Core i7/X58 Overclocking Thread

[Pisklink]

If at any point you cannot boot, increase your QPI Voltage (VTT Voltage) by 0.025V and try again. Keep increasing it by that amount until you can boot at the given BCLK. Do not exceed 1.475V on QPI. For example, if you cannot boot at 190MHz with 1.35V QPI and you increase the voltage twice (2 x 0.025V) to 1.4V QPI and then you boot fine, you need a minimum 1.4V QPI for 190MHz BCLK. If you want to be sure that the given BCLK is stable then take the minimum QPI voltage required to boot and add about 0.025V to 0.05V to it.

Isn't it true that if you e.g. find a max. bclk of 210 with all multi's lower than rated and a QPI/DRAM voltage of 1.35v, this doesn't imply stability when raising the multi's? To put it in other words, wouldn't QPI/DRAM be stressed more after raising the multi's again and subsequently needs additional voltage? Or is that the reason you suggest to add another extra 0.025 or 0.05v?

[hlonipha]

what bios are you using? i cant get 20x200 stable at all, im using F4r bios for the UD5..

So far for me is F5G

[dejanh]

Sorry, been busy with my Honours thesis so I haven't been able to do this yet.

However, I've tried Vcore at 1.400, QPI/VTT at 1.360, VDIMM at 1.660 and CPU PLL at 1.88 before and couldn't get prime stable with 18, 19, 20 times multis with Bclk over 190. I could boot into windows in most instances. I've set Memory and Uncore multis to 6x/(12x or 13x) and 8x/(16x or 17x) respectively. According to GB's Easy Tune 6, at load Vcore dropped to 1.360 and VDimm to 1.616. I am using Prime 95, CPU-Z and RealTemp to stress and monitor.

I know that I should have tried finding my max Bclk, but I was still hoping at the time that I could just get to 4.0GHz with ease (apparently not).

Now I'll have to wait until I get some University stuff sorted before I can try again. Thanks for the suggestions though.

Yeah, I would suggest that you follow a process come time to try OCing again. It will take a lot of guessing out of the equation and we all know guessing in equations is not good

Isn't it true that if you e.g. find a max. bclk of 210 with all multi's lower than rated and a QPI/DRAM voltage of 1.35v, this doesn't imply stability when raising the multi's? To put it in other words, wouldn't QPI/DRAM be stressed more after raising the multi's again and subsequently needs additional voltage? Or is that the reason you suggest to add another extra 0.025 or 0.05v?

That's why I suggest to add extra voltage Everything requires a bit of tweaking, but you have to start with a solid foundation first. Maybe you will even find that you need to add 0.075V, though unlikely unless you are using one of those even multipliers like 20x or 22x. Most chips do not favor those.

[ChaosAD]

I m running 19x200 atm. I m limited by stock intel cooling. I use (all values are written from inside vistax64) vcore:1.23v, vdimm:1.60v, vtt:1.28v. I run prime small ffts 2hours stable, prime blend 2 hours stable, memtest 2.11+ 2 full passes no errors. But as soon as i run LinX 64bit with full memory usage after exactly 25sec system freeze. I dont have any temp issues, using prime i max at 77-78C. Any suggestions on this?

[dejanh]

I m running 19x200 atm. I m limited by stock intel cooling. I use (all values are written from inside vistax64) vcore:1.23v, vdimm:1.60v, vtt:1.28v. I run prime small ffts 2hours stable, prime blend 2 hours stable, memtest 2.11+ 2 full passes no errors. But as soon as i run LinX 64bit with full memory usage after exactly 25sec system freeze. I dont have any temp issues, using prime i max at 77-78C. Any suggestions on this?

Your QPI/DRAM (VTT) voltage is a bit low. I would increase it to 1.35V and then try to run LinX.

[ChaosAD]

If i remember correctly i did try up to 1.34v with same results. If vtt was low, shouldnt i have problems with prime blend or memetest?

[Godmyster]

So far for me is F5G

have you tried the F6 bios? wanted to know how they are so far... i noticed that the F6d came out recently... any feedback?

[dejanh]

If i remember correctly i did try up to 1.34v with same results. If vtt was low, shouldnt i have problems with prime blend or memetest?

Well, I'd look towards your temps then. LinX pushes the CPU more than Prime Blend so I think you just do not have adequate cooling for your setup.

[PC_User]

what bios are you using? i cant get 20x200 stable at all, im using F4r bios for the UD5..

I'm using F5. Everything is fine, but I can't get it to sleep properly...

Overclocking is not just about plugging in random numbers and hoping for the best. There is a process for it that should be followed if you want to have a decent chance finding good overclocks without excessive trial and error.

I agree. I made a spreadsheet with a bunch of combinations. Picked initial clock and voltage values and started incrementally increasing/decreasing some. Took me about 30 trials before I came up with 3 stable settings.

[ChaosAD]

Well, I'd look towards your temps then. LinX pushes the CPU more than Prime Blend so I think you just do not have adequate cooling for your setup.

Thats why i am curious what the problem is. On the first 25secs, that pc freeze, temps are at the same lvl as prime. No signs of overheat till then

[Pisklink]

Try to get some extra fresh air over you CPU (e.g. air from outside using a blower or an airconditioner). Use that as a quick method to find out if it will hold longer in LinX. Probably it will still make it freeze but maybe it will take some longer. If it repeatedly freezes after 30+ seconds or so, you could conclude it is a temp. problem.

I know it is not exactly an empirical way to find out but it's cheap, easy and a quick way to provide you with some information. You could always buy an aftermarket cooler for around 30 euros.

[kpablo]

Hey all...

Built another system based on i7... this time i got a 920 batch #3841A346

Any comment?

[Hooti3]

3841 is a pretty good batch. Should get to 4.0Ghz pretty easy.

I on the other hand, just ordered a D0 i7 920 (Xeon W3520). Should be here in a day or two

Wish me luck

[Efekkt]

Hey guys, I've got a d0 stepping 920 here with the batch number 3848A648. You guys know if it's any good? Oh and I'm probably gonna OC my processor later tonight. Any suggestions on what I should start with? Maybe a guideline for me to start with? I'm on a P6T deluxe and 6GB corsair dominator 1600mhz. Currently running on stock settings.

[Nizzen]

Is there a bios that lock 21x MP on EVGA x58? I use Xeon w3520 d0.

The w3520 is overclocking like a god

[FabricioGS]

People, need some help.

I used to have some generic 3x1GB 1333 DDR3 modules and my system was rock solid (10x Linpack stable) at 3.9GHz (205x19). Now I have them replaced for some 3x2GB 1333 modules from the same brand and they seems to overclock as well as the 1GB modules (verified with memtest) but I can't run the system stable at that clock anymore.

What voltages should I tweak? vQPI? IOH?? I tried a little more vcore too but still not stable.

[rge]

When I went from 3x1gb OCZ on windows XP, to 3x2gb OCZ on windows 7, I had to increase DRAM volts from 1.6 to 1.64 and increase qpi v from 1.31 to 1.35 to achieve stability.

I also had to bump up my vcore by 2 notches, but later confirmed that is because I used linx 64bit with 4-5gb of memory for testing. linx 64 bit using 4-5GB mem required 2 notches higher vcore to run 10 passes stable, vs prime for 14hrs or linx 32 bit (linx 32 bit can use only ~2gb max)

[Efekkt]

Alright guys, my first time overclocking and I managed to get a core speed of 4.2ghz with 200 bclock and x21 multiplier, so I'm pretty proud of myself. :P Question about the bclock though. Unlike the FSB (higher, the better; even with a lower multiplier), the new bclock doesn't work the same way, right? Or does it?

Any good software besides Prime95 to test stability too?

[FabricioGS]

When I went from 3x1gb OCZ on windows XP, to 3x2gb OCZ on windows 7, I had to increase DRAM volts from 1.6 to 1.64 and increase qpi v from 1.31 to 1.35 to achieve stability.

I also had to bump up my vcore by 2 notches, but later confirmed that is because I used linx 64bit with 4-5gb of memory for testing. linx 64 bit using 4-5GB mem required 2 notches higher vcore to run 10 passes stable, vs prime for 14hrs or linx 32 bit (linx 32 bit can use only ~2gb max)

Thanks, that was exactly what I was looking for.

I had to up my vDIMM too but they are stabe at the memtest now, so I'll try higher vQPI and vcore.

[Sch0cK]

I am on Xeon W3520 , Apogee GT 1600, BloodRage
At the moment i do 195,4 x21 for 6 Hours ,
Vcore @ 1,32V, Ram @ 1,67V , Vtt @ 1,28 tems @ 62°-68°C

Max tried without any problems 211Blck

http://valid.canardpc.com/show_oc.php?id=558035

[Efekkt]

Alright, I just tested my CPU for about 6-7 hours on Prime95. What do you guys think of the results? I personally think that the temperature is really high and I really wanna lower voltage, but my computer doesn't seem to boot with anything lower than that voltage. Any tips on what I should do? Also, my bclock doesn't seem to be able to get pass 215. What's up, eh? =/



I'm currently trying for 204 bclock with the multiplier set at 21x. I did some stress testing on the CPU for about 30 minutes, ran without errors, so I'm gonna leave it on for tonight and see how it goes tomorrow.

[Splave]

Thanks, that was exactly what I was looking for.

I had to up my vDIMM too but they are stabe at the memtest now, so I'll try higher vQPI and vcore.

might need to up your trfc a bit too

[The-Fox]

Alright, I just tested my CPU for about 6-7 hours on Prime95. What do you guys think of the results? I personally think that the temperature is really high and I really wanna lower voltage, but my computer doesn't seem to boot with anything lower than that voltage. Any tips on what I should do? Also, my bclock doesn't seem to be able to get pass 215. What's up, eh? =/

I'm currently trying for 204 bclock with the multiplier set at 21x. I did some stress testing on the CPU for about 30 minutes, ran without errors, so I'm gonna leave it on for tonight and see how it goes tomorrow.

Results seems fine compared to other D0 chips which posted results.

Regarding the bClk, what board are you using ?

Anyway, try to increase IOH/ICH voltage a notch or two.
In addition, try to increase PCI-E frequency, around 110Mhz could do it.

And if this doesn't help, start tweaking that BIOS

I always prefer to work with high Multi (x21) and low bClk, i920 likes 21 multi in most cases.

[unclewebb]

Efekkt: 90C is too hot. You need to find a way to run that chip cooler if you want to push it to 4.2 GHz or beyond.

RealTemp reporting a multi of 20.5 is usually a very good sign that Turbo mode is becoming disabled at full load. This can happen depending on your bios if you are drawing too many amps.

I wrote a separate utility called i7 Turbo that does a good job reading the multiplier. It uses high performance timers within Core CPUs and is the Intel recommended way to accurately read the multiplier. You can download and read about it here:

http://www.xtremesystems.org/forums/...&postcount=918

I'd be interested to see what it says when you are running Prime Small FFTs at your present overclock settings. Here's the latest version of RealTemp 3.20 as well:

http://www.fileden.com/files/2008/3/...alTempBeta.zip

[Efekkt]

Results seems fine compared to other D0 chips which posted results.

Regarding the bClk, what board are you using ?

Anyway, try to increase IOH/ICH voltage a notch or two.
In addition, try to increase PCI-E frequency, around 110Mhz could do it.

And if this doesn't help, start tweaking that BIOS

I always prefer to work with high Multi (x21) and low bClk, i920 likes 21 multi in most cases.

I'm on the P6T Deluxe with the latest bios (1403). I'll try what you said. Let you know my results soon.

I tried 205 bclock with the vcore set at 1.34 yesterday and tested it on prime95, woke up with a BSOD.

Efekkt: 90C is too hot. You need to find a way to run that chip cooler if you want to push it to 4.2 GHz or beyond.

I'd be interested to see what it says when you are running Prime Small FFTs at your present overclock settings.

Yeah, it is. I don't know how to lower the temperature, though. I live in quite a hot country, so ambient temperature is quite high most of the time. Oh and I'm pretty sure I messed up my thermal paste. The one thermalright provided me was pretty hardened up, so I couldn't really spread it around properly. I'll probably get new thermal paste, lap my heatsink and see how the temps are after that.

Edit: I did test the overclock settings running Prime Small FFTs. It ran stable for around 6-7 hours.