#overclocking
1 messages · Page 76 of 1
It’s from what 2011 or 2009
IT STILL COUNTS
https://www.3dmark.com/3dm/62861314? final score for only 2125 and stock CORE
sheesh!!!!!
Yeah Imma do over next week too
Damm
Oc the coreee
Getting a better psu on Wednesday or at least ordering it so I’m holding out
At least on big core ocs
Ahh fair enough
Is a 4.4 Ghz base at 1.3V good for a 3700x?
Or should I try to push it higher, still haven’t had any crashes at all
All cores are about 4.4 Ghz
I'm not even sure you want to daily 1.3, I can't remember what Fitz said. I know that's right around the degradation zone
@tall pelican you around?
I brought it down to 1.25 V and it’s still stable with 4.4 Ghz all core
That's better as far as I can remember
Originally when I did Auto OC it brought it to 1.4 V but that seemed too high so I brought it
Yeah 1.4 def too high for loaded CPU
all core loaded
Found it
You want to stay below 1.25 all core load voltage
So you can set 1.3 as long as the load voltage is below 1.25
How do I check load voltage?
Why?
all core OC is one of the worst things you can do to a Zen 2 CPU
Why is it?
I can't tell you the why
But I'd listen to Fitz lol
I'd wait until they're here for an indepth convo
Ok thanks
Actually I don’t think I’m going to bother with OCing since after applying the 4.3 all core there was only a 4% increase in performance on multicore performance
Best to just turn on PBO and raise some of the limits a bit TBH with Zen 2
I love my mobo, like it has its own button to reset any OC settings if they aren’t stable
1.25 for zen2
Fit means absolutely nothing
Temps go up? Fit goes down
Current goes up? Fit goes down
Tune your ram? Fit goes down
Fit is only viable at that current and temps, with smu limits in place
Ahh that makes sense
eliminate all variables except 1
soo (butting in late as usual) what's the degradation voltage for 10900K?
So why is all core OC the worst thing you can do for Zen 2?
Its not the worst, its generally not the best
3600 has a low stock/pbo boost clock, and generally need a manual oc
I’m OCing a 3700x not 3600
Really depends on your conditions, if your board has a current reading it is much better to go off that since the current is a much bigger driver in degradation. I doubt you'll have any direct issues with voltage
The current doesn't mean nothing
If you really wanna be safe you can just match the current
I need help ocing a 10850k, what should I do in bios before just using xtu to do it
Bump your power limits, I believe Intel refers to it as TAU
That's assuming your board can handle it, that chip can pull over 200W
"I overlocked my GT 1030 and used SUES PTGI 60 FPS full raytracing NOT CLICKBAIT!" the youtube video says, and its a video of his just upgrading to an RTX 3090
TAU isn't the power limits, it's an averaging constant used for the PL1 EWMA power calculation
Why are you using XTU?
because i was told to use xtu
what is better then
bios, unless if you know what you are doing
thats the whole point
im not wanting
what?
or im wanting to be able to have something on windows to oc, though some things id need to set in bios before
you don't want to use bios?
Milk — Today at 10:11 PM
im wanting to be able to have something on windows to oc, though some things id need to set in bios before
Why don't you just use bios
because i dont want to restart my pc to access
it
and would rather have something on windows I can oc with
but
it wouldnt need to be changed that much
so maybe bios is fine as long as I dont gotta go in it 24/7
which I wont
Overclocking even through OS will make you need to restart your PC since you'll get crashes
i've read that vdimm only really affects frequency and CL
is that true?
i got 3866c14 stable at 1.52v but all other timings are auto for now, just wondering if it's worth trying to push vdimm lower
1.52 is already down from 1.55
Depends on the IC
rev. e
pretty much yeah
so i should just try to get vdimm lowered at 3866c14 and once i start tightening other timings i can assume more vdimm won't give me more headroom for tightening?
yeah
Just be careful if you have differing ambient temps (or if case airflow is an issue), it's very likely if the ram was 5~10C hotter it wouldn't be stable anymore
i thought rev. e is not temp sensitive?
it is
buildzoid did a vid where he aimed a hair dryer at it and made it 70c and it was fine
it's pretty much in a worst case scenario right now, it's in a half finished build where i haven't put the fans in, but i had to put the glass panel back on to use as a table for my peripherals due to limited space
That was at loose as hell timings
with loose timings yeah you can run the stuff at like 100C
There are even Rev.E chips built to run at 130C, but that won't be with good timings
it's like 3866 14-27-27-27-58 rn lol
😅
Buildzoid tested 4266 cl19 1.45v which is very loose
So of course he didn't get temp issues
I don't know why he tested such as loose config
If Rev.E wasn't temp sensitive the 4266 cl16 xmp kit wouldn't come with a fan
It comes with a what now
my reve would error at 72-75c once tightened down
should be fine then im nowhere near that in worst case scenario
and it's been 100-110F every day this week
Depends on how tight it is.
My Rev.E daily errors at 53C
Because it's tuned to not go past 40C
There isn't a line that you cross then get errors, completely depends on the settings
hmm
maybe i need to install the fans then
didn't want to cause the motherboard has defective usb headers so i'm waiting for the replacement
but i guess it's not that much work
what?
Rev E scaling isnt exclusively tCL
what else scales then
tRFC still scales to an extent.
It doesn't
it does.
1.1V to 1.6V did nothing
I've had difficulty getting 310ns stable at jedec
1,2
1.55 got me to 290
Vtt 0.7
With loosened tRAS and CAS I could go straight to 1.2V from 1.55V nothing else changed and it was stable
and nothing could tighten any further?
nope
how?
I just got an EVGA 500 watt power supply for $28 on prime day
it's non-modular but it'll make testing my cooling loop way easier
Should just get a paperclip and bridge the pins tbh
It's such a mission to disconnect everything. I'd have to take the power supply out for sure because there's at least 4 connections above it
plus I might build my dad a PC down the road & a 500w EVGA PSU for $28 is great if I actually do that
don't get me wrong tho, I appreciate the advice
btw what's the degradation voltage for 10900K? lol

How your cpu score so high?
It says 4416 tho?
bios had it set to 4.425ghz at 1.25v
beat the poop outta that score soon
that was with mem at cl7
its stable at cl5 rn
Nicr
Why gonna strip win 10?
also im lazy
less bloat=more score
Fair
some very interesting behavior when running tRCDWR and tRP too low
set them both to 10
system booted but showed an error on login screen saying some application called WerFault failed to start
then in windows I couldn't open anything due to filesystem error
and everything lost their icons
set them to 11 and everything is fine again ran sfc and dism no corruption found
Rcdwr doesn't do anything since am4 has rcd parity
what does that mean? rcdwr will stick to rcdrd regardless?
Performance and stability will always wr = rd
Doesn't matter
Just know the stability thing you posted earlier was completely rp, and not rcdwr
I see
both tRP and tRCDWR are stable at 12
but I'll compare tRDCWR 12 to 19 and see if there's a negative performance impact after this test
I found that tRAS 32 seems to perform slightly better than the theoretical minimum tRAS 31 (given tRCDRD 19 + tRP 12) and also better than tRAS 33
so I'm testing now if it's stable at 32
There isn't a min theoretical tRAS, below tRCDRD+tRTP you just don't get better read performance. Note it's tRCRD+tRTP, not tRP.
Of course it does something, otherwise the timing wouldn't be present
You can bench it and even see the latency improvement
for writes it is tRCDWR + tCWL + BC + tWR, which will probably be longer then tRCDRD + tRTP
going to try to tighten tRTP now then
I'm just confused now, I've been messing around with my primaries on my dual rank djr sticks, and I swear I was able to boot with 16-16-18-18-34 a moment before, and it just won't boot now...
what freq?
3600
maybe you accidentally swapped trcdrd with trcdwr
With gdm, secondaries set to auto, vsoc set to 1.15v
Hm, possibly, but I wrote it down as I was able to boot with 16-16-18-18-36 and 16-16-18-18-34
(I was just trying to slowly tighten tras, and had the thought of running tm5 on what I knew was bootable)
1.6 is fine
1.6v is fine for daily
Interesting
Might just slap more voltage in then
That might help
It's a 16-19-19-39 1.45v xmp kit
https://www.gskill.com/product/165/184/1620120499/F4-5333C22D-16GVK-Overview
https://gskill.com/product/165/184/1618389754/F4-5066C20D-16GVK-Overview
https://www.kingston.com/us/gaming/hyperx-predator-ddr4
here is the proof that it's fine
Damn okay
1.45v xmp? what freq?
3600
huh
my tRAS won't boot below 31
what is tRC
was 90 when I was trying
that's irregular
I am testing 33 tRAS 56 tRC now
with a tRC limit should easily do 21
tRC boots as low as 51 but performance seems to be bad
ah
So it's a gskill kit Mr1111? And you checked the 042 code right
G skill kit, yes, wdym 042?
Is it G.Skill
tRC at 51-55 consistently added over 1ns to aida latency compared to tRC 90
the sticks have 042 codes that give away the IC, how do you know the IC?
that xmp bin is for sure cjr or djr
no
the memory channels train independently so they are bound to do weird things
AMD terts in general are just a mess
I'll have to check, I just used thaiphoon, how exactly do I check the 042 code?
it's on the stick
starts with 042, last 5 characters matter
might be like 8821D or something
@dull ginkgo
Oh yeah it's correct now that's right
But yeah, it's really important to know since CJR degrades at a little over 1.45v
Gotcha, I'll check it when I'm back in a few hours, I've been doing 1.45v so far anyways
deg has been reported as low as 1.48v
Hm
04213x8821c is CJR?
Thaiphoon can be very special sometimes
mhm
lol 3200c14 c-die
Anyways this is the kit, I bought it at a much lower price than right now but yea
https://www.newegg.com/g-skill-32gb-288-pin-ddr4-sdram/p/N82E16820232861
but yeah if it's reading 8Gb Hynix it'll be JJR, CJR or DJR
JJR's voltage tolerance can be assumed to be the same as CJR since they are process buddies
(JJR is not re-branded CJR though like some say)
tfw the SPD is programmed to PSC C (the real die), but thaiphoon still reads PSC A, stupid program
Ignores the SPD
well typhoon guesses my 3200cl16 corsair c die as a b die ;-;
The part number in typhoon read djr so dunno, I'll have to just manually check later
It guesses every 8Gb samsung stick in existence to be b-die
it was guessed as DJR
Gotcha
I literally was able to delete the entire SPD excluding the part that said it was 8Gb Samsung, and it said B-Die. The SPD didn't even have a jedec memory profile, voltage spec, or anything
interesting, my tRTP can't go lower than 14
that is very weird
my other rev. e kit did tRTP 8 stable
I ran tRTP 12 with tRAS and tRC on auto and got 77 errors in the first second
below 12 no boot
post code F9
Huh
im boutta try 3600 18 10 10 10 40 at 1.35v on c die, should be okay right?
trcd 10? no
Would prob need more voltage ..?
prob typo of 20
18-20-20-20-40 most likely would work, maybe need a bit more voltage, but yea
yeah meant 20 my b
1.35v should be enough
cool thanks
Thaiphoon guesses this to be B-Die 
so if I have tRCDRD 19 and tRTP 14 then tRAS shouldn't be below 33?
Lol
it can be
just doesn't help read perf
It's likely going below that just does nothing at all
okay
I'll test 33
I saw a lower aida latency at 32 vs 33 over a 10 run average but could just be aida being aida
that score tends to go all over the place anyways
What happens when you clear cmos and try and boot with that
bricked spd
board wouldn't know how to read it since the spd revision isn't defined so it wouldn't do anything
And that would brick the eeprom on the stick? why
how did the spd get bricked
you could hardware flash it back but that's the only way
For my ram voltage is it vddp i should be setting to 1.35v?
It's been a couple months can't remember
Which one is for dram voltage then i cant find it lol
Nvm I'm blind
For my soc should i be putting in 1.1v because I'm on gigabyte or what? This bit of the ddr4 guide is confusing me
soc should stay 1.1
Im but should i put 100mv into the soc thing in bios?
How does this look?
are you only doing primaries?
procodt can probably be a lot lower than 60 ohms
but not that big of a deal
Yeah only primaries
Primaries gang
Nice
honestly tfaw seems to have a bigger impact than primaries except tcl
Does it?
it's crazy how much faster tm5 finishes after tightening tfaw
told ryoku to tighten trrds trrdl tfaw from 7 10 41 to 6 6 24 and he said he gained 20% fps in his cpu bound game
that number still a lil sussy tho
most ram can go tighter than 6 6 24
all of my rev. e kits have been able to do 4 4 16 stable
tm5 finished in half the time at 4 4 16 tfaw compared to auto
Damn ok
takes 1.5 hours for all 3 cycles to complete and im still working on primaries
if you can really push 3200c16 c-die to 3600c16 at 1.35v thats insane
oh
LOL
nvm
Wish there's a way to do men oc without having to reboot or reset cmos so many times
Same :p
Like gpu oc so nice to do no reboot required 
the most annoying and time consuming oc
but the biggest gains
ive been at it for 3 days and im still working on primaries
took a long time to land on stable 1933 fclk
arshia was nice enough to teach me what all the stuff in the right column of zentimings means
Also what tfaw would you recommend to try?
had to play around with procodt, drive strengths, and stuff like that just to be able to boot
Ohh sounds complicated
it's only really necessary for pushing really high fclks
That sounds like a job for morning pato
when i oced cheaper kits i'd be done in 2 days
Ahh good then lol
Hm
my laptop's ram oc was fairly simple
barely any voltage headroom
bios didn't allow controlling RTLs/IOLs
Update, I did swap them, 16-18-16-18-34 boots
that was a 2 day oc
still nice that a laptop bios allowed ram oc at all
and the auto trained RTLs/IOLs weren't bad
This is how it looks atm
I think I was just reading off zentimings
which has it opposite from the order in my bios
every time you tighten something check to make sure you actually gained performance btw
for me 51-55 tRC boot but perform worse than 90 tRC
I suspect I'll find the optimal point in the high 50s
Going from 16-19-19-19-39 to 16-18-16-34 right now got from around 75ns in aida64 to 72.6ns
I'm happy that I only mixed up trcdrd and trcdwr and it boots now
My goal is just sub 70ns atm
closest I got was 55.0
Damn
but I'm like 1/3 done
Neat
I'll have to do things one at a time
I have 0 idea how the secondary and tertiary timings work
tRRDS tRRDL tFAW can be changed as a group
try 4 4 16 first if it fails try 4 6 16 if it fails try 5 6 20 if it fails try 6 6 24
Sippy, how you measuring ram perf ? 3d mark r sum?
aida64 memory and cache bench
I've never had a kit of ram not do 6 6 24 tRRDS tRRDL tFAW
Okey will do
Tm5 will end when it's done right? Not a loop?
Update, looks like it is cjr, 04213x8821c
I'll stick to 1.45v then
I guess thaiphoon fails again
What did taiphoon guess?
Djr
Booted with 4-6-16, gonna see how that affects aida64 latency
Run geek3
hard to say when you've tightened multiple things at the same time
I'd recommend just setting everything auto except tCL 18 to see if you can get 3600c18 down
and work one at a time on primaries
gotta be patient
it's only some of the secondaries and tertiaries that you can do in groups
Well uhhhhh
I guess 16-18-18-38 isn't stable
30 errors and a bsod
That's amazing
this kit has some real oddities compared to my other rev. e kit
tCWL not going below 14 while the other kit did 8
and tRC needs to be a lot looser as well, currently testing 58 after getting errors at 57
old kit did 54
granted old kit was running 3733 and this is running 3867
I'm not sure if tRC is at all related to frequency I'm pretty sure tCWL isn't at all
Well, 16-19-18-38 isn't stable either
Sad times
This is looking like I'm going back to xmp timings
Also, what is DRAM CH_A VREF voltage?
cjr?
those are mostly for pushing higher fclks i think
VREF is the reference voltage, it is a voltage that signal are compared to. If it is above VREF the thing is a 1, if it's below it is a 0.
So if VREF is 1V and the memory gets a signal of 2v, 0.9v, 1.7v, 0.3v, 0.8v, it would be interpreted as 10100
with cjr, I never saw mine scale rcd/rp with extra voltage, so just gotta make it looser as you clock higher
tRC 61 crashed, is it possible for too low tRAS to have an impact on tRC even if tRAS is stable when tRC is on auto?
would increasing tRAS have any possibility to allow lower tRC?
Staying at 3600mhz, it's 3600mhz xmp.
so basically
It's finally stable now
the only thing I've changed from XMP is tRAS from 39 to 38, and tFAW tightened
an hour into tm5 is prob good nuff for now
that'll be enough mem oc for the day
Lesson of the day : Spent way too much time seeing what boots and not what is stable
I could've just spent 5 minutes each time running tm5 and it would've threw errors for me
instead of me spending like 2 hours trying to tune primaries and resetting cmos way too many times just to figure out it's not stable
is lower procodt always better provided it remains stable?
like would it allow me to tighten timings more
or should I be looking to increase it instead
if tRC is limiting no
all procodt affects is stability, if you are stable it's good enough
procodt is stable at the minimum setting of 28.2 when tRC is auto (90)
can I potentially gain more headroom for tightening it by increasing procodt
I've booted at 53.3
Every board is different for that, only way to know is to test it
I see
for tRC? no
I put on tRC 65 along with 53.3 ohms to see if it enables that
since tRC 65 crashed at 40 ohms
it's just weird to me how this kit can be stuck at such high tRC when the other rev. e kit is running 54 stable
nope, got an error
back to increasing tRC then
tRC still extremely unstable at 70, at this point it really might take 90 to be stable
on a whim I decided to increase vdimm back and now tRC 55 is working... guess tRC scales with voltage on rev. e after all
one cycle of tm5 has now passed, will update if it finishes stable
ok now give it 3 volts
no thanks
why not
1.55v is the daily limit for rev. e i already feel a bit uncomfy riding right up against it
ok go to 1.53
which is why i tried to drop to 1.52 in the first place
i'm gonna finish tuning all my timings at 1.55 before seeing if any drops are possible now
jus put a fan on it
I liked seeing buildzoid casually put 2.2v on a set of b-die sticks and on stock cooling lol
You see i did that cuz i had oc'd this kit before but lost the exact timings due to bios update
Also yeah it went a good 20 mins with no errors in tm5 with 18 20 20 20 40 so should be okay i feel
I've had bad experiences with walking away after 20 minutes lol
20 minutes no errors
go get some water
come back to 730 errors by 25 minutes
Yikes
Well @short blade I'm trying 18 19 19 19 39 now so 🤞
Left for 5 mins then hopped in shower so will see how it is
Also jeez you're up very early/late lol
What this mean :(
Welp guess's I'm trying 18 19 19 19 40 now anyway
Could it have been the mining that it didn't like and crashed ?
Cpu or gpu mining? Cpu will absolutely hammer the cpu, while gpu mostly just usee the IF
Gpu mining ofc i just heard sometimes you need to leave x amount of ram for windows which i didn't do
What do you mean the IF btw
Hmm it takes 97% of my ram and 100% cpu
Yeah, I've heard of people using xmrig as a memory stability test lol
lol I left tm5 running overnight and now my windows install is gone
that's impressive
was planning to wipe it anyways
Sip HOW
I dunno
I watched it run into cycle 2 with no errors
and went to bed thinking it would be fine
I come back downstairs in the morning and I see a windows install screen
LOL
prob rebooted, and default on your bios was an install stick
you should never hit a reinstall screen unless you have install media plugged in
huh
default on my bios was still set to windows boot manager
I closed it and checked
I do have install media plugged in, my testing software USB has a partition for windows installer
feels good dialing in some really low timings thinking to myself "no way this is stable" but then it's stable
This happened to me mid benching some time last week. I had two win 10 installs on the same ssd, deleted the boot manager entry for the second but somehow benching on the first install deleted its own boot manager entry and then re-enabled the second.
CJR?
Samsung C? really?
nice
theres a few things you should change
I finally finished it
I need to get some of that fancy coolant stuff but I think it came out alright... Anything I should know about OC on a 5900x compared to 10900K? Lol
Very nice
Thank you
dont spill coolant everywhere
It took me from about the end of April until now lol
I sat there staring at it for about 4 hours waiting for drips
I got a 500w EVGA PSU for Prime Day for something crazy like $21+tax
so I used that to run the pump only
hopefully ill have more luck this summer because i dont have school lol
To get a graphics card?
Good, now put up some points for https://hwbot.org/team/wegg/#Members
I should sub more tbh
Yas
I barely have
I'll try to sub some more too
yes
I got mine from Tech with Excellence. At first I thought it was some sketchy scam
but I wanted one enough that I figured I could get my credit card to reverse it if it was
It took about 2.5 months but I finally got the EVGA rtx3070 ultra gaming
Then I got a waterblock from EK sent over from liek Slovenia haha
their website says 8-12 on a PNY 3060 XLR8 rn
Ya that's a long time but when I ordered mine I figured it's better to be like #999 in line than not in line at all
It's this dude & his son. They're nice people. They email you about every week letting you know what they're expecting soon, what other stuff they have, & updates on what you ordered
Originally I ordered a Gigabyte vision 3070
but they emailed me like ya we're getting the EVGA, no ETA on the one you ordered yet but for like $60 more or something I got the EVGA when it came in
I didn't even believe it was actually gonna happen until I opened the fedex box
Nice
Noicee
So ya try them out if you're tired of getting the newegg "you weren't selected" email
I was super sketched out at first
I was looking up businesses in Ohio trying to see if it was legit
I dunno if you ever went on that Falcodrin or whatever dude's discord where he has like a live feed of where cards are in stock
but they all told me I was nuts when I said I was gonna order it form there
lol
hairstonjake@gmail.com is the Dad's email. You could probably hit him up & see how long he thinks it would take to get one.
Good guys
You know of them??
I mean they sound like good guys
Haha ya they're like super Christian, which regardless of your feelings on that, means they should be honest
I feel like they were with me tho
So does 5900x OC better or worse than 10900k? I thought Intel chips generally had more OC potential or something
worrying finding... running a previous profile that passed tm5 anta777 full cycles failed with no changes
sfc reported no problems
should I do a fresh reinstall of windows? is tm5 not reliable enough?
Is your ambient temp higher than when it passed the test before?
What voltage are you running?
B-die I'm guessing then
rev. e
Oh interesting
Rev.e is good for high frequency but not really known for tight timings like b-die, as you've no doubt noticed by now.
You're sure your fclk is stable at 1933?
stable up to 2000
tested with linpack xtreme
dropped to 1933 due to performance loss at 1967 and above
Well sure mine is stable enough to go to 2000 too but I start to get corrected errors at 1933
like wheas?
No those are the uncorrected errors
There's error correction built into the IF and chipset
Your performance drop is due to the corrected errors
I did bench 1867, 1900, 1933, 1967, and 2000 at least 10 times each
all with the same timings
1933 consistently had the highest performance beyond margin of error
and for what it's worth I've passed at least 30 runs of tm5 anta777 with 1933 fclk with various timings
Fair enough but you're probably running on the edge of stability
I'm now wondering if tm5 anta777 is insufficient or if something else is going on
any timings immediately stick out to you?
I had previously set the profile I screenshotted as stable due to it passing
then tried to tighten tWR
tRFC has to go lower than that right?
got kind of suspicious when it had zero headroom for tightening
then put it back auto and to my surprise it failed
tRFC is still on auto
I've only gone up to tWTRS and tWTRL
tRAS and tRC also can go a little lower but doesn't perform as well
300 ns tRFC should work, if not tighter
nope
Weird, I've seen some rev.e kits go down to 8
goes down to 10 but errors out
well
boots down to 10*
yeah I mean this kit seems kinda weird
tRTP cannot go below 14 at all and I'm unsure that tWR can go below 26 at all
then again I guess literally everything is uncertain now if I'm not even sure that the stress test I've been using is good enough
That doesn't seem too unusual at least for that frequency
really?
my old kit runs tRTP 8 and tWR 14 at 3733
I stopped testing tWR to investigate other things after it errored very quickly at 20
26 is auto
If it was 3600 tRTP at 10 is about normal, tWR 20 (at around 1.4v)
You may have a weird bin or something
so tWR 14 is abnormally low?
Maybe, but you do have more voltage so 🤷♂️
this is what i'm currently running
this has been rock solid stable for several months
hmm actually maybe I should try increasing procodt back
I have it at the absolute minimum 28.2 ohms
I wonder if higher procodt will allow me to tighten tWR
any timings that look like they're on the "edge"?
tWTRS?
What about trfc2
Hasn't touched tRFC yet
I see
290-300 ns should be fine when we get there
tCWL 14?
I thought running tCWL at the same as tCL is safe
auto sets it to 14
I have not touched that
Yeah 14, it's at 12 in the screenshot?
Ohh got that mixed up then
try upping trrdl to 5
I did that before I even did primaries
Whoa why is clkdrvstr so high
if that wasn't stable I'm sure it would have manifested in the 30+ full runs that it passed
120 ohm clkdrvstr was the only way to get 3867c14 stable
with all other timings on auto
errors at 24 and 60 ohms
fclk stability was tested independently at cl28 or something like that
Well if it works...
I saw it suggested on the ddr4 oc guide that raising it to 120 can help get GDM OFF stable
so I tried it and it worked
it's at 120 ohms on my old system as well, seems to work
I couldn't tell you what it actually does though lol
Oh I didn't notice GDM off
It's the clock drive strength resistance circuit, directly related to memory and fclk speed.
That's the extent of what I can tell you though lol, I'm no engineer so I can't tell you what role resistance plays.
I am now investigating whether or not raising procodt to 40 ohms will allow tighter tWR
That would have been my next suggestion, 28.2 seems low
and if not 40 then 53.3
since my other rig runs 53.3
that's auto though since I didn't even know what procodt did when I oced that one
You might even set RttWr to disabled
for the rtts I just followed arshia's suggestions otherwise I'm not sure what those do exactly
Samsung yeah
Thanks I'll try those
Jeez don't scare me like that, i left tm5 running while completely out of the house yesterday lol
tl;dr - bench on a $20 ebay ssd
@orchid flame This is my current C die OC. Still under progress, but completely stable with these timings
Alright will use my old sata ssd to bench then, tyvm
G skill right?
Ripjaws
Macrium reflect or acronis are popular within the hwbot community for saving images of benching OS's in case you bork one overclocking
Or just clone the drives with dd if you're a linux nerd
I borked 1
Im starting fresh
hehe
I just moved to nvme and i haven't deleted my old win 10 off the sata so no need but thanks :)
If/when you do do fresh installs, look into using ntlite or WSIM/SCCM/similar to customize your iso https://www.youtube.com/watch?v=PdKMiFKGQuc
Secure your online accounts, and manage your passwords with NordPass. Download it for free by visiting https://NordPass.com/Craft, and a huge thanks to NordPass for sponsoring this video.
Let's face it... Windows Pro hasn't been a 'professional' operating system since the Windows 7 days.Between the lack of control over updates, and the 12 versi...
Saves a lot of time if you do this often either for benching use cases or actual productivity
Thanks
Oh i will. This looks cool.
Modified iso/answers file + chocolatey setup script/xml is the way to go for quick, disposable installs
I always keep ntlite handy, but I found out the hard way that my audio driver interferes with it starting since Nahimic wants to hook into EVERY program for the audio effects, even when they're disabled... Sounds kinda like a virus doesn't it?
TBH I've never heard of anyone using ntlite outside of the hwbot community, all the sysadminds use sccm/mdt etc
Really? I've used it since I started modding Windows 7
LOTS of good mods to make on that OS for performance gains
I've heard that stock win7 is quite good for efficiency when benching
Have you seen blackviper's tweak page?
http://www.blackviper.com/
Welcome to my little spot on the internet. My name is Black Viper and if you are looking for Windows Services information, it is here!
It's ok at stock but there's plenty of services to disable and optimizations to make
You should hwbot lol
I probably should
With how stable my memory is I've been considering pushing more voltage to get cl13 gdm off and tighten up a couple things
Maybe I can push the clock to its limit too on the stock spreaders
I've noticed, so many little things I've picked up and used
Alright I made my hwbot account and joined Wegg
Awesome
@proven canopy hey
Cool news
I’m modding my HOF 3090 tmrw
Capacitor mod for both input and output
Plus, on top of that, I’m changing cooling
AND
What bios do you run
I’m doing smth unusual
SCLs 4
Perhaps change your Rtts a bit
RZQ/2 or RZQ/1 park aka 120 ohm and 240 ohm
And
Wr to 80 ohm aka RZQ/3
Thanks for the tips
Wdym, the rig with it in?
Will try when I get home
Or do you mean the VBIOS
I think he means the vbios
O no mods?
I’m getting an EVC2SX integrated into the card.
Connected to a raspberry Pi at all times
Wats that
Which, if it receives overrides from my Pc, will change the settings
And if it doesn’t, automatically resets the EVC2 settings back
Reason: Bad word usage
It’s perfect, the ultimate card
I like how u dont even care about the warns anymore
The best PCB, with added on best cards, new auto debug functionality, dynamic voltage tweaking on the fly, custom cooling
Hopefully it’ll be even better when I migrate to HC500A daily
Uhhh
Could ya dumb that down for me
All I got was raspberry pi
And voltage tweaking
And custom cooling
Look up Elmor EVC2 and you’ll understand
Maybe
So forks, self resolving EVC2 setup to a card
This is gonna be so good for daily istfg
Aiming for -20c on the chiller setup
Got it
Its a switchy adjustir doodad
It’s a device to override control of the PWM controller
Btw @orchid flame the timings i showed earlier were good for samsung c die right? (18 19 19 19 40) only adjusted primary and tcwl so far rest were auto
Mine cant go below 18 20 20 20 40 at 3600mt/s
I'm gonna take that as pretty good then lol
Tbf I have other things tightened as well
I saw advice from someone saying tWR benefits from being a clean divider of tRFC and that tRTP benefits from being half of tWR
is there any truth to that
Reason: Bad word usage
no
I didn't believe it but I wanted to check
tRFC literally affects no timing other than tREFI
can't even tune tREFI on ryzen :(
||Alatron gonna come and be like “well actually it does affect the self refresh timings” as revenge for all the pedantic stuff I do istfg||
how does it affect tREFI?
is it really possible for one rev. e kit to require double tWR compared to another rev. e kit or am I just doing something wrong
well I haven't tried 22 and 24 yet cause I need some sleep
but 20 errored out
just let the previous profile test for now to check again if it's stable
auto is 26
anything I could do that might have an effect on tWR?
1.548 vdimm
0.694 mem vtt
I overclocked my 1070 ti and didn't realize that afterburner gave it the ability to use a lot more power, I checked the temps and it was at 91C I got it to 85C is that ok?
85c is fine under load
If you want you could undervolt or just use a more aggressive fan curve
What that be?
Custom highly stripped win 10 iso creation
Uwu
Cut it down more
i have opted for undervolting and its now to 81C with slightly higher fps
Nice
By now I'm testing how low I can go and be stable and damn this thing doesn't need much
Are you doing undervolt oc?
Also ya make sure you dont loose perf
Yea I am and so far fps has only raised a bit but that's slowing now
👌
78C with 144+ fps in BeamNG with best settings, damn this card is great
Nice
turns out my board is pretty good at training auto tertiaries
I can't touch any of them lol
I was only able to lower 2 of my tertiary timings, if I touch any of the others I need to do cmos clear lol
I'm not done going down the list yet but most of them can't tighten below auto
even a lot of the secondaries can't
I just had an oc pass tm5 and then hard crash while running aida
I'm starting to have my doubts about whether or not tm5 is actually a valid stress test
happened again
What aida config? Memory only 80%?
there's a config?
I meant what boxes did you check
uhh let me see
I didn't touch any settings in aida
a stable system shouldn't be hard crashing though no?
my other system never had hard crashes from aida
Like this
ah
is there something wrong with it or is there something wrong with tm5
Are you running the full aida memory benchmark? Or just memory? e.g. right clicking on start benchmark and skipping the cache tests
oh I didn't realize you could do that
I've been letting the whole bench run
does that matter? should cache be unstable with cpu at stock?
Regardless - I would just keep testing individual stress tests and try narrow down what's unstable
Then if nothing, figure out what combination crashes it, maybe hotter IMC becomes unstable
idk, never used it
okay
Using multiple tools to test stability is something I always recommend anyway
check with occt memory, the different instruction sets are better at determining if its imc volts, or timings that are off
arshia recommended using linpack xtreme for 1 hour to test imc instability
this is what passed tm5 but hard crashed aida
currently testing linpack using tRC 56, procodt and rrts on auto, tWTRS tWTRL 4 8
oh and tRP 14
any other recommendations?
120 ohms was needed to get 3867c14 stable with all other timings on auto
24 and 60 ohms failed
apparently raising it helps stability with GDM OFF
i don't know why or how but it seems to work
my other rig also has it at 120
damn that board must really be struggling
my b550 aorus elite also needs that to run gdm off at 3733c14
that board is quite bad too
But yeah, optimal drive strength really depends on load
It needing to be high means there is a lot of load
120 is what i needed to reliably pass tm5
so is it like a bad thing that i should worry about
or it's fine
it's fine
okay thanks
high enough to obliterate the cpu and ram in a few microseconds
So like 2V
even getting it at 2V would be very difficult
hmm something makes me think i won't be running that for a daily oc
I'd hope not
on a Ryzen 5900x is CPU Die the most accurate temperature reading?
Ehh should be?
what about voltage? I don't see VRVOUT on HWiNFO
Ehhh there should be total package voltage etc
Your cpu usually shows twice in hwinfo, look at the 2nd tab of it
okay thanks I'll look more closely. I didn't know if that an an Intel thing or something.
Np
Are you Ryzen or Intel?
why did they skip 4xxx? lol
But if you send a sc im sure it's pretty similar on intel
They had 4xxx laptop lol
I just got a 5900x
Idk
Nice
weird I tried setting clkdrvstr to auto and it autos to 24 ohms now
gonna see if it's stable I guess
before I had any timings dialed in except cl it needed 120 ohms to not error
Thanks. Bending PETG is difficult
it's a board thing
the svi2 sensor is correct
okay cool. The board is ROG Strix B550-F Gaming if that helps at all
Imagine not having a 3600x
Ik smh what was i thinking
@lavish tundra what's your 3600x scores in timepsy like btw
Lemme check gimme a few
Kewl i jw cus i gained 100 point after my 3600 18 19 19 19 40 (tcl 16) aint touched anything else oc
Is 8050
Get pwnd
4.425ghz all core
1.25v
:( i only got 7100 or something but need to work on ram oc, also i just use pbo
I also need to work on my ram OC
Arshia gave me some juicy pointers
I not home and was leaving house for work when pic was taken on monday
You can't get any of those to 19 no?
I still needa try make some of my 19s 18
Last time I tried I couldnt

came downstairs to see tm5 still on cycle 1 after 6 hours and not using any ram
💀 rip 
side note, pog
I overlocked my GT 1030, and got 3 more fps in minecraft 
What the timings
xmp, except with tightened tfaw lol
that all you can do? dam
@orchid flame when you recommended these adjustments, I'm looking at soc voltage in bios but had it set to 1.1v or so i thought i the bios
Had soc at 1100mv i assumed?
Because can't enter a . Into that box
Or is that a different thing i should be changing
Yeah i had that figured out
Ah
Just confused cause vscoc in zentimings was 1.085v
But I'm guessing that's not soc voltage
Is it vcore soc?
Any idea why this shows my card as a 3070Ti instead of just regular 3070?

