|
United Kingdom20278 Posts
A friend helped my OC my CPU to 4ghz last year, but i had problems a little while back that i couldnt figure out, so i reset it to 3.06(3.2)ghz because i thought it was a potential cause, but it turns out it was just windows causing mass bluescreening for no reason and it was fixed with OS reinstall.
Anyways, im gonna overclock it again, all the cooling etc is set up, a little worse than it could be (not stable at 4ghz w/ prime95 unless hyperthreading is disabled or its the middle of winter) but hyperthreading is worthless in sc2 and skyrim, and i have also noticed xsplit running on hyperthreaded cores and lagging out stream unless i restart it a few times and it runs on 2 logical cores when assigned them, which is annoying.
So, what should i be aiming for? Im looking up my RAM voltage atm, there seem to be a lot of different ideas around the web about the CPU and LGA1336 is a bit outdated now, wasnt sure where to post a thread to get someone who actuly knows what they are doing, so i guessed TL was the place.
From what i remember ~1.35vcore is about the max i could go for temperature reasons, was using 200x20 OC before (is 167x24 ok or suggested for any reason?) I know the basic stuff, but i dont really know what to go for in terms of qpi/uncore voltage, probably another setting or two. My RAM is rated @1600mhz, 1.65v
Im 30 mins into Prime95 atm, 64c max @3.2ghz and whatever voltage the default turbo uses for reference and also using a gigabyte motherboard.
Il reboot now and take a look at the settings, could prob just throw in 200x20 1.35v and auto the uncore voltage, see if it boots, why not right?
Greatly appriciated though if anyone who actuly knows what they are doing posts here 
|
Netherlands19129 Posts
|
United Kingdom20278 Posts
Thanks, i didnt really want to post it there incase it was considered a random bad thread or something
|
United Kingdom20278 Posts
Aparantly most of the 1336 cpu's had multipliers on cpu, qpi and RAM locked, to confirm, my cpu/motherboard doesnt
|
multi's aren't unlocked. even if you try to change it in your mobo it won't do anything.
|
On November 21 2011 09:45 mahnini wrote:
multi's aren't unlocked. even if you try to change it in your mobo it won't do anything.
The only 1366 cpus that are unlocked are the extreme series processors. Your CPU can hit 4.0 easy, and 4.5 with nice watercooling. If you got a good chip 4.5 is doable with watercooling.
|
United Kingdom20278 Posts
Well, ive got an i7 950, and ive booted before at 167x24 and validated CPU-Z even @4008mhz. The reason i posted that was because people were saying most cpus were locked on some other website but some were not, i think this is one case?
Anyhow i just booted at 200x20, massive problems aside from my RAM randomly being set to 7-7-7-20 latency (manually set 9-9-9-24 1.6v) and forgetting to use the adjust vdroop thing for 1 attempted boot, gonna go run cpuz and try prime95
Also, had to set qpi voltage to auto because i couldnt boot with that, what should it be set to? someone was mentioning 1.3v in a guide while ive heard in others it should be >=vcore, my motherboard has different colored options past ~1.33 but i cant boot with ~1.3 and i dont want to throw like 10 different voltages at it and continuously restart for no reason when i dont know what im doing
|
United Kingdom20278 Posts
Crashed 20 seconds into Prime95 at ~78c with the bluescreen message "A clock interrupt was not received on a secondary processor within the allocated time interval" - Got this a long time ago, im pretty sure that is why i was eventully running 167x24, it was narrowed down to motherboard not handling higher baseclocks well or something? Im not sure, but il test the 167x24, i dropped memory and qpi multis to acceptable levels and booted straight in without issues
edit:
I lost a multiplier to turbo mode (which i have disabled) but that is an easy fix, going from x23 to x24 on same voltages wont affect temps in a massive way as far as i know, heres a validation anyway
![[image loading]](http://valid.canardpc.com/cache/banner/2108612.png)
Stable @85c 15 mins in, temps higher than i would like, i think they could be probably be lowered by taking the qpi volage from auto but i cant get it to boot without it, any suggestions?
|
81 is too high for first gen i7's. The general rule with first gen i7's is dont go over 80 for a 24/7 OC. You're going to need to get better cooling IMO. Hyperthreading adds 5-6 degrees to your load temperatures, if you dont have a great heat sink i'd recommend keeping it off and getting a higher stable OC. Also I'd turn turbo off if you are using the max multiplier of 23.
I get to 3.8Ghz without any voltage changes on my 950 (166x23, with 1333ram, with ram at the lowest multiplier so you might be able to get the same). I'd recommend you try that. Make sure to under clock your uncore multiplier, you can increase it later but it's a very minor performance increase. I'd stick to keeping the bclk low and use a high multiplier but some other configurations will work better. Just dont OC your ram if it's not high end.
I'm 4.2Ghz at 1.26v and I get 70max with a Noctua NH-U12P SE2.
And to Mahnini, all i7's are unlocked...
What stepping is your 950? is it a d0?
|
United Kingdom20278 Posts
On November 21 2011 10:21 Phayze wrote:
81 is too high for first gen i7's. The general rule with first gen i7's is dont go over 80 for a 24/7 OC. You're going to need to get better cooling IMO. I get to 3.8Ghz without any voltage changes on my 950. I'd recommend you try that. Make sure to under clock your uncore multiplier, you can increase it later but it's a very minor performance increase.
I'm 4.2Ghz at 1.26v and I get 70max with a Noctua NH-U12P SE2.
And to Mahnini, all i7's are unlocked...
Thats on Prime95, xsplit will usually be around ~70-80% CPU because you need to leave it plenty of room (if you ever hit 100% it will drop frames and cause system-wide stuttering and performance issues, even with affinity only on ~2 cores), I doubt it will go anywhere near those temps during actual usage.
I also heard that you was fine below ~90, but il take that into account, thanks.
What should i put Uncore at?
What about qpi/vtt voltage?
Sorry to say i dont even know what stepping means
|
Just keep your uncore stock (or close too it by downclocking it after you change your bclk) for testing stability. It barely impacts performance anyways so it's not worth the heat. Qpi doesnt really matter. It's always going to be alot higher, I personally am not given much options to downclock qpi/vtt.
Start by only going for CPU speed, downclock everything else. Work on those after. In general stay under 80. But i'm sure you would be fine with 85. Dont feel like you need to use the high multiplier, if you have a decent mobo you could find a better clock with a lower multiplier.
And the general recommendation is never go over 85 with first gen i7's. Not sure if it was just a stigma or what was with it, but 90 is definitely NOT the number. The intel spec rated temperature was alot lower for first gen i7's compared to the old quads.
Anyways, theres more i7 overclocking guides on the internet than any other generation of processors as far as i've seen. You'll find better information in those written guides then here.
I get 3.8Ghz at stock, 1.35V for 3.8Ghz on a D0 is wayyyy overkill. Try 1.22-1.26V. I think you're not stable because you're overvolting (and in doing so going over 80degrees)
Stepping is just the series of production the processor was made in, a0,b0,c0,d0, d0; d0 was the latest and most power efficient i7 near the end of their lifespan, and with that was the best overclocker.
|
United Kingdom20278 Posts
On November 21 2011 10:25 Phayze wrote:
Just keep your uncore stock (or close too it by downclocking it after you change your bclk) for testing stability. It barely impacts performance anyways so it's not worth the heat. Qpi doesnt really matter. It's always going to be alot higher, I personally am not given much options to downclock qpi/vtt.
Start by only going for CPU speed, downclock everything else. Work on those after. In general stay under 80. But i'm sure you would be fine with 85. Dont feel like you need to use the high multiplier, if you have a decent mobo you could find a better clock with a lower multiplier.
And the general recommendation is never go over 85 with first gen i7's. Not sure if it was just a stigma or what was with it, but 90 is definitely NOT the number. The intel spec rated temperature was alot lower for first gen i7's compared to the old quads.
Anyways, theres more i7 overclocking guides on the internet than any other generation of processors as far as i've seen. You'll find better information in those written guides then here.
Honestly i cant even remember what stock was for uncore and having a little trouble looking it up, stupid problem i know.
There is a flat option in my bios for qpi/vtt voltage which ive been leaving on auto because i have no idea what to put it on, thats why i mentioned it, i want it as low as possible for stability and ive heard auto gives a massive amount of unneccesary breathing room.
I went to 167x24 because my mobo was failing @200x20 and i wasnt sure what it would be stable at, do you have any other suggestions for it?
It is infact a d0
|
Lower your voltage! Stock voltage for i7-950 is 1.181V! You should have no trouble hitting that 3.8Ghz stable at 1.22-1.24V!
|
United Kingdom20278 Posts
On November 21 2011 10:39 Phayze wrote:
Lower your voltage! Stock voltage for i7-950 is 1.181V!
Ok ok i get it shit is it that bad
|
Not really. There should be no damage, just 1.36V + is used for like 4++Ghz overclocks with water cooling setups.
Try 166x23 (3.8Ghz) at like 1.21V or 1.23V, and test for stability. You want to keep QPI/VTT voltage .02 below vcore. So if you are at1.23vcore, try 1.21Vtt. This is because by overclocking your bclk you also overclock your qpi. So the voltage has to be adjusted as such. You should easily be able to stay well under 1.3V under 4ghz with a d0.
|
United Kingdom20278 Posts
On November 21 2011 10:40 Phayze wrote:
Not really. There should be no damage, just 1.36V + is used for like 4.8Ghz overclocks with water cooling setups.
Try 166x23 (3.8Ghz) at like 1.21V or 1.23V, and test for stability. You want to keep QPI/VTT voltage .02 below vcore. So if you are at1.23vcore, try 1.21Vtt. This is because by overclocking your bclk you also overclock your qpi. So the voltage has to be adjusted as such.
Running that @1.265 (i restarted before your post) and qpi/vtt still on auto, testing now
|
Good, you booted with the same overclock, but .1v less. Don't start voltage high and lower it. In general start voltage low and add more based off stability. D0's are OC beasts. Turn HT and turbo off. Turbo will mess with stability if you dont know what's stable and what's not. But thats a personal opinion. I have HT on, turbo off. I personally like stable performance.
|
United Kingdom20278 Posts
Its stable @ 75c after 5 minutes, will probably go up a couple degrees but its still a massive cut, should i drop a bit more along with qpi/vtt?
I cant seem to add a 24'th multiplier because of turbo mode "reserving" it or something, so i think im gonna enable that too on restart
|
Yes, 24th multiplier is reserved for turbo, and only available on the 980/980x+ series. The gulftowns. I like turbo off. 1.26V should be fine with turbo on though. I'd personally keep lowering the voltage with turbo off till around 1.21V or whenever you're unstable and find the minimum voltage you need to run 3.8Ghz. Having turbo on will require more voltage, and doing it this way you can easier identify correct estimate voltages for a higher overclock. 4-4.4Ghz is easily possible with a d0 assuming you stay under 80. Staying at 3.8Ghz is fine, but I feel you are still overvolting a little @ 1.26v.
|
![[image loading]](http://valid.canardpc.com/cache/banner/2108647.png)
Just loaded an old profile. 1.2V with HT on, 3.8Ghz. I used it for months when I first got my 950. Most certainly stable. I was able to use stock voltage too but if I recall correctly I had crashes after like 10 hours of p95. This is why I feel 1.26V is still a little too high.
Once you get your lowest stable voltage at 3.8ghz, start raising the bclk and voltage incrementally. Test for like 5 mins p95. Get to the clock you want, then start dipping back the voltage until your are unstable again. Once you're there just dip back the bclk or increase the voltage a tiny bit. Remember to keep qpi/vtt .02 lower than vcore!. It seems like you (or your friend) just randomly chose 1.36V as a good number and put in a random OC. Since you got a good 3.8 ghz OC almost (imo voltage too high) , it acts as a good reference. As you barely need any voltage to hit 3.8/4Ghz, and when you start going over 4 that's where you're going to need to put in more power.
|
On November 21 2011 10:21 Phayze wrote:
81 is too high for first gen i7's.
They aren't first gens..... But yes the temp is way too high even though its a prime 95 temp.
|
United Kingdom20278 Posts
That was ugly.
Finally narrowed crashing down to turbo being enabled, disabled it again, booted back normally... odd, intel turbo boost (1 multiplier) causing like a dozen varying bluescreens, was confused because i just ran 15minutes p95 and couldnt even each welcome screen with the exact same settings.
My uncore/qpi is lower now, il run tests to see if it lasts 30 seconds then post back
|
United Kingdom20278 Posts
On November 21 2011 11:18 Boblhead wrote:They aren't first gens..... But yes the temp is way too high even though its a prime 95 temp.
Yes they are
|
United Kingdom20278 Posts
Stable, changes seem to have no affect on temps though. It seems i cant set a qpi/vtt voltage without instant bluescreen, i tried quite a few steps between 1.1-1.3 with zero results, its still on auto.
This is ~77c max p95 in ~18c room temp with hyperthreading disabled... Maybe my cooler is worse than i thought? Its the Xigmatek Dark Knight
|
As I said, focus on clock, leave turbo off. You can keep HT if that's what you want in your final OC (for streaming etc). And try to keep uncore stock. whatever the uncore is at stock clocks just keep down clocking it so it sits around there. You can play with it after you achieve your desired OC. Turbo makes it incredibly hard to get the correct lowest possible voltage when overclocking.
And as for first gen etc, I've always seen lga 1366 classified as first gen, 1155 i7 second gen. It's really irrelevant though. Anyways you even said that the multi's are only unlocked on the extremes, and that is blatantly false information.
|
United Kingdom20278 Posts
167x23 (3.84ghz?) runs fine, sits around 77c, but it is again bluescreening with "A clock interrupt was not received on a secondary processor within the allocated time interval" as it does when i try to run 200x20, i bumped voltage back to 1.3 to see if it would help as i ran these settings fine with 1.35, maybe my CPU just turned out shitty and cant run as low as yours... Ive definatly never heard of anything near 3.8ghz on 1.2v.
Doesnt uncore have to be kept at >2x RAM clock, or am i mixing it up with something?
|
Your VTT is still on auto is it not? It has to be withing .02v of CPU vcore. I'm at 1.184V at 3.8ghz and running p95 right now. I'll post a SS of my temps etc in 20 mins or so. I dont think it's your processor.
|
On November 21 2011 11:18 Cyro wrote:Show nested quote +On November 21 2011 11:18 Boblhead wrote:On November 21 2011 10:21 Phayze wrote:
81 is too high for first gen i7's. They aren't first gens..... But yes the temp is way too high even though its a prime 95 temp. Yes they are
I consider the 1156 lynnfields to be first gen 
|
United Kingdom20278 Posts
0.02 is a very small margin, il go and set it now.
|
if your CPU vcore is at 1.26 then your VTT has to be at 1.24 or 1.25. Also make sure your ram is not exceeding its 1600mhz, if it is downclock your ram multiplier. Majority of ram cannot be overclocked with any good result.
http://www.bit-tech.net/hardware/cooling/2011/03/21/xigmatek-dark-knight-review/2
Seems like your cooler does suck quite a bit ;/
Yes uncore is 2x ram clock, if you dont overclock your ram, you leave uncore stock while overclocking(by manually lowering your multiplier, perhaps keep it 1x more than 2xram clock). Dont bother overclocking your ram, find multiplier/bclk combinations that keep you at 1600mhz.
Anyways, heres my 3.8ghz profile after 15minutes of p95
+ Show Spoiler +
I dont see how you need anything remotely close to 1.3V to hit 3.8 Ghz with a d0 950. I dont see how that is possible even with the worst possible binned chip. You must have something set incorrectly. I have never seen any 920/950 not hit 4ghz with ease.
|
United Kingdom20278 Posts
Im underclocking RAM with the multiplier, didnt want to OC it for no reason (cant keep very close to 1600) and RAM speed has almost 0 effect on anything from what ive read
With 1.25vcore my system wont boot period the end before 1.265 vtt, i bumped it to 1.272 or something for stability as i barely reached welcome screen first time, i eliminated most other factors and that 1 voltage causes all my crashes below that value if it is off auto
|
United Kingdom20278 Posts
Thanks for the screenshot, i dropped my qpi and uncore a bit and set 1.2v as it occured to me i dont think i have actuly had any vcore related crashes yet, and it booted fine with qpi/vtt voltage on auto (cant get it to boot at all otherwise, ive tried like 15 different steps, crashes every time, and boots fine when i set it auto)
running tests now. The cooler was bundled with the CPU with a 4ghz w/ hyperthreading guarantee, guess im a fool for not looking into it further at the time.
|
Ok, this is my settings.
QPI clock = 36x
Uncore x16 (2667) for my 1333ram
BCLK 166 (166x23)
mem multi: 8.0 (1333mhz ram)
1.22500 bios vcore
I think I was wrong, qpi/vtt has to be 0.1V to 0.2V lower. not .01V to .02V
Looking at my bios, stock is 1.1875V for vcore and 1.175V for vtt/qpi. So perhaps you qpi/vtt is too high. Make sure you are downclocking your qpi/vtt, this is probably why auto qpi/vtt is letting you boot stable, because you have not lowered that multiplier yet.
|
United Kingdom20278 Posts
On November 21 2011 11:58 Phayze wrote:
Ok, this is my settings.
QPI clock = 36x
Uncore x16 (2667) for my 1333ram
BCLK 166 (166x23)
mem multi: 8.0 (1333mhz ram)
1.22500 bios vcore
I think I was wrong, qpi/vtt has to be 0.1V to 0.2V lower. not .01V to .02V
Looking at my bios, stock is 1.1875V for vcore and 1.175V for vtt/qpi. So perhaps you qpi/vtt is too high. Make sure you are downclocking your qpi/vtt, this is probably why auto qpi/vtt is letting you boot stable, because you have not lowered that multiplier yet.
My uncore is ~3k for the same RAM speed, QPI is also the same. I just booted with that vcore and it crashed the moment i started prime95, so i bumped it back to 1.26 and its running fine now, but havnt really made progress.
What is your exact qpi/vtt voltage? You didnt mention it.
I want to hit as high clock as possible without hyperthreading i think, skyrim wont max cores but it scales almost 1:1 with cpu clock speed increases, 20% OC 20% fps gain etc, someone did a benchmark, and also obviously because hyperthreading is useless to sc2 and seems to mess up with xsplit.
I ran on 1.35v for over a year which would pass 90c in prime95 with hyperthreading pretty easily, but it never crashed out because being maxed for 5-10 minutes on 8 threads just never happened in any real world situation, i dont think ~77c is too high to stop at heat-wise for that reason.
Il lower the uncore a notch but aside from that should i look into raising bclock?
|
1.185V. I just rebooted and set it on auto and I still had no problems. Perhaps try a memtest?
|
United Kingdom20278 Posts
|
Yeah, it should be lower than Vcore. I previously said by .02, as it's been a while since I even though about this. But based off the default stock clocks the qpi voltage was .1V lower than the stock vcore, so i'm guessing I was wrong and it's actually supposed to be 0.1V and 0.2V below cpu vcore.
|
United Kingdom20278 Posts
Every attempt i make to boot with under ~1.25 vcore is met by a blue screen on desktop before i can even touch p95, but if i run @ 1.26vcore it is stable for ~20mins (couldnt test more yet) so i think i cant go lower than that, still having issues with the vtt voltage but whenever i could run at all on lower voltages it seemed to have no effect on temperatures
|
United Kingdom20278 Posts
On November 21 2011 12:08 Phayze wrote:
1.185V. I just rebooted and set it on auto and I still had no problems. Perhaps try a memtest?
you think that would be relevant?
I have no idea how memtest works, it appears to be pre-boot?
Anyway im stable now @72c after 10 minutes, one of my case fans was off aparantly
|
Good to hear. Let it run overnight. What voltage did you settle at?
|
United Kingdom20278 Posts
Anything below 1.26 will bluescreen within 30 sec of reaching desktop, even @ 3.2ghz, so i bumped it up to 1.3 to be sure of stability anyway as ~75c p95 max is like 15c lower than i had before and it was fine even then.
Im pretty sure your motherboard and/or CPU is just a ton better than mine, cooling too, im gonna make sure to get either an NH-D14 or if i dont have the space, a high end water cooler with my next build (ivy bridge? not even sure if there will be quads etc)
|
you wouldnt happen to have all the power saving features enabled would you?
|
United Kingdom20278 Posts
On November 21 2011 13:23 Phayze wrote:
you wouldnt happen to have all the power saving features enabled would you?
None of them... Its a pretty horrible result (3.8ghz w/o hyperthreading) but i guess im stuck with it when i cant move bclck above ~170 without instant bluescreen at any voltage, nor move multiplier past 23, and hyperthreading gives me >10c more heat, which i cant really afford either
|
United Kingdom20278 Posts
It bluescreened. Same bclk, mult, RAM, uncore, qpi settings as you, 1.3 vcore, qpi/vtt on auto... I was running @4 with hyperthreading for almost a year using 1.35vcore, qpi/vtt on auto... a bit hot but it never actuly passed ~80c outside of benchmarks, just doesnt work anymore.
I think my CPU got permanantly damaged or something
|
On November 21 2011 11:34 Boblhead wrote:Show nested quote +On November 21 2011 11:18 Cyro wrote:On November 21 2011 11:18 Boblhead wrote:On November 21 2011 10:21 Phayze wrote:
81 is too high for first gen i7's. They aren't first gens..... But yes the temp is way too high even though its a prime 95 temp. Yes they are I consider the 1156 lynnfields to be first gen
they are part of the generation of architecture. if anythign the bloomfields came first.
On November 21 2011 13:44 Cyro wrote:
It bluescreened. Same bclk, mult, RAM, uncore, qpi settings as you, 1.3 vcore, qpi/vtt on auto... I was running @4 with hyperthreading for almost a year using 1.35vcore, qpi/vtt on auto... a bit hot but it never actuly passed ~80c outside of benchmarks, just doesnt work anymore.
I think my CPU got permanantly damaged or something
some people experience degradation with high voltage overclocks so that might be what's happening in your case.
|
Yep, your friend was quite stupid to have you running that processor at 1.35V for a 4ghz OC on a d0. Look up your motherboard online and see what OC's other people achieved. It could be a limitation of the board just not being able to handle the higher bclk. So at like 1.24V Vcore and 1.22 QPI/VTT is instant BSOD at 166 BCLK? It could be a case of TOO much power. Like remember, stock i7-950 is 1.181V. Google the BSOD code. You want to look whether it's caused by memory or the OC (power related/failure).
|
That CPU cooler should be enough, have you tried reseating it w/ fresh thermal paste?
Each CPU is different, if yours got damaged or not you should probably start from scratch on it's overclock:
If you're new to overclocking, you should probably leave all the voltages except vcore on auto.
with good 1600 memory, the RAM overclock will be limited by the IMC in the CPU, not by the ram it's self. IIRC those CPUS were only rated to run RAM at 1066. but for ease of overclocking, just load the XMP then manually override the memory multiplier down to 8.
If you're aiming for a substantial OC, you should disable Hyperthreading and all the power saving features of the CPU.(HT is like adding 25% increase in performance for highly threaded environments, it generally won't help as much because you'll hit a higher cooler OC without it. It won't help at all for sc2) If you want a stock voltage OC, you can leave HT on.
set vcore to 1.2v, CPU multipler to it's highest, then starting at about 150 increase BCLK in increments of 5 until you can't pass 5 minutes worth of prime.
Then start incrementing vcore very slowly, until you get stability back, then BCLK till you lose stability.
Repeat until you hit your maximum tolerable voltage/temp(I'd stay under 1.3v and 80°C in prime were it my machine), or you hit your desired OC, or stop when increasing voltage doesn't improve stability.
Now you have your highest 5minute stable CPU overclock at that multiplier, try reducing CPU multiplier by 1, and increasing BCLK by about 10-15, to see if you can get a higher stable overclock with that multiplier. after you make that call, you pick your multiplier and start fine tuning the stability:
drop BCLK by about 5 from the highest 5 minute OC, run prime for a couple hours to see if it's stable, then repeat the above process with smaller increments and 1 hour stress tests to narrow down the ideal settings. When you think you're done, run a 24 hour torture test to make absolutely sure it's stable.
as for me, I have an i7 930 @ 3.61Ghz, 1.175v w/ HT on. I didn't bother with a huge overclock because it's fast enough to stream sc2 at stock voltages, and I prefer a cooler running CPU(less heat= less noise, and yes my CPU is slightly undervolted).
|
United Kingdom20278 Posts
On November 21 2011 16:08 Phayze wrote:
Yep, your friend was quite stupid to have you running that processor at 1.35V for a 4ghz OC on a d0. Look up your motherboard online and see what OC's other people achieved. It could be a limitation of the board just not being able to handle the higher bclk. So at like 1.24V Vcore and 1.22 QPI/VTT is instant BSOD at 166 BCLK? It could be a case of TOO much power. Like remember, stock i7-950 is 1.181V. Google the BSOD code. You want to look whether it's caused by memory or the OC (power related/failure).
I tried stock vcore @166x20 mult and it wouldnt boot either with other stuff on auto. The RAM has like 20 reviews from people stating they hit 4ghz+ with it with no problems on the same cpu's.
Thanks for all comments, il have another whack at it later
|
United Kingdom20278 Posts
Ok i read some more stuff and spent a lot of time running prime95, it seems that everything works fine aside from increasing my base clock, at ~150bclk i can run x24 (3.6ghz) with hyperthreading @1.2vcore, but as i increase base clock, at ~160 i start to hit errors in prime95 after around 5 minutes (just 1 thread here and there drops) and @166+, the whole system is just unstable, bluescreening when put under major stress. I tried everything from 1.2 to 1.32vcore and a few other voltage options, but everything works fine, no problem @1.2vcore until ~160bclk, where it suddenly goes to shit.
What should i do from now? Can i push it further, is it a problem with motherboard/cpu etc, am i maybe missing a setting etc, or just stuck @3.6ghz? Heat is fine, 80c @1.3v with hyperthreading when i tested it at that an hour ago.
Thanks for all your help everyone either way.
Im looking a lot more into vtt voltage and running memtest as that seems to be a potential problem
|
United Kingdom20278 Posts
I read some stuff about vtt/qpi voltage and base clock being related, and after another guide i decided to step back and run higher baseclocks with lower mults close to stock frequency to test for stability, it seems to be running fine atm @180 (3.06ghz) so im gonna bump up to 200x15.
More knowledge of what runs and what doesnt cant hurt right?
|
You should not have any trouble as long as you are careful with how much power you pump into the processor. If theres not enough power, it blue screens, you reset cmos, no harm done. Too much power, cause damage, then you have problems. Experiment all you want with different multipliers/bclk combinations but try to stick to multiples that keep your ram at it's rated speed.
|
United Kingdom20278 Posts
On November 22 2011 09:59 Phayze wrote:
You should not have any trouble as long as you are careful with how much power you pump into the processor. If theres not enough power, it blue screens, you reset cmos, no harm done. Too much power, cause damage, then you have problems. Experiment all you want with different multipliers/bclk combinations but try to stick to multiples that keep your ram at it's rated speed.
Yea ive always been watching RAM frequency etc, manually set it to 1.6v 9-9-9-24 and favoring going under the rated 1600mhz instead of OCing it using the mult.
Interestingly enough, it seems i need >1.26vcore to boot at higher base clocks, even on lower overall frequencies...
200 instant crashes regardless, but 180x22 @1.31v seems pretty stable (166x(>20) is not) so im gonna keep testing for a bit.
On a side note, when do i evolve into a reaver?
|
United Kingdom20278 Posts
Dropped a thread in Prime95 after 40 mins, rounding error.
Im at 3960mhz (180x22) 1.31v with RAM underclocked to 1440mhz, floating 74-80c across cores, should i bump up the vcore a bit?
|
United Kingdom20278 Posts
Had and dealt with some more issues, through a lot of trial and error i learned that my system is not stable at >~3.6ghz with under 1.3vcore, and i got errors in prime95 after half an hour or so at ~1.31, so eventuly with vtt voltage tweaks im running @166x24 with hyperthreading @1.35vcore, seems stable, I dropped P95 after 5 minutes because temps were approaching 90c very slowly and started stress testing with xsplit, hasnt hit 80c in an hour or so through various stuff, replays at x8, skyrim gameplay while streaming, etc, so i think this will work for me, it sucks i have to use such high vcore though... I can boot @4ghz with like 1.26 but it just instantly crashes when CPU gets to high load.
Hyperthreading cuts 10c off those temps, so i was considering running Prime95 with it disabled at the same voltage, would that be an accurate test of stability though? Im not sure what effect disabling hyperthreading would have on potential errors (would they not occur with the reduced load/heat?) but if you think i should il run it for like 8 hours or something.
It took me 2 hours to get a second thread to crash @1.3v with error (after i couldnt even run at 1.28) so i think 1.35 should be stable, provided temps are not an issue (and i havnt found anything capable of driving them nearly as high as p95)
|
Why do you have hyperthreading enabled in the first place? If you're increasing voltages to get a better OC, that should be disabled regardless of what programs you're running. It sounds like you wasted a lot of time with the trial and error method. While the systematic approach in my earlier post might take a couple days of work to find the limit of stability, you're going to get better results.
|
United Kingdom20278 Posts
On November 22 2011 15:51 SoulWager wrote:
Why do you have hyperthreading enabled in the first place? If you're increasing voltages to get a better OC, that should be disabled regardless of what programs you're running. It sounds like you wasted a lot of time with the trial and error method. While the systematic approach in my earlier post might take a couple days of work to find the limit of stability, you're going to get better results.
Because i dont loose anything by enabling it as far as i know. I wasnt going to use it for xsplit/sc2/skyrim anyway, actuly disabled atm.
I cant run at lower voltages anyway, because the system is unstable. 1.3v gets me some 3.4-3.5ghz, anything less will throw up prime95 errors or flat out bluescreens. I tried your approach, but i cant get anywhere near 4ghz with it, HT on or off.
It might be a problem with my motherboard (or just a shitty motherboard) or i might be missing something, but i cant get anywhere on the lower voltages mentioned. 166x23 on 1.26v instant crashes in p95, ~1.3v throws up errors, and this is 3.84ghz without hyperthreading.
My 1600mhz RAM is @ 1440mhz at rated timings and voltage meanwhile, and i tried various uncore multiplier settings but it had zero effect on stability
|
well, you can find out what's limiting you by minimizing all the multipliers, scaling up BCLK till you lose 5min stability, then go through the multipliers one at a time first just CPU, then just RAM, etc. to see exactly where the instability is coming from.
for further reading: http://www.overclock.net/t/538439/guide-to-overclocking-the-core-i7-920-or-930-to-4-0ghz
That guide works just as well for an i7 950, you just have the option of higher multipliers.
|
Well, I just got my D0 950 recently, but haven't got the time to overclock it (I'll do it in two weeks or so).
I'm shooting for 4GHz. I also have a trio of Kingston's HyperX Ti 1800, so I'm gonna ask for some early advice on what should I tweak to get the maximum from it. (as you can see, I'm not very experienced). Are memory timings > memory frequence? What should I shoot for?
btw, My heatsink is a ThermalTake SpinQ VT.
|
Would it be possible to try to run it at stock frequencies at 1.181V? I'm starting to think your processor needs the extra voltage to be stable. maybe the mobo just doesn't want to run with a lower voltage, bios update might help (with all settings at stock i might add). As far as hyper threading goes, your processor will run 5-6 degrees hotter with it enabled. If you're putting extra volts in you probably should disable it.
|
On November 22 2011 20:52 necrosed wrote:
Well, I just got my D0 950 recently, but haven't got the time to overclock it (I'll do it in two weeks or so).
I'm shooting for 4GHz. I also have a trio of Kingston's HyperX Ti 1800, so I'm gonna ask for some early advice on what should I tweak to get the maximum from it. (as you can see, I'm not very experienced). Are memory timings > memory frequence? What should I shoot for?
btw, My heatsink is a ThermalTake SpinQ VT.
http://www.overclock.net/t/538439/guide-to-overclocking-the-core-i7-920-or-930-to-4-0ghz That guide will work just as well for a 950, as for memory, you won't notice a difference in performance outside of huge spreadsheets or other niche applications so it probably isn't worth the time to optimize. as for timings vs clock, that depends on how much more clock you get from loosening the timings and the reverse. Considering the CPU isn't going to like running memory that fast, you're better off optimizing for latency. If you really want to optimize your RAM, try 7-8-7-24 @ ~1400 and work your way up.
|
United Kingdom20278 Posts
On November 22 2011 19:46 SoulWager wrote:well, you can find out what's limiting you by minimizing all the multipliers, scaling up BCLK till you lose 5min stability, then go through the multipliers one at a time first just CPU, then just RAM, etc. to see exactly where the instability is coming from. for further reading: http://www.overclock.net/t/538439/guide-to-overclocking-the-core-i7-920-or-930-to-4-0ghzThat guide works just as well for an i7 950, you just have the option of higher multipliers.
I did exactly that, 160x19 is unstable at 1.2-1.24v, up to ~150 runs perfectly, 155 threw some errors.
I have to massively ramp up vcore to keep a base clock anything higher than that stable, think i wrote before, that bclk would limit me to 3.6ghz as im limited to a multiplier of 24 (found out how to bypass turbo mode thing and run permanant)
First paragraph of your guide, "Everything in the guide should still be the same although from what I can tell there is a much larger vcore range for the 950 with some people needing closer to 1.4 to hit 4ghz."
Probably related?
Yea he goes on to say "You may now need over 1.3v to hit 4.0 ghz on a d0" which is very specific
|
United Kingdom20278 Posts
On November 22 2011 21:11 Phayze wrote:
Would it be possible to try to run it at stock frequencies at 1.181V? I'm starting to think your processor needs the extra voltage to be stable. maybe the mobo just doesn't want to run with a lower voltage, bios update might help (with all settings at stock i might add). As far as hyper threading goes, your processor will run 5-6 degrees hotter with it enabled. If you're putting extra volts in you probably should disable it.
Yea that works but as i wrote just now anything over ~150bclk murders vcore requirements.
Im still doing tuning, but @1.31v i was getting prime95 errors with a few 4ghz setups that are now completely gone at 1.35, though my temps are high and after 10+ minutes of prime95 its on the edge of 90c with HT disabled which is really damn high, but i figured that was ok just because im not gonna use HT (sc2 and skyrim 2 threaded, xsplit appears to act as if the logical cores were not there) nor approach P95 stress levels (xsplit lags out your entire system if you assign it 2 cores and it EVER peaks to 100% usage, im not sure why, but the entire system stutters really badly, so you have to be careful with increasing encode quality)
Going to be tuning vcore a bit more and trying a few qpi/vtt voltages with 1-2 multipliers down on uncore, that guide says 2:1 with RAM is best stability while i was kinda under the impression from some other stuff that 2:1 was flat out minimum and a stability risk
|
United Kingdom20278 Posts
Confirming i am 100% stable at 150x24 (3.6ghz) with hyperthreading at 1.26v, but 166x20 is hitting errors on prime95 at 1.325v, 166x24 stable @1.35v though on the edge of thermal limits with hyperthreading off
|
This is not completely scientific, but I've read in one overclocking column that uneven multipliers are more stable than even ones, so you want to be @x21 or x23. (not including turboboost mode).
Personally, I also have a 950 D0, and I'm right now testing 3.8GHz (181x21) at stock voltage (1.2V) with HyperThreading and Turboboost off. 25min at prime95 and nothing wrong yet. I'll let it for a couple more hours.
Then I'll shoot for 4GHz. Probably will take some voltage. It everything goes right with 3.8, I'll try 4GHz@1.25V.
EDIT 1:
1 hour test on Prime95 @3.8 on stock vcore was stable.
Got a BSOD on 191x21 with stock vcore by initializing windows. Upped vcore and uncore by 20mV (@1.22V). Testing stability now.
EDIT 2: got a rounding error in prime 95 in the first test. Upping vcore by +20mV (@1.24V) and see what happens.
EDIT 3: BSOD x9C. Upping Uncore.
EDIT4: So far so good. I'm @ 1.25 vcore and 1.18 uncore. I'm running a 15min test just to see if it'll crash and will do an overnight test afterwards to check for stability.
EDIT5: Crashed in 8 min ): I'll try this again later, gotta run. Think Ill up the vcore to 1.28 and the uncore to 1.20.
EDIT6: Got a stable 1 hour test on prime95 @ 1.28 vcore. Running 4GHz (191x21). I'll see if its really stable tonight when Im not using the computer. :D
4GHz yaaay
|
I decided to try for a bit of a higher OC on my 930, right now I'm half an hour into prime blend 183x21 @ 1.2375vcore
HT on, 77°C.
I know I can hit 190 BCLK stable with a lower CPU multiplier, so I'm only tweaking vcore at the moment.
Edit: currently testing 188x21 @ 1.275v it's probably bench stable, but probably not 24h stable.
Edit2: backed off to 185x21 @ 1.25v because of heat, currently 79 C a few minutes into prime blend.
Edit3: bench stable 3.9Ghz: 186x21 1.26v HT on, will run a longer test tonight. if it's not solidly stable I'll see how high I can get with HT off.
that OC was NOT completely stable, currently testing 4.2 GHz HT off 1.31vcore 75°C in prime blend
|
|
|
|
|
|
EPT![[image loading]](http://i15.photobucket.com/albums/a376/Modem69/tempsyo.png)
