Those are CM Hyper N520s with dual 68ish CFM fans on each. Temps aren't a problem.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
SR-2 Optimization Thread
- Thread starter musky
- Start date
Those are CM Hyper N520s with dual 68ish CFM fans on each. Temps aren't a problem.
SR-2 arrived this week, been tinkering and testing non-stop. Wanted to say to all of you who have contributed their time and effort into this thread:
Thank You!!
It's been a *huge* help. (Been reading this thread for 10+hrs the past few days). Any-who, expect to see me poking around in this thread, as I'm also trying to stabilize/push my 2x E5620s to the ~4Ghz mark for folding.
Fold [H]ard, or go home...
Currently running "stable" (aka, still running IBT just to make sure):
210 bclk
19x uncore
1.3125 cpu core
1.4 cpu vtt (hope to get this lower...)
1.68 cpu dram (ditto)
ioh = 1.5
ioh pll = 1.9 (unsure what this one should be)
Thank You!!
It's been a *huge* help. (Been reading this thread for 10+hrs the past few days). Any-who, expect to see me poking around in this thread, as I'm also trying to stabilize/push my 2x E5620s to the ~4Ghz mark for folding.
Fold [H]ard, or go home...
Currently running "stable" (aka, still running IBT just to make sure):
210 bclk
19x uncore
1.3125 cpu core
1.4 cpu vtt (hope to get this lower...)
1.68 cpu dram (ditto)
ioh = 1.5
ioh pll = 1.9 (unsure what this one should be)
No that wasn't itbut thanks, I saw that as well
I never said my posts make sense heh.
Me too. 3.96 (180x22) is the highest I can get without going into the 1.4x vcore range, and 3.9Ghz (195x20) requires 1.425 VTT for each chip. So far neither is showing much advantage in folding performance, as each is within 2 seconds of each other on a P2685.
I'm sure with the uncore running slower I can either lower VTT for higher BCLKS, or use these BCLKs without the magic disappearing DIMM.
Have you tried a BCLK of 187 with Turbo? I have ans use a X5650 and I find that 187 BCLK is a very nice sweet spot, and allows me to get 4.1GHz with VTT and vcore at or below 1.4v, only issue was I had to push IOH to 1.425v to get it stable (My IOH likes the voltage, linpack woul fail within minutes at 1.4v IOH, but be 24/7 stable at 1.425v everything else the same.
In any case, try out the hacked A49 BIOS and play with the uncore, I personally was able to bring my volts down at the same clocks (187x22 Turbo) from 1.3825vcore and 1.4vtt to 1.375vcore and 1.35vtt(IOH 1.425, tried lowering it and linpack started erroring again, oh well)
ebolamonkey3
[H]ard|Gawd
- Joined
- Sep 2, 2008
- Messages
- 1,237
Yea, I've seen some 700D and 800D mods, but after everything's said and done, I'd imagine you're pretty close to $400 anyways, and that's a brand new Lian Li V2120...
Yea, really looking at the V2120, but it's $400 bucks... It's like, a case, or WC for both CPUs and the mobo chipset lol.
Compaq 9000 series, 42U...
http://www.avid-edge.com/pc/SR2_3.JPG
http://www.avid-edge.com/pc/SR2_4.JPG
That is EPIC LOL! Sometimes I'm tempted just to build a bench table out of wood from Home Depot, but I can't stand the noise and wires
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
CHILLYYYYY! Long lost x5650 brother!!
(sorry, bit excited to have another x5650 buddy to share with...)I agree about 187 being a sweet spot - I have been working that the last few days trying to get a summer-proof overclock, and this is rock solid and looking good. (14 hours linpack + 24 hours of 6701s at 26c ambient is proof!)
What is your QPI?
I am at 187 x 22, 1.3625 vcore, 1.35vtt 1.4 IOH, 19x uncore on the unlocked bios, and QPI -85 and -16. So pretty damn close to what you have in other words.
Interesting on the IOH - I have never tried above 1.4, I might give it a whirl.
This is encouraging. That is the highest bclk I have seen for Westmere chips.
Compaq 9000 series, 42U...
http://www.avid-edge.com/pc/SR2_3.JPG
http://www.avid-edge.com/pc/SR2_4.JPG
The big fan made me drool. Is that HP or Compaq branded?
Sounds like a good idea. I will try that BIOS out this weekend and get some benchies going. I had 4.0Ghz at 182 BCLK and my folding performance went down a tiny bit. I then found the culprit. "Lost" a DIMM and was only seeing 10GB in Windows which fubar'd my performance with NUMA off.
That uncore hack looks promising, and I'll let you know how it turns out. I definitely need it methinks, because even at 182 BCLK my CPUs need 1.375VTT and 1.40VTT (primary and secondary respectively) with 1.38125 VCore, so the first CPUs temps get a bit high (low 80s on a couple cores folding).
BTW what RAM timings are you using? My x5650s seem ultra-sensitive to this.
Thanks for the input
Based on the results I've seen, it appears that IOH voltage may be the magic number when overclocking Westmeres. I'm hoping to get my X5650s running today, so I'll experiment further. What is the maximum safe IOH?
Sadly I had to back down on that a bit. Still expirmenting, will post results within the next 24h or so.
1.40 is listed in Shamino's overclocking template, but others have gone to 1.425 without negative results. It might be a temperature related thing more than max safe voltage, but I'm just hypothesizing.
I've pushed IOH no further than 1.5 w/o any temp issues, forced air cooling. Was still trying to get'er stable at the time, so it was only at that level for ~10 minutes or so.
Speaking of safe, anyone know the max safe IOH PLL? I haven't seen anything showing that it helps/hurts with the OC, so I've given up and left it on auto. It usually sets itself around ~1.8
Speaking of safe, anyone know the max safe IOH PLL? I haven't seen anything showing that it helps/hurts with the OC, so I've given up and left it on auto. It usually sets itself around ~1.8
I lost my OC settings when my BIOS failed to load (following a Windows Update that caused a BSOD, for whatever reason), but getting them back was as simple as the first time I set it up.
After reading over this thread, I got my SR-2 and immediately tried 200 BCLK @ 1.35v and it worked, much to my surprise. I needed to add VTT (I'm at 1.4v, although I don't recall if it will remain stable below that) and I entered the XMP timings for my G.Skill RAM manually (9-9-9-24-10-88-6-5-6-1T), set uncore to 2133MHz, changed IOH to 1.4v, disabled NUMA...and voila: LinX and -bigadv stable at 3.6GHz *knocks on wood*.
It sounds like I need to tweak something with my RAM in order to get my TPFs on par with the rest of you. I never really understood the MCH Strap setting so I just left it at auto...anything you guys can think of?
Most of my research on the matter leads me to the same conculusion, the IOH seems to be more temperature sensitive then voltage sensitive(in other words, you can pump the volts to it all day, as long as you keep the temp down).
I suspect if I can find better cooling for the IOH(perhaps adding it to my watercooling loop) I can push the BCLK up up higher, and 4.4GHz(200BCLK, @ 22x turbo multi) should be within my reach. I was able to get 4.4GHz bench stable by pushing 1.5v into my IOH, 1.435vcore and 1.375vtt into both cpus. It only started becoming unstable when the ambient temps rose(due to the heat the system was putting off)
I don't have any money to cool the chipset better at the momment to test my theroy. Even if I did, getting a proper memory configiration is slightly higher on the priority list if I did come accross any spare cash.
may i be worthy: I have my QPI set at 4.8(18x) and other QPI settings at -76 and -16. So in summary;
187 x 22, 1.375 vcore, 1.35vtt 1.425 IOH, 14x uncore on the unlocked bios, and QPI -76 and -16
10e: As far as my ram timings, I have a VERY mismatched set, but in any case am running them at 1496MHz(748MHz) @ 9-9-9-24-10-88 at T1 Command. at 187 BCLK that means I am using the 1066MHz divider(2:8) and the 1333MHz Strap with voltages set at 1.62v in BIOS giving me 1.65v Actual.
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
IOH - I agree with 10e (which rhymes) I am beginning to think if temps are ok, then go for it. (I am far happier to risk a $600 motherboard than a nearly $3000 of cpus!)
QPI -78 - is that the max you can do? I ask because Shamino said you want the highest (well lowest negative number) you can reliably boot with. Hence I have -85 on both mine. But then again others have done fine on auto. But you never know what might be a bottleneck...
QPI -78 - is that the max you can do? I ask because Shamino said you want the highest (well lowest negative number) you can reliably boot with. Hence I have -85 on both mine. But then again others have done fine on auto. But you never know what might be a bottleneck...
Here's a weird question. How many times does your SR-2 beep when it boots? My old one beeps three times, and my new one beeps five times (and seems to POST a little slower).
I'm wondering what the beeps mean. My only guess is they may have something to do with attached USB devices, since it beeped when I disconnected and reconnected the USB keyboard/mouse while in the BIOS screen.
I'm wondering what the beeps mean. My only guess is they may have something to do with attached USB devices, since it beeped when I disconnected and reconnected the USB keyboard/mouse while in the BIOS screen.
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
Yes = I get extra beep when I have a usb stick in to boot to memtest!
- Joined
- Oct 23, 2007
- Messages
- 6,105
Ok, someone help me before I pull my hair out.
I'm running new ram, not oc'd, and I still cannot get NUMA disabled and boot.
Running 12gb of Corsair XMS3 2000 9-9-9-24 1.65vdimm, settings dialed in, they run like a top, no problems.
I go and disable NUMA, same thing happens that happened with my 1333@1600....starts to boot, gets to 2A on the LED, clicks, starts boot over again. Does this 3x, powers off. Powers on, rinse and repeat. Eventually i have to force power off and clear CMOS and start over again (or rather choose to load my settings and pick up where I was, with NUMA enabled).
Any ideas? anyone??
I'm running new ram, not oc'd, and I still cannot get NUMA disabled and boot.
Running 12gb of Corsair XMS3 2000 9-9-9-24 1.65vdimm, settings dialed in, they run like a top, no problems.
I go and disable NUMA, same thing happens that happened with my 1333@1600....starts to boot, gets to 2A on the LED, clicks, starts boot over again. Does this 3x, powers off. Powers on, rinse and repeat. Eventually i have to force power off and clear CMOS and start over again (or rather choose to load my settings and pick up where I was, with NUMA enabled).
Any ideas? anyone??
Cool. I used a USB DVD drive to install Windows (which the BIOS detected as a normal DVD drive, which was pretty neat) and I'm also hooked to my KVM which has a USB hub. I'll try removing the DVD drive when I reboot and see if I lose a beep.
Ok, someone help me before I pull my hair out.
I'm running new ram, not oc'd, and I still cannot get NUMA disabled and boot.
I can't help other than to say I saw something very similar on my new SR-2 when I tried a gentle overclock (160 BCLK, x10 RAM) so that makes me think NUMA may have been my problem. The new box has 6 x 4GB of G.SKILL 9-9-9-24, so I'm expecting it to be a little tougher to get maximum memory speed.
BTW, running 20GB LinX passes takes almost 15 minutes per pass at stock clocks.
I think I'll make another attempt at overclocking now....OT, but I can push mine to the -85, but I've got different procs than you two guys.
Mine beeps 2 (or 3?) times, with a USB keyboard/mouse plugged in via a KVM, and a single video card. BUT! I notice it'll beep when I swap to a different KVM source or if I unplug the video cable.
Running 100% stock? Latest BIOS? Clear your CMOS and re-arrange your RAM. I know this sounds stupid, but I swapped some sticks around (slot 1 and 3 swapped, for example) and could get it to POST properly. It was almost as if it didn't like a specific stick in a specific slot.Ok, someone help me before I pull my hair out.
I'm running new ram, not oc'd, and I still cannot get NUMA disabled and boot.
Running 12gb of Corsair XMS3 2000 9-9-9-24 1.65vdimm, settings dialed in, they run like a top, no problems.
- Joined
- Oct 23, 2007
- Messages
- 6,105
No, I mean the ram is running @ stock 2000 using 2:10 divider. cpu is oc'd to 3.6 using 200bclk.
A47 bios, not latest.
Had the same problem with my 1333@1600 using 2:8.
A47 bios, not latest.
Had the same problem with my 1333@1600 using 2:8.
I can go up to -92 with no issue, fully stable and reliably boot, however it doesn't do anything for me stablity or performance wise at 187BCLK so I just keep it at -76 for no partcular reason. I do put it at -92 though whenever I I play with things and try to squeeze more out and etc.
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
@zero2dash - can I ask why the old bios? It might be worth trying the A49 and or A49 hacked - remember you can install 3 at a time using the switch to 3 banks. I have my best overclocks yet on the hacked A49 on my x5660 rig...
Otherwise post more settings like MCH and uncore etc
@chilly - righto - I don't know how much qpi effects things, but if you are held back it is worth a go.
Otherwise post more settings like MCH and uncore etc
@chilly - righto - I don't know how much qpi effects things, but if you are held back it is worth a go.
Just removed the USB DVD drive, and now I get four beeps instead of five, so apparently the beeps are indeed some sort of USB device detector.
Flashed to A49 with unlocked uncore, so I'll see what OC I can get quickly before I fall asleep...
Flashed to A49 with unlocked uncore, so I'll see what OC I can get quickly before I fall asleep...
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
Changing uncore on the official bios does not seem to actually do anything - check NB freq in CPU-Z and it is the same.
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
More than than the usual 10c higher than CPU1?
OH! Well F me, no wonder I was having such [H]ell with that. Ugh. Where would one obtain such a BIOS?
My CPU0 is a bit hotter, too, but I thought that was "normal" with the SR-2?
Last edited:
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
MIBW, you are the man. Thank you!
Edit: And I'm dumb...this seems to be for the X56xx procs. Probably won't work on my E5620s right?
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
MIBW, you are the man. Thank you!
Edit: And I'm dumb...this seems to be for the X56xx procs. Probably won't work on my E5620s right?
Hmmm, got me there. I honestly don't know.
But as it is a bios, you could just flash one of the 3 banks and see what happens. I don't think it is possible to permanently damage a CPU that way, but who knows.
I was happy to be the [H] canary and try it first because I have two SR2s, which meant no matter what I would still have one to cuddle while I cried over the smoking ashes of the second.
Good point -- I'll probably give that a try tomorrow (lol, later today) and see if it helps. I am still learning with this SR-2 beast, so having another option probably wouldn't hurt.
Hmmm...I might volunteer to be a [H] canary as well, since my E5620s are *much* cheaper than what you [H]ex boys are running. Only downside is, unlike you, I won't have a backup to cuddle with.
Hmmm...I might volunteer to be a [H] canary as well, since my E5620s are *much* cheaper than what you [H]ex boys are running. Only downside is, unlike you, I won't have a backup to cuddle with.
Hmmm...I might volunteer to be a [H] canary as well, since my E5620s are *much* cheaper than what you [H]ex boys are running. Only downside is, unlike you, I won't have a backup to cuddle with.
Actually, they aren't cheaper than the L5640 "extra spicey" chips most of the dual hex SR-2s around here run, which is the main reason why there are so many dual hex (and single hex for that matter) machines. Unfortunately (for you and everyone else who did not jump on these), the supply dried up.
I have one E5620 in-house (well, in Vaulter's house) and was about ready to pull the trigger on a second one when the hexes became available. Seemed like a nice chip, but I only ran it long enough to flash a bios. Hopefully you get some good numbers out of yours.
Ahhhh, OK -- that explains things! At that price, I'm jealous and impressed all at the same time.
So, I type this from the modified A49 BIOS system, and starting from the ground up on the OC, just to be sure. AFAIK, it looks like the unlocked uncore worked.
Currently testing a 181bclk x 19 multi, with the uncore @ 14x. CPUz shows NB freq at 2535, which seems correct. The QPI link shows 3259, and I've got it set to 4.8(or whatever) in the BIOS.
Will keep pushin' and see where we land.
Edit: Did a quick mm'bay search, and there was one, but it was for roughly double the price of what the E5620s go for, but still ~$400 less than retail. Might have to keep an eye out hehehe
So, I type this from the modified A49 BIOS system, and starting from the ground up on the OC, just to be sure. AFAIK, it looks like the unlocked uncore worked.
Currently testing a 181bclk x 19 multi, with the uncore @ 14x. CPUz shows NB freq at 2535, which seems correct. The QPI link shows 3259, and I've got it set to 4.8(or whatever) in the BIOS.
Will keep pushin' and see where we land.
Edit: Did a quick mm'bay search, and there was one, but it was for roughly double the price of what the E5620s go for, but still ~$400 less than retail. Might have to keep an eye out hehehe
As an eBay Associate, HardForum may earn from qualifying purchases.
- Joined
- Oct 23, 2007
- Messages
- 6,105
I lowered mine to 16x, my NB freq. is 3200. looks like it worked fine to me I'll try A49 later. Need to sanitary erase my Vertex and do some fw updating before reinstalling 7, *then* I'll go ahead and update to A49 and see if I get anywhere on the NUMA front.
Half the cores on CPU0 are 10C higher than the other half under load. I thought something was weird when I mounted the heatsink, so I'll take it off and readjust.
edit: I think I identified the problem. CPU0's HSF was touching the mobo heatsink on the edge of the board. I remounted the HSF leaving about a hair's width between the two, so it should be firmly seated now. Overclock attempt #3...
edit 2: OK, weird discovery of the day. Both Core Temp an E-Leet have the temps for CPU0 and CPU1 reversed. I thought my temp problem was on CPU0, but after remounting the heatsink nothing changed. I noticed the CPU1 heatsink felt hotter, and I confirmed this by disabling CPU0 and then CPU1 using the mobo jumpers and running LinX. CPU1 is running hotter than CPU0 and has half the cores running extra hot. My temps are 60C on CPU0, 65-72 on CPU1. It makes sense that CPU1 would be a little hotter because the CPU0 fan is blowing hot air on it. BTW, everything else in E-Leet looks correct, just the individual core temps seem wrong. Could this be a side-effect of the hacked A49 BIOS?
Last edited:
may i be worthy
[H]ard|DCer of the Month - December 2010
- Joined
- Aug 17, 2010
- Messages
- 1,186
Ok, you just did my head in, so I checked - CPU0 and CPU1 as labelled in the SR2 manual are reversed to how all temperature monitoring programs are seeing them! CPU-0 (reported) is the hotter one on both SR2s, but this hotter socket is the one labelled CPU1 (closest to the PCI ports) I turned off/on a fan to confirm which one is which in the monitoring.
grrr, this makes it so confusing as to which one to refer to - the documented in the manual CPU0 or the monitored CPU0...
Thanks for making me check something I should have worked out months ago!
Yep, I just figured this out last night.
I took off the heat sink fan on the back (secondary) CPU and watched CPU0's temps go nuts on both the LED board on the mobo, and HWMonitor. I put it back on before reaching heinously high temperatures, but what the hell?
The uncore lowering trick helped get a few degrees down, as right now I'm at 14x uncore, but temps on "CPU 0" were still flirting with close to 80 celsius. I've since added a second fan on the back of the heatsink which has helped get cooler, but this is just plain silly.
I took off the heat sink fan on the back (secondary) CPU and watched CPU0's temps go nuts on both the LED board on the mobo, and HWMonitor. I put it back on before reaching heinously high temperatures, but what the hell?
The uncore lowering trick helped get a few degrees down, as right now I'm at 14x uncore, but temps on "CPU 0" were still flirting with close to 80 celsius. I've since added a second fan on the back of the heatsink which has helped get cooler, but this is just plain silly.