Message boards :
Number crunching :
System Tuning
Message board moderation
Author | Message |
---|---|
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I have been tweaking my rig to get the most out of it. Running hyperthreading at 4.26 GHz and higher has been a struggle. I was fine doing it with Einstein, but on switching back to SETI I have had problems, even though Einstein seems to be harder on the CPU. Earlier today I did see a video driver recovery from a remote log in. I tried upping the GPU voltage several times, which really didn't work. Now, after the last BSOD, I upped the CPU core voltage to 1.45 VDC. I may have to do some more tuning, but what struck me is that now I'm pulling 922 watts with just the tower. I haven't seen that much since I last ran Furmark. I hope it holds, as my CPU temp is only 53°C, and as my coolant heats up, I don't expect it to go over 62°C Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
I have been tweaking my rig to get the most out of it. Running hyperthreading at 4.26 GHz and higher has been a struggle. I was fine doing it with Einstein, but on switching back to SETI I have had problems, even though Einstein seems to be harder on the CPU. Earlier today I did see a video driver recovery from a remote log in. I tried upping the GPU voltage several times, which really didn't work. Now, after the last BSOD, I upped the CPU core voltage to 1.45 VDC. I may have to do some more tuning, but what struck me is that now I'm pulling 922 watts with just the tower. I haven't seen that much since I last ran Furmark. I hope it holds, as my CPU temp is only 53°C, and as my coolant heats up, I don't expect it to go over 62°C Steve: I know with my i7 920 at 3.8G I can run it with 1.22v but if I go to 4.021 then I have to use 1.35V. Above 4.021 I have to do a minimum of 1.40 and sometimes 1.50 for 4.4G. It just seems the chips have good spots and once you go past those the voltages go up in a very steep curve. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I've still got more room to work. The problem is that it may not crash for 12 or more hours. At least it's the weekend, and I have more time to tweak. I had it at 4.5 GHz for a couple of hours. My RAM is running at 1705 mHz at the moment, but I have had it stable at 1774 MHz. Ahh the fun of tweaking. I just want it to stabilize before I try advancing it further, but time will tell. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
I've still got more room to work. The problem is that it may not crash for 12 or more hours. At least it's the weekend, and I have more time to tweak. I had it at 4.5 GHz for a couple of hours. My RAM is running at 1705 mHz at the moment, but I have had it stable at 1774 MHz. Ahh the fun of tweaking. I just want it to stabilize before I try advancing it further, but time will tell. You might want to read up on your motherboard also. I have read/seen some motherboards have problems at certain multipliers with a high BCLK (usually above 215). I know both of mine (Rampage III Extreme, Asus P6T Deluxe V2) have issues around 218 BCLK. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I'm running the Rampage 2 Extreme at 170 BCLK, and 25 as a multiplier. If I disabled one thread it would work fine, but with all 12 going along with 3 wu's per GPU, I really want to see what 18 wu's at a time will do for me. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
I'm running the Rampage 2 Extreme at 170 BCLK, and 25 as a multiplier. If I disabled one thread it would work fine, but with all 12 going along with 3 wu's per GPU, I really want to see what 18 wu's at a time will do for me. Then I would look at CPU voltage or memory not liking that speed/voltage combination. How long have you run memtest86 to verify you are not having memory issues at that speed? I know sometimes it takes 7-8 runs through before you start seeing errors. The other issues can be the memory controller on the processor not going past a certain rate but at a 170 BCLK that shouldn't be it. |
-BeNt- Send message Joined: 17 Oct 99 Posts: 1234 Credit: 10,116,112 RAC: 0 |
Have you tried messing with your clock skews and your cpu pll voltages? That's what I did to get my ram working at 1780Mhz and my cpu at 4Ghz with only 1.28 volts. Another trick on my Rampage Formula has also been changing the pci-e bus to 101mhz instead of the 100mhz seemed to help stability here. I know mine is just a Core2Quad but overclocking between machines is pretty comparable as far as what settings add stability. Of course you are playing around with a lot more ram than I have (only 8Gb here) and an i7 with triple channel and two extra video cards, so your machine is a whole different demon from mine. Traveling through space at ~67,000mph! |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I haven't run memtest86 yet, but I'll give it a go sometime this weekend. The RAM is Corsair DDR3 2000, and I have had it running stable at 2000 Mhz, but not at this CPU speed. One issue I am aware of is that I'm running 12 Gig of RAM. From what I've read, 12 Gig is much harder to maintain at that speed than 6 Gig. The problem with that, is that this rig is not just a cruncher. It will do everything. I've actually built my dream machine. It just blows through anything I throw at it, and that is satisfying. I am sure I will find the happy spot, then kick it up some more. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
Have you tried messing with your clock skews and your cpu pll voltages? That's what I did to get my ram working at 1780Mhz and my cpu at 4Ghz with only 1.28 volts. Another trick on my Rampage Formula has also been changing the pci-e bus to 101mhz instead of the 100mhz seemed to help stability here. I know mine is just a Core2Quad but overclocking between machines is pretty comparable as far as what settings add stability. Of course you are playing around with a lot more ram than I have (only 8Gb here) and an i7 with triple channel and two extra video cards, so your machine is a whole different demon from mine. You just mentioned PCIe bus. That may be a problem too, as mine is running at 120 MHz. If I crash again, I'll lower it. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
That would explain why it had no trouble with Einstein, a more severe CPU test, and started crashing with SETI. Excellent thought! Of course my 480's are running at 860 MHz as well. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
Have you tried messing with your clock skews and your cpu pll voltages? That's what I did to get my ram working at 1780Mhz and my cpu at 4Ghz with only 1.28 volts. Another trick on my Rampage Formula has also been changing the pci-e bus to 101mhz instead of the 100mhz seemed to help stability here. I know mine is just a Core2Quad but overclocking between machines is pretty comparable as far as what settings add stability. Of course you are playing around with a lot more ram than I have (only 8Gb here) and an i7 with triple channel and two extra video cards, so your machine is a whole different demon from mine. Yeah, once you go over 10% you get into some tricky issues with video cards not being happy. Then you toss multiple GPU's into the mix and you are pulling hair out for weeks trying to nail down the issue, which from your pic might take a while to grow back to full length. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
One thing I can state easily. Heat is not a problem. In addition to the two cooling loops, there is a virtual tornado going on inside my case. I bet the PCIe buss is a wee bit too fast. It took me a long time to tune it in, and I was using Einstein wu's to do it. Einstein uses only about 7% of my GPU's, so never stressed my tweakings. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
One thing I can state easily. Heat is not a problem. In addition to the two cooling loops, there is a virtual tornado going on inside my case. I bet the PCIe buss is a wee bit too fast. It took me a long time to tune it in, and I was using Einstein wu's to do it. Einstein uses only about 7% of my GPU's, so never stressed my tweakings. Yeah, I would think Furmark would test out the GPU's but it doesn't put a lot of stress on the CPU's. It is tough to find something that stresses the whole computer the whole time in a multi threaded environment. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
One thing I can state easily. Heat is not a problem. In addition to the two cooling loops, there is a virtual tornado going on inside my case. I bet the PCIe buss is a wee bit too fast. It took me a long time to tune it in, and I was using Einstein wu's to do it. Einstein uses only about 7% of my GPU's, so never stressed my tweakings. I'm really beginning to think that PCIe buss is too fast. I don't want to lower it unless it crashes, becaus as soon as I know it's stable, up goes the CPU clock again. I love these never ending cycles. I think SETI is the best app I've seen for testing the whole system. It was down for so long, I was determined to charge out of the gate, but like the new servers, some tweaking required. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
One thing I can state easily. Heat is not a problem. In addition to the two cooling loops, there is a virtual tornado going on inside my case. I bet the PCIe buss is a wee bit too fast. It took me a long time to tune it in, and I was using Einstein wu's to do it. Einstein uses only about 7% of my GPU's, so never stressed my tweakings. Yeah, it can be a never ending bit of geek fun testing, crashing, testing and either you go insane or it stabilizes. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34748 Credit: 261,360,520 RAC: 489 |
I'd also suggest you try and pull that PCI-E bus down to, restrict it to 100-110MHz as it's just like the old PCI bus, as it'll not only corrupt video performance but also your hard drives and anything else running off it (no more than 10% o/c is the rule of thumb generally accepted though some will try and push the boundaries). Cheers. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I'd also suggest you try and pull that PCI-E bus down to, restrict it to 100-110MHz as it's just like the old PCI bus, as it'll not only corrupt video performance but also your hard drives and anything else running off it (no more than 10% o/c is the rule of thumb generally accepted though some will try and push the boundaries). I couldn't wait for it to crash again, so I lowered it 5 Hz. I also kicked up my CPU clock to 4.3 GHz. I bet it will be fine now. I will watch the CPU temp and make sure it doesn't exceed the spec. I really think that was the problem. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
Brkovip Send message Joined: 18 May 99 Posts: 274 Credit: 144,414,367 RAC: 0 |
I'd also suggest you try and pull that PCI-E bus down to, restrict it to 100-110MHz as it's just like the old PCI bus, as it'll not only corrupt video performance but also your hard drives and anything else running off it (no more than 10% o/c is the rule of thumb generally accepted though some will try and push the boundaries). Well if your CPU doesn't go past 70C you are safe for 24/7 operation. |
SciManStev Send message Joined: 20 Jun 99 Posts: 6652 Credit: 121,090,076 RAC: 0 |
I'd also suggest you try and pull that PCI-E bus down to, restrict it to 100-110MHz as it's just like the old PCI bus, as it'll not only corrupt video performance but also your hard drives and anything else running off it (no more than 10% o/c is the rule of thumb generally accepted though some will try and push the boundaries). At the moment I'm only at 48°C, but it takes a while for my coolant to heat up. The aquarium chiller is an awesome device for driving the coolant below room temp. I've had this CPU pulling 150 watts, which can over power the 1/5 hp chiller, but for the moment things are looking good. I use TThrottle to protect my CPU from over heating, which the coolant won't let it do. I always strike a balance from speed and heat, making sure I never exceed spec. Steve Warning, addicted to SETI crunching! Crunching as a member of GPU Users Group. GPUUG Website |
-BeNt- Send message Joined: 17 Oct 99 Posts: 1234 Credit: 10,116,112 RAC: 0 |
That where linx comes in, run it with furmark running and you are testing cpu / gpu / then put on a mix prime95 running a blend test. If it doesn't crash it would be one heck of a test for sure. Normally I just run OCCT and put it on a full system test for 24 hours after I OC to make sure it's stable. Steve it would be hard to imagine what kind of power that pc has without seeing it in person. I can let Seti crunch in the spare cycles on my machine while playing a full screen game and have live tv going. The live tv hiccups ever so often but the machine does it! Traveling through space at ~67,000mph! |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.