#overclocking
1 messages · Page 7 of 1
Ohhhh yikes I’m blind
Leave mode synced
I don't see how effective this would be tho as it's more likely a measure of your infinity fabric
Also I noticed the settings “Low latency mode” and “XMP/EXPO High Bandwith Support”
Do I leave those alone?
I have zero zen 4 experience I'm not sure
Maybe wait until there's someone who has

Yeah best to just leave those alone for now then lol
Well with the voltage change it seems to be running perfectly
"perfect" I wouldn't be so sure
I would do a memory test
Download occt
Then go to the cpu test section
On the bottom left
Try cpu test large, extreme, variable
Change sse to avx2
Then go to advanced settings(?) And choose start at cycle 18
I ran testmem5 anta777 extreme for 10 minutes and it was working when before it crashed instantly, I'll do that as well to see
You can do that test instead tbh
I ran it for a while and it was good but I'll do this as well just to be sure
Supposedly not as accurate as occt but if you know how to run that one then I would do that
Run it for at least a couple hours
Yeah that's how we figured out it was memory errors, using TM5
It would blue screen the whole system
For AM5? Yeah that's great
It’s sitting around 80 rn
AMD decided running at 95C was normal and expected lol
Typical
I got a beefy cooler and good case at least
Gonna run this for an hour and hope I’m good then 😭
Running at around 5950 mhz CL 36
Beautiful 🥲
I would do more than an hour
OCCT has a "power test" that loads the whole system
Yeah you won't be fully loading the cpu
And you can run each test individually at the same time
But the ram and imc? Ye loaded
It’s saying 100% 😭
Ye but it's not a full load
Maybe I'm not making a lot of sense rn
Yes it's 100% utility but it's not full tilt
I’m not the smartest with stress tests so I’ll assume you’re right
Stress test loads use different instructions sets to try to trigger problems
That makes sense
Full cpu tilt would be occt small instead of large
Or prime 95
(P95 isn't too hard on zen tho)
OCCT CPU test would use the AVX instruction set normally unless you tell it otherwise
It doesn't need AVX instructions to move data around the memory
I’ll open a chrome tab hopefully that adds some more load 🙏
No don't: 💀
It is saying physical memory load is only 20%
Is that ram or the memory on the cpu?
Yes physical memory is more related to cpu iirc
You'd want to tell it to use 80-90% before a memory test starts
I might’ve uhhh not changed it lol
So memory tab at 90?
Alright now I’ll run this for a while lol

It was at 80% before weirdly enough
But clearly not because it was way lower on the actual test
Pray for my pc 
It says cpu usage 100% but it’s sitting at 43 C lol
Weird zen 4 moment
Yep that's because it's fully utilized cpu power but it's not using a stressful set of instructions to do so
Not like games or a CPU test would do
Interesting to know
How will I know if there’s a problem?
I’m assuming if I blue screen
idk, something else seems off about that
Those cores should be running hot, probably looking at the wrong temp
Agree
avx, avx2, avx512 - even amd's avx512 will draw the most power
Maybe it's not loaded the test properly?
It'll report errors in the log on the left
@sterile ridge Just post screenshots
And the green check will turn to a red x
The memory test isn't as power hungry on the cpu
Is sse the default for the mem test? @sudden torrent I don't use occt
once i get it all sorted out though I do wanna run some more general benchmarks just to see how good my PC runs, just dont know what programs to use
AVX2
I had it on auto
I'll switch to that then :)
ok well I tried to switch to the all instead of just temperature tab and it crashed the program LOL
his screenshot shows sse
Yeah that happens when the memory is full
i checked settings and it was on auto, ran AVX2 and it was the exact same so idk
Hmm, maybe they changed it in a recent version because my auto did AVX
My version predates AM5 though so 
What's wrong with the pc?
Memory errors and blue screens when using expo but not jedec
Other people reported increasing voltage helped
I did so and it seems to be helping, just running tests to make sure
Will do 🫡
I'd open hwinfo on the side, send a screenshot after 30s
This is a very hot test, expect it to hit thermal limit in a few seconds
Benchmark pi?
Run the bat
Not quite sure how to do that, do I just open with y-cruncher?
Put the bat in the ycruncher folder, double click the bat
Says "Invalid Parament: -VST"
oops, delete the -
edit it
@proven canopy Did you mean for that to be for a 16GB kit because he has 32GB and it's only using 68%
Good core+mem quick stress test
Put the vst.bat in the ycruncher folder and run it.
http://www.numberworld.org/y-cruncher/#Download
Run this for a few minutes to get memory in the ballpark of stability, then move on to tm5/memtest overnight.
Yeah, larger probably catches things faster
I think this is the best quick mem test, then move on to overnight tm5/memtest
Forks weird question, how do I test for game clocks?
wdym
Run a game?
If render or synthetic is too strong, can I reliably test with something that would be similar? Let's say, GT2 TSE loop
It's a weird question but tryna make a little sense of the KS when I do VF curves
Time Spy has the advantage of giving easy to read reports on max and average clocks during a test, as well as graphs
tse cpu test is very light on core and doesn't touch mem much
Sure, that'll give you repeatable results too
Ye I know, that's my point
I want it to be light intentionally
Or is this a moot idea?
Try valley
valley is a gpu test?
kinda
i just use y cruncher n32,n64, and vst for imc.
telling people to open y cruncher and type
0, 1, 15, 16, 18, 0
for each command is helpful

this chip was so bad.

oh apparently i forgot to sub some scores I got during the october meet LOL
👀 11135mt/s ram go brr
xp prob
LN2 memtest 
Imagine dialing back ur windows 15 years just to get the most stability
Best windows tho
Is it only 15 years? I thought it was close to 20 by now
You know what you're right
You reminded me that it's 2023 not 2020 and covid did indeed happen

🙃
DAMMITTT
Alphacool released a reference waterblock today
not the xfx aib

AAAAA-
Tell them to do gooder
all reference go brrrr
41w too much
I'll 41W u in a second
Oh no you'll make me a whole 1c warmer xD
Or 8w more your choice
I'd enjoy 8W more
But weirdly changing boards with the same cooler is giving me a little more headroom
No idea about clocks tho, haven't tried anything in this one yet
That's pretty standard from what I've heard
Probably a tolerance thing
are these your avg 6700xt clocks or are they higher then usual?
Thats average
If you unlock power it will boost a lil higher
Got mine to 2.85 - 2.9 rn manually
@ancient bridge danny
Yea
So when I had OSD on and I was playing game said my gpu was doing 2.8Ghz core and mem at 11Ghz at with my gpu voltage at 1.075v so is that boosting
for any owners of the msi pro z690-a ddr4, don't update to newest bios v1A0, it's a complete waste of time, stay on v190
my 13700K is getting 50mV more than the last bios with the same settings, 1.380v adaptive in bios now results in 1.41v idle compared to 1.36v idle
does it affect SA voltage?
no
I wonder why they upped vcore then
maaaaybe for KS microcode?
or maybe the boss asked them to work on U-X BIOS so they had to come up with something else to do
Sounds like they didn't mean to anyway
I have killed my gtx 260 
i killed mine within my first bench, core or vram ded
My xlr8 was oc'd to 700ish I think & still working 2+ yrs before I upgraded to a 660 of all things. 
i attached voltmeter and benched it once and it passed with expected results at stock
then i reran for a 2nd baseline and it crashed and artifacted

vram or core dieded
Someone made a bet for 5 dollars that I couldn't beat their AIO r15 g3258 sub with one of my g3258
Guess I'm getting a free 5 dollars for minimum effort 
r15? really? ez clap
What's wrong with r15 lol
Nothing, just older hardware likes it better than newer like r23
Well yea, 2 haswell threads would make r23 insanely boring not to mention using it over r15 would be redundant
cb 2003 
i'll bet you this cooler
no way this is real
Seems legit, that's totally what those official logos look like
It's the thermal paste that's 5.99 the cooler is 76
Dracaena CPU liquid cooler adopts the design of separate pump head to form a high-efficiency direct current. Make the heat absorption and discharge more fully, and improve the heat dissipation efficiency.Heat is generated inside the computer tower by various hardware such as CPU, video card or ev...
They just listed the thermal paste on the same listing as the cooler to grab your attention, clearly it worked 
i hate that so much
It's officially below freezing outside. Time to move the PC to the porch and do some benching 
ill be expecting good scores
take pictures

Yep 48F in my house wasn't cold enough. Although, pc helps slightly warm bedroom ( 1-2F 
Run salad for free heat
have you subbed anything to hwbot before
R23 crashed, moar volts
whats your current best sub
You can check my hwbot profile, I don't have internet while I'm benching
Same as here, I'm on team wegg also
oh you're 3rd fastest for that chip on r23 multi w/ AIO
https://hwbot.org/submission/4820725_lorichic_cinebench___r23_multi_core_with_benchmate_ryzen_7_5800x_16883_cb this is first place guy
barely lol
1.35v crashed at the very end of the test
(╯°□°)╯︵ ┻━┻
I'll let it chill out and try again
I'm amazed what a difference a little freezing weather will do
Guys in 1st/2nd place with same cooling as you were running >1.4V
do ice water, more worth the effort than just cold air
Were they clock stretching though
as in?
Like it's set for 5.2 but effective is 4850
no
When I get my custom loop
LF2 420
I don't have a bucket big enough for this behemoth
well a cheap plastic bucket from walmart will be plenty good
cooler i mean
get a cooler lol

does this still count as enthusiast league?
or does it pump u to apprentice
ive always been hazy about apprentice rank specifics
Yes because it's ambient cooling 

6561 on R20
benching 5800x?
congrats
Y cruncher straight crashed the system, not even a blue screen just dead
cinebenches r probably what i do most consistently coz theyre fun
i dont really have any chips that get my many points in them tho lol
average y cruncher moment
As long as you get a good amount of points compared to others with the same chip 👍
i dont have any popular chips 😔
All the more reason to bench them for free points
i used to do that a lot lol, burnt out of benching for 6 points a sub
When I get around to it I'll bench my laptop, I upgraded it to a dual core turion lol
idek what that is
Exactly
hah
No one ever had the CPU because it was never an option in the s3g4 socket except as an upgrade
Back when you could upgrade laptops
im not old enough to even know the turion name
Circa 2006 laptop moment
yeah, i was 3
new pb
This board is crying rn though
320w average in timespy gt2
peaks of 360w
on the begger intel LE board
i cant believe ur putting effort into ur arc card
Boredom does many things
im trading mine for a 2080ti trio tomorrow lol
gonna do some nvlink stuff
2nd best card for 2 way, right behind titans
Who trades there 2080ti for a arc card😭
Probably arc+cash or something
in synthetics arc performs same as a 2080ti stock at much lower power draw
but in actual gaming perf is all over the place
and behind 3090 and 3090ti
i forgot those cards were a thing 🥴
im trading a750 + $100 for a 2080ti lol
Called it
should be good points
hope card can bench at least 2160 average range without cap modding it, tho the card does have a bunch if empty pads
Intel lover duh
Also arc was good said brutus
when has brutus ever been correct tho
It is, on modern dx12 and vulkan games. On dx11 support gets more hit and miss, dx9 support is almost non-existent.
There's a dll you can download that converts older games to vulkan instructions, I'd be interested to see how that goes.
yea
Brutus reacted with :thinkHD:
oml
Actually dx9 was majorly improved in a driver update
I'm sure it has been. But there's still plenty of games I play still that have no support in sight.
If I were to try arc gaming I'd be making heavy use of the dxvk translator.
Dxvk is nice to have on steam deck too btw, since vulkan is cross platform it handles the dx instructions elegantly.
im sure it'll get pretty good with time but i just cant be bothered with the card, matching 2060 performance at a $500 aud price tag is doodoo
expecially when u factor in all the issues with it, main one being having to have a mobo that supports rebar
That rebar thing is a non-issue if you have a relatively recent platform. There's even mods for X79 bios I've heard that enables it and that's 10 years old. Officially speaking you need Intel B360 or AMD B450 or newer (5 year old platforms), and corresponding HEDT platforms.
bigger issue is needing to disable csm if you have a legacy install of windows
Some boards like the p9x79e-ws I have support rebar natively 
It's also easy to get working on x99
Miyconst has a lot of videos about stuff like that
Why would you need an fm radio in a model t?
did not know anything older than skylake could do rebar, wild

$5 for 1 point
lmaooo


is pushing 1.4V with 4xDDR5 more risky since it's 4 sticks?
Wether a voltage for a die is safe or not matters has nothing to do with the amount of sticks just the die, the imc voltage would depend a lil bit running 4 sticks instead of 2 sticks
as in imc would need to be higher? wanting to get my sticks all at 6000MHz since that's their rated spec but I saw errors with default XMP settings and had to drop everything down to 5800MHz which has worked fine so far
Z690 board?
4 sticks is pretty hard to run in general, if you can you should return the kit and get a 2x16gb kit btw
probably want to increase vccsa / io first
If it's Z690 then you'll be lucky to get 4800 with 4 sticks
it is z690 and 5800 seems to be working OK, maybe I'll stress test it more
I guess more recent bios versions could help with the issue, it's been a while since testing was done that I've seen
But that's definitely why 6000 isn't working
I haven't tried anything over the 1.35V yet so it would be nice to at least know if 1.4 would work I just don't know what safe increases would be when it's 4 sticks since it seems like most people are using 2
Number of sticks doesn't affect safe voltage range
But it may affect how much VCCSA you need
the sticks are sk hynix die, not sure about anything beyond that. so is that what i'd base safe voltage ranges on?"
Yes
hynix ddr5 doesn't really have a "safe" voltage limit, you'll just stop scaling or run into temp limits before you hit it. Most ddr5 cpu/vdimm voltages are like that.
well my bios turned the numbers red when I put it at 1.4V so I dropped it down to 1.39V and it's passing stability tests now at 6000MHz so I'll take it!
Red numbers just mean "this is overclocking be careful"
Nice that it's working for you though
1.4 vdimm turned red? That's strange, sure you didn't set something else?
There are plenty of kits with 1.4 vdimm xmp on ddr5
Might be a silly question, but does overclocking CPU, GPU, and RAM lower the lifespan of any of them? And if so, is it a reasonable amount?
Assuming it’s all stable
1.4 turning red doesn't mean anything, hynix can go way above that just fine
Only if you push unsafe voltages
Gotta question for any ddr5 "experts" I bought a completely generic corsair vengeance ddr5 5600mhz when it released and have it at 6000mhz with the timings as low as it can stably memtest but recently I heard that some pmics don't always work well with all vendors so does anyone have a recommendation for what sticks I should buy for the best pmic or maybe ddr5 chip manufacturer for a gigabyte z690?
For all I know someone's already tested that hynix or samsung may have an advantage on gigabyte bios and gigabyte training
Does "unsafe voltage" change depending on the individual product?
Id* say yes and no many people daily 1.4 and 1.45 for daily on cpu and ram but past 1.5 you can get degradation especially with poor cooling
And most gpus like modern amd uarch where you can manually adjust voltage usually caps at 1.2 which is definitely safe if you are worried about lowering it
But otherwise by the life span of your computer feel safe If a special 1.4 or 1.45 ram kit has your eye
Completely, 1.5v is safe on some ram dies, not on others and if you pushed 1.5v into a cpu or gpu 🔥
So I could reasonably go to 1.35-1.4 on ram and cpu?
I haven't heard anything along the lines of this although yea if you can get a refund and a different kit i would, what country?
It highly depends
You just have to go on a component by component basis
USA if I'm reading that correctly
ryzen would 💥
Yeah I’m on 7700x 😭😭😭
yeah no, don't go setting manual voltages on zen 4
stick to pbo and co
let the cpu figure itself out while you tell it the limits basically
If you're aiming for 6000mhz
https://pcpartpicker.com/product/kRZ9TW/teamgroup-t-force-vulcan-alpha-32-gb-2-x-16-gb-ddr5-6000-cl38-memory-flabd532g6000hc38adc01
But personally I'd spring for this especially if you wanted to manually overclock
https://pcpartpicker.com/product/d7t9TW/gskill-trident-z-32-gb-2-x-16-gb-ddr5-6400-cl32-memory-f5-6400j3239g16gx2-tz5s
Pbo and curve optimizer is great ye
Yeah I guess making a 5600mhz kit run at 6000 and tightening it's timing definitely wouldn't beat trying with a plain higher bin so thank you
eh not particularly
the gains you get from running a better bin won't be significant
definitely not enough to justify a whole new kit imo
Well the 6400 kit could probably get up to around 6800mhz if you're lucky
Well as i said only if they can return the og kit
Working with a 12900k could I look into 6600 or 6800 and if it won't boot at full speed downclock it?
what mobo?
Z690 aorus xtreme
I'm not familiar with aorus xtreme specifically but gigglebit z690 has had a lot of ddr5 performance issues
Nono that was ddr4 issues at the start iirc ddr5 was fine
As of the newest bios I can't lower the timings as far so you got that right😂
nyope
Oh ddr5 issues too?
ddr5 has been on gigglebit z690 always the lowest of the lot
don't ask me about the way out there brands like colorful tho
Can confirm
Oh yeah my round trip has went from 60 to 70 from release bios to f21
Definitely not good
You have a ddr4 msi board no?
Yes, but it was a big issue gigabyte had on z690s
Anyway with a hynix kit (the 6400 one) you should be able to tighten timings and up mhz
They fixed it though
so I would say you'd probably hit the limits of the board quicker than the kit, the binning may help a little bit but imo not really gonna gain much at all
I think I'll take that recommendation it was an 800 dollar board so I'm not keen on replacing my overpriced copy of aorus master lmao
Man i cringe when i see people buying 250$ boards ouch that hurt my wallet xD
a $250 z790 would probably do better in this scenario
lol
won't look as clean but ☠️
Two slot board you think?
2 slot boards are on the whole far better
only aqua oc and unify x are the ones with any troubles afaik
If we're basing this on ddr5 performance
even then they still go ham
to give good perspective
a z690 kingpin does 7200 hynix m
you're struggling with 6000 💀
I mean like I'm already owning only 2 sticks so yeah I could transfer it
but you need to also ask yourself this
is it worth not just the physical but also the time investment
gonna take you a long time to do it all over again lol
you sure you have hynix sticks?
samsungs also do 6000
No these are samsung die we were just talking about replacing them with 6400mhz hynix
ah
bleh yeah
changing to hynix might give a little boost then yeah, but only really if you can net zero or close
Very few people will actually benefit from faster ddr5
yeah, we are talking about single digit differences at most
A 2 ns difference is enough to get my interest 😂
so yeah it'd be board and ram then which isn't really worth it
@modern walrus Got this stable through occt, had to loosen rdwr's and wrrd's to get it to boot, any thoughts on what I should try for those?
Primaries are bad but I'll tighten primaries later
6400g2

Actually I'll go tighten primaries now and try to get faw and cwl down
I usually boost trefi to 65000 or so for a good boost it was my old limit before gigabyte fully unlimited it
The upper limit is like 200000 and you can lose data around the top isn't it?
mhm
top is only a little above 128k on mine
I'm still testing tho but seems stable
cinebench is like, completely lethargic to memory
you'll gain maybe two digits of points between a full overclock and xmp
It goes up as I clock the memory from 5600mhz so must just be a bandwidth thing
I'd say that was something else at play lol
My score with looser timings was like 28256 and before the oc It was 27890 or so
that can be difference in ambient temperature lol
I can't use Aida since that shows inaccurate speeds so I just went for cpu benchs
your kit is XMP 6400? go for 7k+
I'm using an lga 1700 contact frame and an h150i with thermal grizzly with a .08 undervolt it only gets to about 67°c
I'm running 7000c32 with 6000c40 XMP lol
You have plenty of thermal headroom then you should oc a bit
I like efficiency so I've only overclocked the e-cores which can do 4ghz with the .08 undervolt
They can do more like 4.4
4.3
Close enough
or 4.2
At the cost of the undervolt?
Forgot it was 12th gen
E cores basically take no wattage so doesn't matter much if uv usually
oh alright not bad not bad
I went from 240 watts to 177 watts so should I try it without raising things?
which motherboard btw
Z790i edge
Rip
I guess it is fair to say I didn't exactly try to see how higher the e cores could go at the lowest voltage my p cores would
ycruncher lol
I might update bios and see if that does anything for me
I see you dont know the
of setting jedec and yeeting the cores higher
usually newer BIOS is better but not sure
how ya been? long time. Everything good?
lol how POS are we talking?
I got a windows vista with a turion x2 that is BSODing
I started by trying to get the highest frequency to POST with auto VDD2/CPU VDDQ/SA. Once it stopped POSTing I went back to the highest it POSTed & checked hwinfo to see what voltage it was running then tried to add a little to them & keep climbing
tRCD/RP won't tighten?
tRTP won't do 12 & tWR 48?
tCWL = tCL-2
memtweakit reads better than Dragonball I think
Looking good
that's the newest one I have
Currently testing highest freq I can get with auto voltage, 7000 and 6800 haven't worked so far
Wait I forgot to set vdimm
Haven't tightened them further, I was mainly curious about tfaw of 16 being stable or not, considering some sources say tfaw of 32 is mininum
Faw really helps with tm5 times and I'll prob be running tm5 later as extra validation
Primary may not make the biggest difference but the tertiary are still noticeably lower
I think I'll stick to 6400 for now and I'll try those
& not trying to tell you what to do just suggestions
You don't need to set those at all
I would try tWR & tRTP together; tRRD_L & tWRWR_sg together; & tWTR_S/L together
ya they're meaningless it's just for decoration lol
It's for dfferent rank/different dimm
So unless you have 4 sticks or two dual rank sticks, nope
I do often set them to 1 to make the other timings easier to see though
Aren't all ddr5 dimms dual rank?
are 2x32gb sticks DR?
Wait
like none of them are lol
No, two 32 bit subchannels
it's quad channel
Not necessarily, but probably
Some people set it
tXP?
Yeah
your board doesn't have the option for Powerdown disabled?
I haven't seen it on my gigabyte board no just my "old" msi board
So I just manually set the start and stops since I don't have a power down toggle
tPPD to 1 is the same as powerdown disable I believe
Oh cool if I try to change things I'll use that
Changing from 5600mhz samsung to 6400mhz hynix should give me some nice fresh tuning to waste my time with some mindless tech entertainment
Hm it was unstable
Might've been thermals, lowering vdimm has made it stable so far
(reverted faw to 32 too just in case)
usually HWiNFO will show you PMIC temps
Good idea, I'll check that
usually its checked in as a "Im not going to spend money to fix it, but fix it"
bdie heck ya that was a BSOD temp lol
I think it was a walk in the park for Micron tho
I'm trying to get OCCT running so I can see my temps
I'm using water cooling tho
the thing is, iirc the like overtemp protection in my Unify-X BIOS is at like 125C or something nuts
Yea but I assumed Hynix was less sensitive
Dunno, maybe thermals is the actual issue
It crashed due to some driver thing, likely caused by ram, but I'll run tm5 again and see if it logs errors
I'll keep hwinfo64 logging too so I'll see the temps
alright looks like temps might actually be the issue
Put a fan on it and it stopped errors?
going from 1.5v to 1.4v has made it go from 5min to error to 15min to error, erroring in the high 50cs
ITX, no space
just need to decrease voltage and keep testing
Low profile 60mm fan should be fine still in itx
aio tube gets in the way
What you see as an obstacle I see as a mount point
If I run it with side panel open I can fit a fan in but I'd like to daily whatever I get
lol
Still a fan for testing if it's temps is in order I think, even if it's not permanent
I'm just going to keep lowering voltage, prob need to loosen some timings and prob undervolt cpu too just in case
While that would prob be best, I don't have an extra fan around
The only case fans in the system are the aio fans
And they're already mounted weirdly and I'd rather not have to mess with it again
Will prob try dropping vdimm to 1.35 and SA to 1.15 and vddq to 1.25
I hate that my ram temp still keeps slowly climbing up during testing
I doubt I can get the timings stable at below 1.35v vdimm
Hoping it stays steady at 54c as it is now and doesn't error out
Hey that memtweakit you sent was a whole year newer then mine and includes the new 13th gen "trcd wr" so thanks for that
4x16 with no active cooling let me run 1.4 vdimm on m-die 6000c32, I think you can do it
npnp
got bored
started ocing a hd 3450
weirdly the newest drivers are worse performing and oc stability wise than the windows 8 ones 🤔
You'll probably want to be running xp, win7, or 8.1 for most benches with that card.
hmm
noted maybe I'll see if I can free up an hdd
saw most of the top ocs on hwbot on 7 but just assumed it was due to time/age
W10 22h2 is pretty good for some modern hardware
It's actually good on x79 apparently. I haven't tested it very thoroughly compared to a stripped 2012 r2 or w7 though. Just ran a few benchmarks between them
It's more to do with drivers than bloat for modern stuff
Or for specific benchmark compatibility/effi
I'd go 20H2 to avoid the performance penalties from exploit mitigations
And 20H2 should still be compatible with all the new drivers
Yea, I'm just testing this stripped 22h2 out of curiosity. I'd go back to 2012 r2 and w7 sp1 if I'm actually going for tough scores
Though I wouldn't be surprised if W10 is the best option for superpi now because of the tweaks
For x79 specifically I mean
Should test it on lga1150 too 
afaik you can disable any OS-side mitigations.
Possibly but how many know how to do that?
Just use inspectre, good enough
Ah yea I forgot about that
Is that in the wegg drive
Probably, but you can google inspectre
Well I already have the wegg bench drive downloaded so I was wondering
crtl f
@steel nymph Board is pro z790-a CPU 13700kf
what xmp bin
is it 6000 cl32?
6400 32-39-39-102 1.400v
yeah that's hynix
According to bios
Where do you find those
bios?
Oh wait found it
and vdd2 on auto
I was thinking it was in a different yab
see if 6400 with auto timings trains and is stable with those timings
if not play with the vccsa and cpu vddq
both of those have a sweetspot that is ymmv
generally for cpu vddq the lower you go you can gain stability, but the lower you go it is harder to train
The two I don't see is vccsa and vdimm
CPU sa?
yes
Kk
yes
ideally just set the primaries manually
with manual
no xmp
and voltage
XMP on can do some weird things in my experience
let the mobo train the subs
affirmitave rebooting now to bios again
@steel nymph ram is at 50c is that normal while running testmem5
Yes, but DDR5 gets hot
You want to stay under 45-50c
Around there you start getting temp errors
Slapping a fan directly on ddr5 is highly recommended even for xmp
Your temps are technically fine as long as you don't get errors
Degrading temps are far higher than when you get errors which is nice
its staying around 50c
baha turning my fans from silent mode to normal dropped the temps to 48
Yeah I had to waterblock mine
It really doesn't get that hot outside of a stress test
But, the important part is that the heat from your gpu can cause issues
As heat rises
yeah having the kingpin i think kinda helps there also i have 15 fans in my case rn
phanteks 719 is the case
The problem is direct airflow on the ics
And heatspreaders can cause the ics to be warmer
As the "thermal material" often isn't thermally conductive
My sticks had double sided tape and foam

i hope this is enough ground

it's just a 650ti boost, how much could it possibly draw 
guess what platform im OCing memory on rn
the way the vrms were set up gave me an aneurysm so i just insulated what i dont need
x99
lol yes
i know, im watching u from ur window
ok what loop am i on
2


what kind of dimms am i using
the red ones (i do not know ddr3)
ok but imagine i got it right
x99
ram oc
is there a problem
zen 1 simulator
zen 1 is worse...
isnt it just tuning more than actual ocing lol
ocing is tuning
whats the max ram freq u can get stable on x99?
i dont see your point
i mean more tuning than freq increasing bozo
well
u do that with every platform
YOU GET MY POINT
bozo

well im running ~3333 C11 rn
not really benching for subs rn, just pre-testing
and for funzies
mem oc is still fun on X99 for the practice of it if you know what to expect. it just doesnt clock high because duh, it's first gen ddr4
ended up putting vcore and ground on the back instead
gonna use 10awg wires for just getting it working
copper plates afterwards
dcuii 680 vrm
3.3v and 5v probably injected thru sata coz i can not be bothered to setup buck converters

Why is the 7zip world record for the 5820k so bad
It's some mess of a sub
On ambient aio
7zip is years younger than the 5820k and only recently got hardware points
Yea but still
I'm surprised nobody has even done dice on it
Welp, more easy
for me
How do I over clock a i7 6700k?
Are you on a Z motherboard, Z170 or Z270? Any other chipset can't overclock

What gee pee u
Only 460w smh make it go moar
- Ignores the 599W on the right *
Me blimd
oh come on thats weak
@kind walrus come on pull 630W
Efficiency go vvvvvvv 
While coil whine goes brrrRRRRRRrrrrrrRRRR
my 3090 strix has awful coil whine even at 60 fps/200 watts
with 3 different power supplies
pcb issue I guess
i hear nothing BTW
this is GTA lol at 4k
my power rials eating so much power
that's nothing for a 4090 though...
I'll regularly hit 700w+ in hwinfo with just ow2 at 1440p
AYO
WTF
max i hit is 630W rn
diving into the psyche of a ddr3 fanatic
OW is a good power supply test, heavy on everything
YEA IS THERE SOEMTHING IWROONG?
yes thats my saying
thats why i want 1000W gpu and 500W cpu plz
@kind walrus i got more of the funny ram would you like to see
Real full amd pc 
factual facts
None of this g skill nonsense
g.skill pssshhh
Id take off the heatspreader but it'd probably rip off the ICs with it
and im out of LN2 so i cant do it that way 
could be worse
I'm trying to put a fan on my ddr5 and I'm out of fan headers 💀
what do

you dont even have a splitter
lazyyyyy
oh i was gonna link you something but forgot you're au
au
you could always solder the two fan cables together 🤓
I have splitters and controllers but heckin lazy
proper skill issue
proper idotot
Get arctic pst fans ez

Alright I've opted to ditch my 570 for the time being
Yes
Arctic go brr
If I want to optimize the perf of my 3000G/Vega 3 what would be the first thing I would want to oc?
Get your 570 back?
Uh ironically best thing you can do for integrated gpus is faster ram but faster ram costs more money but it's pretty useless sinking money into an igpu
I've been dealing with issues with it for a looooooooong time I'd rather not deal with anymore
I'm just gonna sell it and put the money aside for a new gpu
Can just oc the ram I have then, that's why I asked :P
Ram oc is ehh not something you should take light heartedly
I've done it in the past, tbf, started having issues and reset it a long time ago idm giving it another go
Tell you what, get the detailed report of the sticks from thaiphoon burner and I can tell you roughly what to expect when overclocking them
Or if they're g skill send a pic of the sticker on the stick
They're adata
Yep no useful data on those stickers for finding memory die
Download Thaiphoon burner and get the report, should look something like this
You'd just hit "read", select a slot with memory in it, then hit "report"
it's detected as a virus
By what?
Hynix M
I believe sometime then the best I could do at the time was like 2999Mhz with decently high timings iirc myself
The best those will do is around that frequency on 8Gb MFR, 2933 or 3000
Those ones are pretty bad by modern standards
i'm pretty sure i've had H8M at 3400 before
Probably a newer stick than 2016 then
but in general you buy those kits to run at xmp freq/primaries
can still gain perf by tightening seconds/terts
MFR
?
your IC is 8 Gbit Hynix MFR
lol the name of the die
commonly abbreviated to MFR or H8M
ah
Hey, they were free so I cant complain
You got your money's worth then
they do their job
just not great overclockers
i've bought 16gb kits of similar tier ICs for $20-30 for cheap flips and they work fine
being that their base is 2400 any oc is a good one on these 😭
Poor 2400 stick though, I've seen DDR3 XMP there
lol, there's ddr3-3200 xmp
i've got a DDR3-2400 kit just chilling on my desk in my dorm
waiting for my friend to send me his z97 board
of course there is
MFR 
thats what ive been binning
Have no 2933 or 3200 bin dimms though since they're hard to come by and really overpriced
wonder if their DDR3 IMCs are any better than like, haswell
@proven canopy might know
he barely touches ddr3 
The boards aren't good
he told me about using 9900K for DDR3 igpu competitions
ah, that makes sense
wonder if you could DIY a good LGA1151 DDR3 board
i wouldnt put much faith in the memory controllers anyway if broadwell is anything go off of
DIY mobo 
like frankenstein something together
i'm not nearly smart enough to know if something like this is remotely possible
A pattern being repeated currently, there's a reason all the best DDR4 scores are still on an 11900K

about to buy an apex
not having fun
7800 stable save for some secondary or tertiary timing on whack
sorry 7800 c34
first round of OW2 for the night

Good core+mem quick stress test
Put the vst.bat in the ycruncher folder and run it.
http://www.numberworld.org/y-cruncher/#Download
Run this for a few minutes to get memory in the ballpark of stability, then move on to tm5/memtest pro overnight
will do. ill share results after
Yes - it's a very simple batch file
which named folder should I put vst into?
Put the vst.bat in the ycruncher folder and run it.
And you're sure you don't have core instability? Try R23, then occt cpu tests
I would assume your ram has either always been unstable or you didn't successfully reset your bios to stock
Yes, it's unstable, you don't need to keep running it
The ram has been very stable until recently
I never really reset bios, I simply updated it to the latest version
well...i did do default settings a few times, but that didnt fix anything
Updating the bios will clear all your settings automatically
Is AI over clocking better than manual overclocking
No, "AI" has nothing to do with it
It's just a table of frequencies and voltages it expects to run
They call it AI to make it sound smarter but most companies have some sort of auto overclock
OH.
The issue with it is using those disables core boosting behavior by setting a manual all core frequency
Oh.
hmmm higher numbers
✨ marketing ✨


)


