherboard values for PPT, TDC, EDC

[zenith palm]

Yeah you can tweak your own ones, decent improvement depending on mobo

[modern walrus]

My motherboard is eh

Strix B550-f

What goes PBO overdrive scaler do?

[zenith palm]

Just leave at 1x for the moment iirc

I'd say strix has alright mobo pbo settings but for max perf still tweak

Or auto while you mess with everything else

[modern walrus]

I know I can't go over 1.2v

So how the heck do I get this thing to start cooking?

I have stupid thermal headroom & nothing to use it for

I gotta send this 5900x back & get a 5950x lol

[zenith palm]

Why lol?

[modern walrus]

Damn why did I have to even think about that

because I have like a Ferrari cooling system on a nice corvette lol

Corvettes are awesome. The new mid engines are cool af

But Ferrari man. Ferrariiii

Haha I'm being ridiculous my bad

Haven't slept muchn

[zenith palm]

Ehh I've only done my 3600 pbo but basically you wanna take note of ppt edc and tdc under load, then up them

You fine lol

[modern walrus]

Well rn I'm clearing cmos because I dunno why

But no heckin POST

But that probably cuz I messed with some memory crap.

I never learn

But how much can I bump? Like 10 at a time? Will PBO even use the extra amps?

I get the advantage of PBO but it's like holding my hand

[zenith palm]

Yes it should

Ehh i think 10 is fine 1 sec lemme check

[modern walrus]

because it stays right at the 140 limit all the time

[zenith palm]

@modern walrus this is advice i got for my 3600 but i would bench and see what values you get first then increment in 5 or 10s

[modern walrus]

The concern is what? Thermals or voltage? Both?

[zenith palm]

Ehhh I'm not entirely sure but voltage/amps most likely

[modern walrus]

What happens if I change switching frequency?

It's stock 200mHz. How high should I go?

[zenith palm]

Think it caps at +200

[modern walrus]

So 400 total?

Should I max out BCLK? It's at 100 max value is 118

I'm taking shots in the dark here let me cool my jets real quick

Motherboard ignores the increased values. Still capped 95/140

[zenith palm]

I mean the boost clock override maxes out at +200mhz afaik

Idk bout bclk, haven't touched it before

[modern walrus]

I have it set to +200

I just can't figure out why it's ignoring my VRM settings

[zenith palm]

Wdym your vrm settings?

[modern walrus]

I mean like EDC, etc

It ignores that I went beyond the motherboard values

& enforces the motherboard values anyway

I wonder if I have to manually configure platform thermal throttling

I think I have ppt too low is the problem

[zenith palm]

Hmm weird

[modern walrus]

It was hitting 100% ppt

It was set at 180

[zenith palm]

Mb, run with auto and see what values you get then pop them in and up

[modern walrus]

okay I'm gonna do that now

Same limitations

This motherboard is so wack. It's like there's doubles of every setting (except these of course)

Wait a minute

I think it does have doubles of those

That has to be the issue

Found them

When I try to change max cpu boost in AMD's settings, highest is +200. Here in the ASUS version, it's just blank

But there's no curve optimizer

I wonder if I tell it +300 then change AMD's to auto if it'll do +300

Now we're in business

[modern walrus]

Problem now is I'm getting 1.23-1.24v under load

The correct voltage reading is "CPU Core Voltage (SV12 TFN)" right?

[lavish tundra]

thats fine

[modern walrus]

Under load?

I thought Fitz said over 1.2 was degradation on 5900x?

[lavish tundra]

I dont think he would say that...

zen3 is good till 1.3v iirc

[modern walrus]

Really? Maybe I misread

I'm seeing 1.263 with all cores at 44.9 rn

On CB R23. 200

Core zero is at - 30 magnitude now but I'm not crashing.

& I'm +200 wth

I wonder if I increase amps more if it'll lower voltage at all or just increase the clock speed

[south sky]

No it isn't

[lavish tundra]

Oh fr?

its 1.25 as well?

[modern walrus]

Voltage seems much more closely tied to TDC than EDC at least on CB...

But so does ratio

Should I not mess with LLC when doing PBO?

[clever epoch]

Leave LLC on Auto. I haven't used a Strix but I have the C8Hero and the PBO settings for that are under Advanced > AMD Overclocking

That's where you should be working. Don't touch BCLK, you can slap a +200 frequency override, and that's where you should put in your PBO limits

I'm running my 5900X at 220PPT/130TDC/140EDC which are fairly conservative

[clever epoch]

Go into BIOS. Look for AMD Overclocking. Change PBO to "On"

[grim shore]

Oh no i didn’t purposely oc to 5.8 I did set it to XmpII and tweaked some settings the speed and percentages varied quite a bit and it ran for about an hour I put it under some load, speed and usage jumped got hot and froze I have it set to auto off if over 100c also so I just did ai over clocking

It runs Minecraft pretty good 12 chunks 45-120 fps and valorant low to medium settings 60-120 FPS also I should note I don’t have a gpu and I’ve been trying to stream Minecraft lmao

[zenith palm]

Auto or "ai" overclocking is generally 💩💩 btw lol, I'd say that hurt your chip sightly depending on how you long before it froze

[grim shore]

Yeah

It say something like “turbo avx 5200mh” and “avx 4800mh” I have turbo on but no speeds above 4.8 ghz

[grim shore]

i did it

temps 40c 80 max

but

as always

theres a catch

its not going below 5.18 ghz

ok so it seems the voltage is playing a large role in my speeds

if i drop the offset voltage by 0.008 max speeds 4.9ghz

interesting

i put it back to 0.0000v offset max still 4.9

cpu cor voltage 1.47 avg

when it was over 5ghz voltage over 1.6

imma just leave it

[proven canopy]

1.6?

[tall pelican]

smells like vid readings

okay, so it sounds like you're doing a few things wrong:

1. that turbo speed that you're reading sounds like you're reading from the SP predictions
2. 5.18ghz is task manager reading frequencies, not an actual sensor reading
3. you're overclocking just by changing the frequency and not manually setting voltage accordingly (which is what you should be doing)

[modern walrus]

What's your voltage look like under load? Does having TDC & EDC close keep voltage down better? What's your core ratio under load?

[clever epoch]

I haven't tested with the 5900x much, but a 5950x I had earlier was about 1.2V at 4600 all core with similar (scaled) PBO settings

TDC should be less than or equal to EDC, otherwise only EDC matters

[modern walrus]

I have 46.7/all core under load but the darn thing is getting like 1.24v

ah okay

[clever epoch]

TDC is your "sustained" current, EDC is peak current

[modern walrus]

once I'm over 110a my voltage starts getting way past 1.2

isn't 1.2 my goal?

[clever epoch]

PBO does its thing. Keep in mind that the SMU (the part of the processor that handles boosting) has access to much more information than is visible to us as users.

There's temperature probes in every core, and it has an internal model fed by the temperature probes that it follows to mitigate hotspots

I haven't really benched the 5900x that much, I became quite busy IRL by the time I got it, and had spent a bunch of time with a (actually 2) 5800X's prior, and a 5950x

[modern walrus]

So PBO won't damage the processor automatically?

[grim shore]

Thank you for the tips, I made sure to change the physical pin on my mother board for over volting, I was changing some of the voltage settings i just didnt want to push it too much as my pc started to show some in stability.

[clever epoch]

Yeah, as long as you don't start messing with the throttle limits

and I'm using the loose definition of "damage", as simply the act of having the CPU on causes "damage"; degradation is always happening while a processor is powered on.

But using the looser definition of "degradation" in that you probably won't see a performance impact after ~5 years

[modern walrus]

ya I'm not touching throttle limits. If I did I probably drop it to 85 or something anyway

[sterile flame]

you would have to change scalar

1-2x is safe generally

[remote portal]

anyone got a good 5900x OC
on the b550 taichi

[sudden torrent]

Like an overclocking profile? Every chip is going to be different

[remote portal]

ya

but i need some things to throw at it

i need to get my voltage down

[sterile flame]

just turn on PBO, tweak the curve optimizer

[left blade]

Klutch#5130 has been warned

Reason: Bad word usage

[remote portal]

PBO is bad

[sterile flame]

not on zen 3

[remote portal]

runs my cpu at 90c if it can

thats on a 360mm in a o11 d

[sudden torrent]

PBO is the best way to max out the cpu, that chip is fine up to 105

[sterile flame]

must be a pretty bad 360mm

[remote portal]

nzxt

[sterile flame]

i had to put 200W into my 5900x on a 240mm to get to 90c

[remote portal]

but whats your oc

plus i can lower the voltage on a manual oc

[sterile flame]

i had a bad bin

4.6

[remote portal]

voltage

[sterile flame]

don't remember

[remote portal]

u g h

[sterile flame]

my voltage would get you high as crap

https://cdn.discordapp.com/attachments/610602203849162772/830366966257680384/unknown.png

this is 1.35

go 1.3 or lwoer

no higher

[proven canopy]

cursed

[remote portal]

got it

[modern walrus]

What does scaler do?

https://www.hisevilness.com/articles/technology/amd-overclocking-terminology-faq.html?showall=1#:~:text=What is VDDCR SOC Switching Frequency%3F Switching frequency,the heat output but lower the transient response.

This guy know what he's talking about? lol

[sudden torrent]

I think the numerous misspellings speak for themselves

[modern walrus]

oh I didn't notice

oh crap that says save not safe

tf is a save voltage?

[lavish tundra]

I dont know who this "community" is thats recommending these numbers but they are not true

[sudden torrent]

Not even close

[lavish tundra]

This "community" is full of knuckleheads

[modern walrus]

save =/= safe? lol

[lavish tundra]

No safe = safe

that list = unsafe

[sudden torrent]

Unless they mean idle voltage, in which case it's completely useless

[lavish tundra]

Either way they did it wrong

idle on stock goes past 1.4v

[modern walrus]

Are those actually AMD's recommendations?

[sudden torrent]

No

[lavish tundra]

they didnt even do those right

[modern walrus]

Where would one find AMD's recommendations?

[lavish tundra]

if thats what they were trying to do that is

[sudden torrent]

There was no official "max safe" voltage for zen (1000)

[south sky]

or for any cpu really

[modern walrus]

I thought everyone was asleep

anyway, it sucks safe is 1.2 but I gotta live with it I guess

[south sky]

On Intel there are the recommendations from their resident morons, the OC lab.

It's better to just go off current these days

[modern walrus]

how much is too much current?

[lavish tundra]

I mean overclocking is technically against warranty so they arent gonna help you void your warranty

[south sky]

see what you get with PBO and that stuff and if you don't go over it you are completely safe

[modern walrus]

Is 105.804 a lot?

[sudden torrent]

Depends on how long the load is sustained

[lavish tundra]

FIT testing right?

[modern walrus]

uh-oh 3 people typing I facked up

[south sky]

Then there is Intel's OC lab
"1.4V is safe for all 14nm cpus"

[sudden torrent]

Which CPU exactly? That's within spec for many

[south sky]

[modern walrus]

5900x

[lavish tundra]

Lmao

[sudden torrent]

105W Package Power is within official spec for the 5900X, I've seen up to 200W be fine

[modern walrus]

oh that was amps. watts was 164.614

problem is voltage hits 1.231

max temp was 70.5 so I'm pretty good on that front

[sudden torrent]

Oh, well that's still fine. My 5800X can hit that.
1.23v isn't bad

1.3v is probably as high as I'd go under load

[south sky]

if voltage hits 1.23 with pbo and stuff it's fine

[modern walrus]

hmm and are PBO curve offsets the best way to up ratios?

I miss all core OCs lol

[lavish tundra]

Leave everything auto EXCEPT PBO that you want on

then start prime95 on smallffts

[sudden torrent]

Yep. I had a lot of success doing curve optimizer. I can do over 5GHz all core.

[lavish tundra]

and monitor voltage through hwinfo for like 20-30 mins

see where the voltage settles

[south sky]

PBO2 fully tuned is the way to go

[lavish tundra]

thats whats safe for ur cpu

[south sky]

how tf

[modern walrus]

I got it decent but it wasn't getting over 45.1 in CB so I upped the PPT, TDC, EDC because motherboard had it locked at 180/140/95

[sudden torrent]

Good bin I guess but yeah, even 4.9 should be doable

[south sky]

5ghz all core under what load?

what is the voltage for that

[modern walrus]

then voltage started going bananas like 1.32 or something

so I'm redo-ing offsets

[sudden torrent]

Cinebench r23, around 1.28v or so

[south sky]

jeez

[lavish tundra]

@modern walrus

do that

to see what voltage is safe for you

then set that

[modern walrus]

the only problem with that is the motherboard will keep the PPT, etc. too low to go past 45.1

[lavish tundra]

and go from there

[lavish tundra]

You can set the power limits and stuff in PBO

[sudden torrent]

Well yeah you have to up the power limits of course

[lavish tundra]

^

[sudden torrent]

How far it can go depends on motherboard

[lavish tundra]

power limits all the way up

[modern walrus]

ya but my motherboard is weird man. There is this like redundancy where ASUS has almost identical options to AMD Overclocik

minus Curve optimizer

[lavish tundra]

I mean

all boards have that?

[sudden torrent]

Some boards will tap out and start throttling vrm at 140W, depends heavily on model

[modern walrus]

oh for real?

[lavish tundra]

from what I know pretty much

[modern walrus]

I'm already hitting 164+ so I don't think it'll tap out but I dunno how much higher it'll go

[lavish tundra]

just change in both

[lavish tundra]

Do what i told you

[modern walrus]

so set PBO to auto & max out EDC, etc

got it

[lavish tundra]

PBO manual

[sudden torrent]

Yeah my MSI X470 has 2 menus for overclocking, except one is mostly in hex so f that

[lavish tundra]

set it to on

raise limits

test it out using the method I said

[modern walrus]

those are the limits lol

[sudden torrent]

boost clock override is the last thing I do

[modern walrus]

alright I set that to 0

[sudden torrent]

My CPU is unstable at +200, curve optimizer first

+175 is where I'm at

[modern walrus]

and for all my values in core optimizer... set them to 0 again?

[sudden torrent]

Do you know which ones are your best cores?

[modern walrus]

yes I got them from HWiNFO

and POS Ryzen Master

which needed like an orthopedic surgeon to remove

[sudden torrent]

Good, set the best ones to -25, the rest to -10

You can play with it from there

[modern walrus]

Best is the top 4 or top how many?

[sudden torrent]

With your CPU I'd say top 4

If that's unstable top 3

Just need to play with it really

[modern walrus]

okay

ya I did hours hours hours of corecycler

to figure out 7 was crap

but ya 1,3,8,&9 are best supposedly

oh crap I'm reading the stupid 10900K chart I made best are 1,4,3,5 on 5900x

[sudden torrent]

Woops lol

[modern walrus]

gotta label my excel sheets better

[sudden torrent]

Sounds like CCD0 is epic on your chip

[modern walrus]

what do you mean? lol

0 has 5th best silicone rating supposedly

okay so now I run CB, note the voltage under load, and

and then what? lol

oh wait

you said run prime95

that's right

CB was anywhere from 1.219-1.244

okay wow this is intense. Voltage 1.212 max; CPU PPT max 200.4W; TDC/EDC max 140.482 amps;

[sudden torrent]

200W is normal when overclocking that chip

[modern walrus]

max temp freaking 85.4

[sudden torrent]

CCD0 is the first core cluster, cores 0-5

85 is fine, really. I don't worry until it's over 95.

[modern walrus]

That's what prime95 Test 4 is showing

[sudden torrent]

Only 45x? Do you still have the manual core ratio set?

[modern walrus]

No I haven't had it set since I was doing per core OC

[sudden torrent]

Hmm. Try dropping the third and fourth best cores to -15

[modern walrus]

alright I'm turning off the prime95 microwave then I'll check it out

but the best I got messing with curves was 46 on every core at 1.237v

with +200 but isn't that for single core boost anyway?

CPU Core ratio says auto

that was with scalar at 2x btw

[sudden torrent]

Think of it like this:
Every core has its own table it uses to find the boost clock it should be at. If you shift the graph to the left by 200 MHz, every boost clock is 200 MHz higher.

[modern walrus]

should I disable ASUS Performance Enhancement?

[sudden torrent]

I don't have any experience with Asus so I couldn't tell you

[modern walrus]

so how am I cooking at 84 degrees & only getting 45/core?

ya 200mHz doesn't really do much

[sudden torrent]

The problem is heat, the closer it gets to throttle limits on hot spot temps the lower the boost clocks it chooses

[tall pelican]

Now to see if 4500 uF of caps help

[sudden torrent]

You have to balance it, that's why +200MHz isn't always best

[modern walrus]

but I noticed with like CB v. Prime95, it'll run slightly higher clocks on CB--probably becasue of the heat?

CB gets me to like under 70

[sudden torrent]

P95 is an unrealistic power draw

Cinebench is a little closer to "gaming load"

[modern walrus]

It just seems really mediocre

especially when I do 3dmark & end up like 400 points below average

bin?

[sudden torrent]

You can do +100 offset (or even +150 if you have good bin) while you work on the curve optimizer, but I need to get some sleep

[modern walrus]

I had it at +200 before

gn thanks

[sudden torrent]

Yeah but that was before you played with CO right?

[modern walrus]

noo

that was after hoursss

of playing with CO

[sudden torrent]

Oh I didn't realize you'd already done a lot of this

[modern walrus]

but with EDC/TDC/PPT limited by motherboard settings

ya sorry I come & go in this chat. I'm not quite the random noobs but I'm nowhere near the mainstays

[sudden torrent]

Yeah once you eliminate the power limits it comes down to 2 things: Binning and Motherboard VRM

[modern walrus]

there's a setting to allow VRM up to 130%. I left at 110% cuz I dunno what I'm doing entirely

Maybe up those to full 130?

[sudden torrent]

If you have good airflow over your VRM then you can do 130%

[modern walrus]

that's the clusters left & top of the CPU right? lol I might throw another 70mm by them

I don't even know where VRM temp readout is

but my bad don't let me keep you up

[modern walrus]

all 3 make up vrm?

[sudden torrent]

VRMs don't usually have their own temp readout, but the board knows what the temp is

[modern walrus]

uhh I wish I'd have known better. I would've gotten another Aorus board

[sudden torrent]

Which board do you have exactly? Some of them don't like to have too much current, even with good airflow.

[modern walrus]

Strix b550-f gaming wifi

I think I remember reading overall current limit was 700amps; 600 for cpu

[sudden torrent]

That one should be ok around 200W current (package power in hwinfo) and a little airflow over them will do wonders

[modern walrus]

ya it looks like it's maxing at 202 rn. I tried it again with 130% set

but only up to 45.1 now on P95

[sudden torrent]

See how it does on cinebench, if it's at least 4.7 I'd call it a win for now

At any rate I'm about to pass out, g'night

[modern walrus]

lousy 45.5 lol

gn

only 156 watts tho so heck I dunno

[sudden torrent]

You could try a voltage offset, I do -0.0125v on mine and it makes a big difference

[modern walrus]

alright thanks. I'm gonna go figure out where that setting is

gn

[dull ginkgo]

By the way, it looks like we might be starting to develop some of those more 'exotic' cooling methods, which will be quite interesting for the future.
https://www.tomshardware.com/news/tsmc-exploring-on-chip-semiconductor-integrated-watercooling

[modern walrus]

sounds expensive

they should just replace the copper & gold inside them with water. It's conductive lol

well technically 100% pure water is not

[modern walrus]

I cannot break out of 46 ratio; can't even get over 46.4. My bin sucks I guess

If I increase EDC, amps go up to whatever the max is, voltage will try to tick up with it but core ratio will barely move .1

oh my Lord it's a miracle. I put every offset I could at -30 & got 47. I doubt there's any headroom left after this

heck.crash

[modern walrus]

What's going on here?

6 cores at 4600 & 6 at 4200

[tall pelican]

Ccd1 and ccd2

[modern walrus]

is there normally a difference like that?

[tall pelican]

yes

cant really see it in effectives, since it was a short time frame, but max frequencies show it

[clever epoch]

ye microfluidics is an emerging topic that looks pretty interesting. We're gonna need it, really, considering that we're already running into head density issues

[dull ginkgo]

Yea, I wonder if they can rush development for use in more current chips that aren't vertically stacked yet.

[cold jolt]

me when theres too much pressure in my cooling loop and my cpu literally effing explodes

[dull ginkgo]

Lol

[dull ginkgo]

Actually, do you know if any companies have any timelines or forecasts for when they'll start stacking logical cores?

[clever epoch]

Certainly no official announcements, and I wouldn't expect to see any until we get proven performance.

We'll likely see patents first as a hint it's coming, I wouldn't expect it any sooner than on the order of 5+ years

[dull ginkgo]

Fair

I haven't seen much news on Intel with their odi or anything, at least not as much as TSMC with their 3dfabric and all, was wondering if they're doing much with that

[clever epoch]

Direct chip cooling is very much in the exploration phase right now I would say. I wouldn't expect it in mainstream until at least a year or two after we see some server chips using it

[orchid flame]

this is dumb af btw

there are much better ways to do this

[dull ginkgo]

like?

[orchid flame]

the simplest would be integrated microheatpipes

since heatpipes are about 100x more conductive than pure copper alone

bring those to the top, and cool it and you're done

[dull ginkgo]

I mean, this is pretty small for a heatpipe here..?

[orchid flame]

and because of the pressured nature of heatpipes you dont need to worry about speed or evenness of cooling

[orchid flame]

not really

it can very much be done

TSV-based 3ds often have several tenths of milimeters of area between layers

thats plenty to fit a microheatpipe

https://www.furukawa.co.jp/en/product/catalogue/pdf/microhp_e.pdf

[dull ginkgo]

I mean, from what TSMC says, directly cooling the silicon with water seems to work better than to add a layer of interface and then dissipating the heat through fins or whatnot

[orchid flame]

this is for 2d and 2.5d solutions

3DS requires different cooling entirely

[dull ginkgo]

I believe it's supposed to be a pillar setup that can expand upwards, but not too sure

[orchid flame]

this solution would introduce channels to cool the coating

I'm proposing a solution for 3ds / 3d die stacks

well

i'm nort the one that came up with it

this solution would have higher logevity though and would mae sense in facilitating 3d die stacks

[dull ginkgo]

Maybe, but I believe there would be a performance benefit from direct water cooling

[orchid flame]

there would also be one here

since you could more easily do direct mounting to the heatpipes

and you cut down on overall distance between max temperature point and the cooler

so coldplate - tim - coldplate

but the coldplate would instead be ultrafine microheatpipes

(on the chip side)

[dull ginkgo]

I mean, I believe what TSMC is doing is basically microfins on the chip

[orchid flame]

yeah since thats honestly an easy solution

coating has always been the cheapest phase

changing it a bit to account for this wont change much for cost

the concern for me is regarding future 3ds implementation

TSMC already seems to be very fond of playing it extremely safe

example? N3 is finfet

[dull ginkgo]

As instead of using gaafet?

[orchid flame]

yes

as SS is doing

well, samsung is doing an extended gate all around setup

3GAE

still.

dont know if I'm allowed to share this here

might DM it to you

[dull ginkgo]

alright

[orchid flame]

https://images-ext-2.discordapp.net/external/C8ztvWOjDet079PCbAoXRqdXFSSs53Njb3wTtYgfybM/https/fuse.wikichip.org/wp-content/uploads/2021/07/gaa-sides-slidew.png

https://images-ext-2.discordapp.net/external/n-54PkEXJY3GFNQQ8KeIcyiQx9KcFUlqxDZsHcbaFjY/https/fuse.wikichip.org/wp-content/uploads/2021/07/ss-3nm-4nm-compw.png

https://images-ext-1.discordapp.net/external/VRRjINnhYKm1JYvi1ujkOjYDD9DG-QbiiXQ48xdvYcE/https/fuse.wikichip.org/wp-content/uploads/2021/07/ss-3nm-sram-pn-ratio-configw.png

[dull ginkgo]

Then the reasoning behind TSMC not doing GAA would be just lack of experience?

hm

The stuff I've been reading might've been way too old

just realized one of the slides I was using for info was from 2007 and about 45nm planar fets

[orchid flame]

no

its cost

[south sky]

it's also just not necessary with their design, performance improvements would be minor for them

[orchid flame]

it provides clear benefits to freq scaling

and decreases losses

[south sky]

past a certain point

clearly their design doesn't pass that point so the benefits would be minor

[dull ginkgo]

huh, hwinfo got a whea logger now?

[tall pelican]

always had one

[dull ginkgo]

Never noticed it

[sudden torrent]

It's just aaaaalllllll the way at the bottom

[modern walrus]

okay I'm very confused. All I have set is PBO. I've been messing with curve optimizer & well... long story short PBO is putting me at 1.3v+ on CB regardless of how low/high my offsets are

that's under load running CB

Presently, highest magnitude is -13

highest magnitude -18

is it because I increased EDC/TDC to 150A? usually motherboard limits to 140

[red junco]

hey guys... i need alot of assistance overclocking my gpu, its an nvidia geforce 650 ti boost made by some brand called "GALAXY". i really dont know what to do and would like to know if someone could walk me through it. i need to overclock because my gpu is always on 100% usage in every game :/

[jade pulsar]

I think it good to have 100% gpu in games tho, you dont want less than your full gpu when playing games lol

[proven canopy]

no

[jade pulsar]

Not typically anyway

[proven canopy]

input lag

[jade pulsar]

When your at that point then ya there is input to consider but if you are on a 650 ti realistically you wont really be able to go that far in modern games

[south sky]

https://gskill.com/product/165/299/1620978762/F4-4600C20D-64GTRG this looks like AJR at 1.55V

[zenith palm]

Download msi afterburner, and heaven benchmark and bring the power slider as far as possible then start increasing the values slightly like try +50 core and +300 mem originally, wait a sec, if you don't see any artifacts or crash, up the core in 25s until it crashes then bring it back down 25, then the same thing with the memory basically

[vapid ridge]

How do I over clock im in the bios rn what do I do

[dusty falcon]

If anyone here has a 5900x, what do you get for your average all-core clock while running cinebench r23?

I only manage to get 4.26 ghz personally.

[zenith palm]

What cpu, motherboard , cooler and ram do you have

[zenith palm]

I'd recommend prime 95 or something, cinebench can be a bit variable, also what cooler do you have ? Do you have pbo enabled ?and are you reading the frequency from task manager? If so use hwinfo 64 or something else instead task manager isn't great at reporting frequency

[dusty falcon]

Ek 240 DRGB, I have -10 offset on curve optimizer and a boost override of 100mhz. Temps are about 70c when running multicore tests. The frequency is from OCCT.

[zenith palm]

What are your other pbo settings?

[dusty falcon]

Auto

Everything at default

All done in bios

Anything higher than -10 and I crash constantly as well.

[zenith palm]

@dusty falcon i would try something along these lines

[dusty falcon]

It get SUPER hot when I try and adjust the power limits.

Even 165 115 140 causes 80c under all core load.

And I only get a 100mhz oc.

[zenith palm]

80c is fine, anything under 90c is still fine, also make sure your pump is connected to the pump header or if it's connected to a fan header make sure that header is set to 100% and you can always adjust your fan curve

It can go 95c even but 90c is where i would get uncomfortable

[dusty falcon]

ok

[tall pelican]

Its down to silicon quality, cooling, and somewhat motherboard

[red junco]

i tried downloading msi after burner and when i try launching it, this hapens

[wheat lintel]

I’ve been getting around 4.6-4.7

[red junco]

in my case... its expected, but what i meant is it can barely run games on low settings and on newer titles it just gets like 20 fps on all low settings like warzone, also @proven canopy is right because when i am playing games that are using all my gpu and still asking for more, i feel alot of input lag. i bought an overkill cpu for my setup to hopefully combat this(ryzen 7 3700x watercooled with an aio) but it seems to do nothing on newer games

[zenith palm]

I've never seen that, mb google the problem

[sudden torrent]

Lag is often memory related on Ryzen from what I've seen

[jade pulsar]

Ye i forgot that part, but it’s your decision at that point, either aim for higher fps but high input lag or same or possibly lower fps and less input lag, personally in the games i play high refresh could be more beneficial up to a point depends how bad the lag is. You have a 3700x with a 650ti??!!!

[sudden torrent]

You have a 3700x with a 650ti??!!!
I'm guessing that wasn't their ideal choice

[jade pulsar]

Lol

[red junco]

yeah lmao not at all

[red junco]

sadly 😦 it was pulled from an old system

[jade pulsar]

While I’m here does anyone know the name of the setting to force use igpu in rog intel bios

[tall pelican]

make igpu primary display output, or just have it on while using a dgpu?

[jade pulsar]

Primary for boot os

[tall pelican]

should be something along the lines of IO ports or periphs in settings tab of bios

[jade pulsar]

Oki i check when i get home

[wheat lintel]

I just ordered what I think is all the stuff I need to delid my 6700k

How should the application of LM look?

Or anyone know any good videos?

[tall pelican]

enough to coat, but not enough that you hold the chip sideways and you can see it forming a drop

[wheat lintel]

Ahh ok makes sense

[faint tangle]

I'm too tired I read that as 6700xt and was so confused

[wheat lintel]

XD

[short blade]

are there any valid alternatives to stability test an 8700K? vrms on this board can't handle sustained small ffts avx2

throttles after 30 minutes at 265W cpu power

[proven canopy]

You're not thermal throttling at 265w? Delidded?

[short blade]

yeah, cpu temp is fine at 265w sustained

delidded + lm above and below ihs + 240mm aio

85-90c all core

it's the vrm hitting 110c and throttling it to base clock

once it goes back to 3.7ghz it's no longer a cpu stability test

this is just going to be a gaming system so i feel like hammering it with sustained 265w power virus isn't necessary?

tried realbench but it seems too light

currently trying prime95 small ffts SSE and that can run without vrm throttling

[proven canopy]

Point a fan at the vrm

[short blade]

there are 2 exhaust fans almost directly touching the top vrm heatsink

running 100% speed

and i have the case open

[proven canopy]

https://www.ebay.com/itm/162723781023

You probably don't need all 10 fingers anyway

[short blade]

so the only choice is to put a server fan on the vrm heatsinks?

that fan almost as expensive as the mobo

[proven canopy]

What's the current rating / cfm spec on the fans you're using

I posted that as a joke - but something like a .3 amp fan might be enough.

[short blade]

75 cfm

[proven canopy]

oh, that's pretty high

How many amps

[short blade]

0.17A +/- 20%

[proven canopy]

I'm surprised .17A has a spec of 75 cfm

[short blade]

meh cfm spec is kind of ? anyways

[proven canopy]

Maybe re-pad the vrm heatsinks?

Probably not worth spending a ton on, but some Arctic thermal pads are probably miles better than whatever's on there.

[short blade]

doesn't really seem worth it

the vrms only throttle when pushing 265 watts

it's a gaming flip

it's never going to be pushing a load like that irl

[tall pelican]

I personally only use small for 5 mins

For game stable, sse/avx for 5 mins, avx2 for 1 min is stable in 99% of games

Games that use avx2 use it at load screens, and only for a few seconds because of how much faster avx2 is, than sse

[short blade]

it passed 30 minutes before hitting vrm throttle

didn't crash but time at 3.7ghz doesn't really count

30 minutes at avx2 that is

i kind of want to be a little stricter than that for a flip though

can definitely afford to be less strict with "stable" for a personal build

i've decided to try per-core oc approach instead

now testing 5ghz for 1-3 cores, 4.9 for 4-5, 4.8 for 6 core load, -2 avx offset, 1.4v llc second highest (drops to 1.328v max load)

this is probably the worst bin 8700k in existence so yeah the -2 avx offset seems necessary

5ghz 1.4v max llc (drops to 1.376) is not stable even with small ffts SSE

[proven canopy]

p95 or occt

[tall pelican]

P95

Oh, and dont bother with avx offset, nearly everything uses avx today

[short blade]

i am completely aware

but this 8700k is seriously the worst one ever made

[tall pelican]

Which board?

[short blade]

4.7ghz all core is not stable

have to go 4.6

z370 extreme4

this is the same 8700k that won't boot with ram above 2666

and requires 1.2 vccio 1.25 vccsa to hit 2666

[tall pelican]

Yeah asrock has some stupid weak llc settings

[short blade]

ehh it seems ok there's just no completely flat option

[tall pelican]

Nah, level 1 llc (should be max) is equivalent to asus 5, gigabyte high

Im guessing no vr vout?

[short blade]

idk about those all i know is that level 1 goes from 1.4v bios setting to 1.376 vcore

there is a vr vout reading in hwinfo yeah

is that more accurate than vcore?

[tall pelican]

Go off that, small avx2 should drop ~.1v

Vr vout is die sense

[short blade]

okay gotcha

[tall pelican]

Vcore is superio guessing

[short blade]

that is a lot of droop yeah

thats llc2

that VR VIN is a lot lower than 12v, is that okay?

this is running small ffts SSE

been going for 31 minutes

ambient temp around 25C

[tall pelican]

That's fine

Vrvin can go to like 11v and be perfectly fine

Also, you can try putting a fan on the vrm for testing purposes

[short blade]

this is what i'm getting on my ryzen system

but it's not running a load

i assume it droops under load

[tall pelican]

On ryzen, svi2 tfn is the on cpu sense, but the top one seems close

Bottom is soc vrm

[short blade]

yeah i'm familiar with svi2 tfn

just less familiar with intel

this is the PC right now, top rad fans are touching vrm heatsink

[short blade]

well

4.6ghz all core crashed in prime95

at 1.4v llc2

[ionic spoke]

can someone help me using ctr 2.1

[south sky]

Just don't

That's the best advice you can be given regarding it

[ionic spoke]

why tho

what should I do instead

[south sky]

It's unsafe and doesn't give you an optimised oc

What CPU do you have

[ionic spoke]

5800x

[ionic spoke]

^

[south sky]

Use PBO2 and the curve optimiser functionality

https://youtu.be/dU5qLJqTSAc

[ionic spoke]

thx

[proven canopy]

Great vid

[short blade]

yes avoid craptuner for ryzen 2.1 before it degrades your cpu

or don't take my word for it, it happened to @lavish tundra

[lavish tundra]

GRRRRR I HATE IT

Can confirm my cpu is degraded after ctr

[clever epoch]

What CPU out of curiosity? I know it degraded somebody's 4650G, and apparently outright killed a 3950X of somebody else

[lavish tundra]

3600x

[clever epoch]

Neat. I might make a spreadsheet of this since I know some people who insist on using CTR

Even the CPU "rating" is bunk IMO

Can go from gold to bronze by changing LLC apparently...

[sudden torrent]

TBF changing LLC changes load voltage

[lavish tundra]

Like the degradation on my chip is real

Went from being able to run 4.475-4.5ghz all core at 1.25V to struggling with 4.35ghz at 1.25V

[ionic spoke]

Weird

Btw watched that vid and oh boy it takes alot of time

[sudden torrent]

Yep. That's how you know you're doing it right.

Just wait 'til you do memory overclocking and have to wait 1 hour between tweaks to see if each small setting is stable

[south sky]

what did it run?

and how long?

[lavish tundra]

4.375 @ 1.35V

[lavish tundra]

on and off for a month (more on than off)

since CTR said it was stable

I was even more of a dumdum back then

[south sky]

makes sense

1usmus the dev thinks 1.55v is fine for zen 2 and 3

not sure if its a troll or not

but 1usmus is a malicious troll in other ways

[sudden torrent]

Unintentionally malicious is the worst kind
"Oh it'll be fine don't worry" "Oops, it exploded in your face? Your problem."

[south sky]

nah intentionally

Like when the zen 2 xt chips came out

[short blade]

1.55v fine for zen 3?

[south sky]

hell nah

[short blade]

at 1.55v im sure your house will burn down from opening minesweeper

[south sky]

1usmus claimed the zen 2 xt chips all did 2000+ fclk, based off their testing, then just said lol when people wasted their money buying them for no gain, after encouraging them to buy them

[short blade]

LOL

i wouldn't be surprised if 1.55v caused degradation just from booting

[short blade]

hahahahahahahaha that's cringe

linus needs to take back the endorsement of CTR

he has a huge audience that includes many casuals that may just take it at face value

[south sky]

1.25v might not be safe and this nut case is saying that 1.55 is practically fine

[sudden torrent]

"Read my guide, trust me not them. See my guide for references."

[short blade]

I believe ctr always includes one 1.35v profile

linus's video about CTR has 1.6 million views

if 5% of viewers downloaded it and 60% of those chose the highest performance profile offered

that's 48,404 cpus degraded

[lavish tundra]

i was one of em

[short blade]

you saw it from linus?

[lavish tundra]

yep

I no longer trust LTT as much

I watch em for entertainment only

[sudden torrent]

Good. Make your own decisions. Find your own facts. General life advice.

[south sky]

all of ltt's oc stuff is garbo

they recommended 1.4v on a skylake chip, 1.35 on zen 2

just yikes

[tall pelican]

trust me bro, when I release hydra (the new version of CTR), it'll be so much better

[south sky]

oh no, the saddest thing is how much he makes off scaming people

[lavish tundra]

LOL it uses all the same things as ctr

[tall pelican]

nonono trust me bro its hydra now

kill off one cpu, two take its place

[steel zephyr]

For a while CTR has used 1050mv as the base voltage and only uses 1.3v for the single/dual core OC profile.

[south sky]

1usmus makes like 50K a year off patreon which is so sad

[short blade]

thanks for confirming it's still using unsafe voltages

[steel zephyr]

How often do you get sustained single or dual core workloads?

[short blade]

considering I ran sustained single core workload for 12 days straight

I dunno man

[steel zephyr]

What runs 1 core at max speed 100% load for 12 days straight?

[short blade]

corecycler

xd

[steel zephyr]

So basically no normal workload.

[short blade]

doesn't matter. cpu degradation can happen in 10 minutes

stop drinking 1usmus koolaid

[south sky]

cinebench

[short blade]

setting manual freq/voltage disables the cpu's ability to protect itself

[south sky]

you've gotta push like 1.6v+ to actually get deg with one core though

[short blade]

going on to set voltages above FIT is not safe

[sudden torrent]

Most motherboards will straight up warn you in some way if you push voltage too high, usually above 1.5v

[steel zephyr]

Proof that it happens in as little as 10 minutes at 1.3v? I've never seen anyone provide proof of CPU degradation at 1.3 volts for short/intermittent workloads.

[south sky]

it won't happen that quick

unless if temps are slightly high

(130C or so)

[short blade]

not at 1.3

degradation in 10 minutes can happen at 1.4

[south sky]

then there is my board

1.4v probs not

[short blade]

but it doesn't matter what specific numbers are being used

[short blade]

happened to me

[south sky]

damn

Some people ran 1.5V for benching ambient for a few hours and were fine

[short blade]

10 minutes of cinebench, superpi, and prime95

had to increase 3 cores by 1 offset

[south sky]

Yeah, and 1usmus says 1.425~1.45 won't deg at all

[short blade]

there is something inherently wrong with a software marketed to casuals that disables CPU protections and then applies voltages above safe levels

[south sky]

What was the last chip 1.4V was actually safe on? 2500K? haha

[short blade]

if you can't see that then go ahead and keep using it

[south sky]

zen 1 did okay with around 1.4v which was crazy

[short blade]

hmm

[south sky]

zen 1 was so tolerant

could run 1.375 ish on those things

[steel zephyr]

I ran a 3950X with a BIOS OC of 1.3v all core and 4.35ghz OC for a year with no issues. I guess I got a golden chip. lol

[short blade]

I've seen a lot of people say 1.4v is the upper limit from haswell up to comet lake as long as you can cool it

[south sky]

Let me guess you set 1.3V

[south sky]

Haha no

[steel zephyr]

set 1.3v in the BIOS as the fixed CPU voltage as I said above.

[south sky]

Yeah set voltage means nothing, don't refer to it because it's misleading

would have dropped probably sub 1.2 under load

[short blade]

what would it be then? 1.35?

[south sky]

cml can deg at 1.25v

with aggressive settings, high load and not great cooling

[steel zephyr]

The various monitoring software showed CPU voltage at about 1.3 volts +/- a little bit

[south sky]

same with skl, kbl, cfl and cflr

and skl x, ccl x

[short blade]

1.25?

[south sky]

yes 1.25

[short blade]

what temps are we talking about

[south sky]

pretty high

depends on the chip though

[short blade]

8700k at 1.4V LLC1 (1.376V VR VOUT at max load), delidded with liquid metal above and below ihs, 240mm aio, 80C all core maximum

degradation territory?

[south sky]

Yes

for sure

that's insane

[short blade]

it's already the worst 8700K ever, maybe previous owner degraded it

[south sky]

what is the power consumption on that?

250W?

[short blade]

235-265

[south sky]

yeah that's about the same per core as my 9900K at 1.377

I get 340W

[short blade]

I pushed a 10700K to 335W and the 360mm aio gave up

[south sky]

that poor 8700k

[short blade]

the 8700K stays pretty cool though

[south sky]

192A though

138A chip

[short blade]

if I have to lower voltage at all I'd be better off running stock

[south sky]

like 40% over spec

[short blade]

straight up

[south sky]

what clocks?

[short blade]

it can only do 4.5ghz all core stable

[south sky]

What

What board

most 8700K's do like 5ghz 1.25v

maybe you degraded it

[short blade]

and will not boot ram above 2666 with 1.25 vccsa/vccio

[south sky]

also that is one hella leaky chip

[short blade]

it was like that as soon as I bought it

bought it used 3 months ago

[south sky]

sounds degraded as hell

[short blade]

been powered off for most of the time I've had it

[south sky]

did you get it delidded?

[short blade]

no, I delidded it

[south sky]

Oh

was it always that shit?

is it stable stock?

[short blade]

if it was degraded it shouldn't be stable at stock no?

it is stable stock yes

just the worst bin ever I think

[south sky]

these chips have massive headroom

something like 150mv

my board undervolts my 9900K by 100mv stock

and I can still go another 70mv

[short blade]

if I load uefi defaults and run stress tests

perfectly stable

[south sky]

What board?

[short blade]

have not tried to undervolt stock

z370 extreme4

[south sky]

that makes sense

A: That board is crap
B: It doesn't undervolt stock

[short blade]

meh

price was right for a flip build

$58

[south sky]

If it isn't degraded that is the worst 8700K on the planet

I haven't heard of one not doing 4.7ghz 1.2v

[short blade]

I delidded it just so I could try delidding

4.7ghz 1.2v

meanwhile

[south sky]

mine does 5ghz 1.22v

[short blade]

this one crashes at 4.6ghz 1.4v

[south sky]

or was it 1.19v

can't remember

[short blade]

is that small ffts avx2?

[south sky]

haha no

cinebench stable

1.19 cb stable

it wasn't even a very good chip, and it wasn't delidded

[short blade]

I asked here before if there were any alternatives to small ffts avx2 that would be "good enough" for a gaming flip

4.7ghz 1.4v gets through y-cruncher consistently

[south sky]

Don't sell it with that OC

It will die

[short blade]

5.1 gets through cinebench, don't remember if it was 1.35 or 1.4

5.0 gets through 2 hours of realbench

5.3 gets through r23 single core

[south sky]

so it does 4.5 stable but can run 5.0~5.1 somewhat stable?

[short blade]

tbh

I cannot tell you if 4.5 is stable

[south sky]

that chip is definitely degraded

[short blade]

I stopped testing out of despair after 4.6 crashed

IMC seems degraded too

I tried 2 ram kits on it neither could boot xmp

have to manual oc to get something stable

the highest it could boot also changed depending on the kit

rev. e could stable 3200c12, hynix mfr can boot 3000c15 but highest stable 2666c12

[south sky]

well yeah rev.e is a lot lighter

mfr is heavy

[short blade]

1.2/1.25 vccsa/vccio, 1.45v dram voltage

[south sky]

very heavy

[short blade]

then again

I haven't tried yet if 1.2/1.2 works

1.15/1.15 fails boot

1.2/1.2 does boot

anyways I got the 8700k for suspiciously cheap price so I'm not really surprised that it's degraded

as long as it'll run stock that's ok

you think it would be best to just let it run at stock?

[south sky]

well yeah

4.5 1.37 would degrade it more

plus it's already degraded so it can handle even less

[short blade]

I could probably get at least 4.8 @ 1.25 through cinebench

but for a flip it's probably better to not go with questionable stability

anyways this is gonna be the last time I buy a used cpu

the headache is not worth the savings

what would you say is safe on haswell?

benched 4770k at 4.5ghz 1.4v was gonna do 4.4ghz 1.35 after

[south sky]

low 1.3's set with a reasonable load line is pretty good, not sure how much higher you can go

[short blade]

alright I'll keep it in mind

thanks

gonna sleep now it's 3am

good night

[south sky]

good night, sleep well

[lavish tundra]

makes sense makes sense

[clever epoch]

Yeah I know, LLC will affect your oc stability, but the fact that CTR looks at set voltage instead of load voltage is telling

[finite frigate]

Predict max clocks for my 280c

*280x

Stock clocks are 1070 and 1600

Hopefully can get it today

[jade pulsar]

I must run benchmarks on my pentium

[sterile flame]

Can someone tell me how to overclocked the uhd 630 from a 10700k

[faint tangle]

Wait

K cpus don't have integrated? Unless you mean f or kf

And idk if you can overclock integrated graphics

[zenith palm]

k is overclock, f is no igpu

[zenith palm]

you can

[finite frigate]

Download intel extreme tuning utility

[sterile flame]

Yea but I don't know where to overclock the graphics

[finite frigate]

Should be in graphics tuning or whatever

[dull ginkgo]

"I prefer to tell you when you have done something wrong, rather than doing it for you, and if I'm wrong, well, too bad for you, your computer is dead"

great quote

"I recently learned that CTR is changing its name and being renamed Hydra. Why would you do that? OCCT has been named OCCT for 18 years, I don't have anything to be ashamed of in my past, why would I change my name since it's well known? Ask yourself the question."

[proven canopy]

Because people are finally realizing that CTR is hot garbage

Dangerous* hot garbage

[finite frigate]

do you guys know what sensor afterburner is pulling the voltage number from

hwinfo is reporting 1.2 and this is 1.025

[tall pelican]

which gpu?

[finite frigate]

R9 280x direct cuii top

[tall pelican]

I'd go off hwinfo then

[finite frigate]

im thinking afterburner is reporting correctly

and hwinfo isnt

as the temps go up and down as i change the voltage etc etc

[tall pelican]

are you looking at the right sensor in hwinfo?

[finite frigate]

should be VDDC right?

[tall pelican]

Im thinking more vrm voltage out

[finite frigate]

alright ill pay attention to that rather than ab

[tall pelican]

would be like 1.056v requested, 1.001 received

[finite frigate]

actually ab is close enough i doubt itll matter too much

[finite frigate]

yeah this checks out

[finite frigate]

anyone know how to force 0db mode/0% fan mode?

ive already tried forcing the vbios to go down to 0% but that aint working

[short blade]

something tells me this is bugged

nvm, apparently negative tWTRS is actually a thing

[zenith palm]

So it's displaying negative as like 4 million?

[clever epoch]

signed number represented as unsigned causes apparent overflow

[short blade]

I know, just didn't think negative timings were actually a real thing

[short blade]

2^32 - 1

4294967295

interestingly, bios correctly labelled it as "-1"

[zenith palm]

Close enough

[finite frigate]

vrms getting mad toasty

[south sky]

It's not negative it's 1 7

You can do negative but you need to to further

[short blade]

yeah I heard someone said intel adds +2

[orchid flame]

77c is nothing

If it’s below 95c, it’s fine

[finite frigate]

yeah

[vague pine]

Welcome to bad PCs

[sudden torrent]

If the tWTR_S is negative does that mean it's so bad it's good?

[south sky]

no

it means it is so good that it's good

[modern walrus]

well guys I think my overclocks have really worked out well

New World Record right here:

I think I broke it lol

[sudden torrent]

Submit it to HWBot for infinite points lol

(until it gets reported 😛 )

[modern walrus]

Maybe it'll help them somehow

time for a little sfc action

[sudden torrent]

Such a weird number too, it's just above 16MB so not even a buffer max

[modern walrus]

The weird thing is it worked fine until the final render

My chip bin sucks I can't even get to 46 all core

[sudden torrent]

That's probably not binning, what motherboard do you have?

[modern walrus]

ASUS B550-F Gaming Wi-Fi

I can put TDC/EDC/PPT way up so that's not the problem

[sudden torrent]

Heatsinks look good physically, but it could still be the VRM is getting too hot

[modern walrus]

I have it at 115 (TDC)/160 (EDC)/PPT like 190

Is there a readout for that on HWiNFO?

[sudden torrent]

Nope, not on most boards

Lemme see reviews

[modern walrus]

I'm custom loop cooled maxing at 74.9

[sudden torrent]

Good, but core temp doesn't matter here. If the airflow over the VRM heatsinks is poor they'll overheat.

[proven canopy]

[modern walrus]

hmm

I got an extra 70mm bgears little crazy guy I could probably aim at it

3500rpm. Those little things move some darn air

[sudden torrent]

I slapped an old 72mm 5k rpm fan onto my GPU backplate, now THAT moves air. I have it turned down to 50% lol.

VRM looks good on that board so it must be something else

[modern walrus]

haha ya I have a 5500rpm fan that's so freaking loud I can't stand it

213cfm from a 120mm tho

Has to be the bin then no?

[sudden torrent]

Or the procedure you used to overclock it

[modern walrus]

PBO fun time

I'm +200

& uh

here's my attempts

[sudden torrent]

I do offset last. I'm only stable at +175 with my curve optimizer settings.

[sudden torrent]

+200 and I start regressing on performance

[modern walrus]

I'll drop it & see what happened

happens

You think I should do 150 or just dive in at 175mHz