Ah, okay. I see. The 6.4GT setting.
Printable View
Ah, okay. I see. The 6.4GT setting.
been trying for 4k uncore on 2k c8 mhz dom GTs @ i7 920 on this board. Been pulling my hair for a week now and it won't stay stable on prime blend custom 5400 gb in diagnostics mode. Any guidelines on how to do this?
Put everything to full auto (load bios defaults and reboot).
Then manually set the CPU multi to 13x, and in ram setup pick the XMP profile.
Pick a multi that gives you 2000 or close to that (1:14 is closest i think) but dont go over it.
Reboot, go in BIOS, see the values it has picked for you. You'll certainly have to correct VTT (which XMP seems to overshoot). Vdimm should be 1.65V for this kit.
Boot in OS and Memtest, run stress tests, lower VTT, repeat until crash.
Now you have your minimum memory setting but remember if you want to OC your 920 you'll have to lower the RAM multi since you'll be raising the BCLK, or you'll get over 2000MHz which results in a whole new world of problems.
Is your 920 able to run 4GHz Uncore without RAM being at 2000?
Maybe you could do yourself the favour to fill out (copy it here and fill out) the BIOS template for the classified:
http://www.evga.com/forums/tm.asp?m=...&key=%EF%BF%BD
:D
It's not easy to find the right balance between vtt and vdimm and the fitting RTL configuration. Subtimings matter as well. So please be detailed and it would definitely help to get you a signature with your system specs so that we can see what you got at first glance. ;)
Seems to be dead link now, maybe you mean this one ?
Here are the screenshots. I'll edit this post with Bios template in a bit. Edit: Checked the bios, for 4 ghz screenie and everything was on Auto on Voltage menu except Vcore at 1.6250 and Vdimm at 1.550v. Cpu features all off except all cores. I'm setting minimal voltages and rerunning linx 25 runs. my bios settings and e-leet read seems to have a very huge difference. =.="
Signature updated too ^^ Thanks a lot for helping me. I didn't expect to get so many responses so quickly.
Yep, link was from OP. I dd not check whether it works or not. :D
No problem, mate. Here for help <- me.
BTW, I cannot see anything on those tiny screens. :p:
/edit. Better now^^ :up:
//edit.
k. I would say your settings are fine from what i could see in the screens so far (ain't much ;) ). I'd recommend to let realtemp run while stresstesting. I'd further recommend to open up an eleet window showing set voltages for such screens. It's much easier for us to see if something's wrong.
@4.5GHz you cannot use the 10x ram multiplier without ocing your DomGT far away from specs. I guess it can be done with the right sticks and settings, but it depends on temps and the willingness to drive voltages to their limits. :D
///edit. :D
The signature is visible in your profile, but not below your posts. Pls check if you enabled displaying the signature in user control panel.
I'm actually aiming for 4 ghz (20x201)CPU/4k uncore/Ram at 2k mhz 8-8-8-24 1T then only go for 4.2 ghz (21x201)/4k uncore Ram at 2k mhz 7-8-7-24 1T... 4.5 ghz HT on was already on 85C max temp on linx so i'm just gonna leave it as it is. Screenies were just to show it runs fine at 4 ghz 3200 uncore without touching ram multi as per your question earlier. I'm rerunning linx with voltages set instead of Auto and will edit screenshot and signature as per your advice too. Sorry i'm new here and wasn't sure how to add it to the signature.
Hmm, I think you misunderstood my request. You should try to run 21x200 8x Ram and 20x Uncore first to make sure your uncore can take it. Next step would be to up the ram multi to 10x if stable.
omg i got it all wrong. so I can use the settings I posted but raise uncore to 20x and raise VTT notch by notch only until its stable? Also which stability test is better for uncore? Prime95 blend custom 5400 or linx all ram?
For Uncore/RAM testing prime x64 w/5.5GB mem usage is best IMO. Be careful with Vtt and don't go over +300mV. Under water those IMCs get damn hot, especially if you increase vtt + vdimm, that you could burn it priming for too long. The highest I used was +275mV resulting in 1.459V loaded Vtt under watercooling.
can anyone confirm/deny that westmere works on 759/760/762 without hardware mod?
i want to run my x5677 in a 762 if possible.
I can only tell what I read... for Rev.1.0 boards there have to be soldered two smds - you can find a descriptionon kingpin cooling forums.
Rev.1.1 is said to work with Westmere without these mods. Ask shamino (Lardarse) for confirmation.
thanks for the warnin SAE
how long should I run prime for with the above settings? I ran linx and it passed 25 runs @ CPU VTT Voltage 1.2v at 1.37500v in e-leet voltages tab before u replied now I'm running prime and so far its stable. What do I do next once it passes prime?
So you made sure 4GHz Uncore are stable, you can now tell what vtt is needed the least.
I would say you set the memory to xmp in bios and watch the timings being set in memory menu. DO NOT let it use xmp though as it sets Vtt way too high. You may be able to just change voltages using xmp, but i am not sure. Use xmp timings, +175 Vtt, 1.63-1.64V Vdimm - depends on what you get from overvolting vdimm circuits (eleet is a good indicator for vdimm and relativley spot-on). You should not go over 1.65V real vdimm first, as specs and binning from corsair are being "conservative" imho. There's often much more in those GT sticks. Don't forget to set the ram multiplier to 10x manually. xmp may set it to 14x.
BTW, you may need one or two notches more vcore with this kind of imc stress.
Yup earlier prime95 crashed BSOD 0x0000003B. I had to increase vcore 1 notch. Running in for 2 hours now. I'm just gonna let it run for another 15 mins before running xmp with edited voltages and ram multi as how u advised. Will edit this post with results.. if it crashes, does it mean I can rule out lack of vtt as the reason?
edit: xmp sets trfc to 110. shouldn't it be 88 for c8?
edit: yeah I'm testing with xmp first. looks good so far. I set vdimm to 1.63. e-leet reading is 1.69. Is there anyway to open 2x e-leet windows?
Probably the reason I'm pulling my hair for a week is because if trfc settings being set to 88 and the rig refuses to stay stable for 2 hour prime 5.4gb runs.
edit: Prime95 froze @ 7th test. increasing vcore 1 notch and rerunning it. increasing vcore 1 notch caused it to crash at 2nd test. backed down vcore 1 notch now and increased vdimm to 1.64v.. prime95 restarted.. its looking like its the same as when I'm running trfc 88..sigh. Now trying vcore up 2 notchs and vdimm at 1.64v. prime froze again after 3 tests.
110 is more on the safe side, but a lil slower. Pls test with xmp timings first to make sure, everything is working as intended. You can tweak based on these settings later.
Vtt errors have often been failing cores in prime x64 for me. Sometimes you get bsods with that ...0x00000124 line.
so each time 1 worker stopped means increase vtt?
May help. If not, I would try vdimm or vcore.
But there are various settings going hand-in-hand. Important for me were:
- IOH / QPI PLL 1.15V both
- CPU and IOH PLL 1.5 or even 1.35V were most stable
- ICH 1.075V was best
- keeping vdimm 0.235 - 0.275V higher than vtt has worked best for me
- Vcore had to be increased by 3 notches for tight timings at 2000 and needed high uncore vs. 8x ram and 16x uncore at same clocks
thanks a lot SAE. I've never seen my rig this stable with that cpu clock together with that uncore and ram freq. played with these voltages the whole night and was still unstable. Got pissed and turned on turbo. turned up some vcore and tweaked DIMM DQ vRef to -50 and used 1.61 vdimm which was the one that gave me the most runs b4 it crashes. I was just trying my luck and don't even really know when to adjust DIMM DQ vRef lol ...and got this.. this is just wierd =.=". Gonna let it run for one whole night and see what happens. if this is stable overnight gonna try for 7-8-7-24. Is that possible with this sticks with just air cooling on the ram?
7-8-7-xx should be possible, of course. Depends on your sticks though. Aircooling is enough, my 1057 7-8-7-21 1N were on air as well (stock Corsair dual fan). You might have to up vdimm a bit to run 7-8-7-21. Try it out if stable. Big thumbs up for ya. :up:
Are the voltage monitor readings on ELEET overstated? Like the VTT? I'm on Bios 59.
It seems that when I use +100mV it says CPU VTT is 1.4v but in the voltage section it says 1.3v? I think the latter is correct. Just wanted to check
Eleet overshoots some readings such as VTT and Vdimm yes. BIOS readings are more accurate.
SAE, we can definitely set voltages manually with XMP profile. That's what i am using (with great results on benches so far) so XMP actually sets the timings, and i set the voltages myself.
So far my memory benches are more than excellent so i will keep these seettings.
Thanks. Is it just the vDIMM and VTT volts? I'm too lazy to buy a multimeter atm...