![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
I posted this over in the memory section, but only because it stemmed from a shorter question I had. That thread died down, so I'm editing my post to move it to this more applicable forum in search of help getting my system stable at 4.0Ghz.
I built my computer, or rather the majority of it, about 8 months ago. Here are the specs of those initial parts.
PSU: Antec Earthwatts E650W
Mobo: Gigabyte P35-DS3L (I can't find the revision # on the board, at least not where it says in the manual)
CPU: Intel Core 2 Duo e8400 C0 stepping
RAM: Corsair Dominator 2x2GB 1066Mhz (TWIN2X4096-8500C5D)
GFX: Sapphire Radeon HD4870 512MB
OS: Vista Ultimate x64
When I built it, I didn't do much in my BIOS, only really setting boot order and PEG graphics primary. I made sure my RAM was running at 1066 (which it was), but didn't adjust the memory timings (which were apparently quite loose on auto - I think 5-8-8-24? or something similar), or voltages (all auto including DDR2, which came out to 1.8v). I installed Vista and the latest catalyst drivers. For the first six months, the only real optimization I made was overclocking the video card within the catalyst control center (using the auto-tune tool) which resulted in a 790Mhz/1100Mhz card for 3d apps.
I did add a sound card, keyboard, mouse, and headset to my system during the next couple months, but those were without incident. I really only started changing things this month. I bought a new case (Coolermaster HAF 932), a 40 mm fan for my NB (Scythe Mini ultra kaze), and a new CPU cooler (Xigmatek Dark Knight s1283v) + Articlean and OCZ Freeze. While I was waiting for these to ship, I did two things.
First I tried to push my video card a bit more. I downloaded ATI Tray Tools, and using ReadyDriver Plus to get around driver signing issues in Vista, I pushed the card to 840Mhz core clock and dropped my memory clock down to 1000Mhz for 3D apps. These settings seemed stable in benchmarking tests as well as game use and produced no noticeable artifacts.
Second, I adjusted my BIOS where I found out about my loose memory timings / voltage. I fixed this, and ran memtest86+ overnight with no errors to ensure the RAM wasn't going to be an issue. Over the next couple days I encountered no problems running at the correct timings, browsing the web and playing games that would probably BSoD me now.
I then took apart my computer and put on the new CPU cooler and NB fan. The build was still in my old case at this time (I was impatient, and the new case wasn't going to arrive for over a week), but temps were never and issue (and still aren't - the DK s1283v is awesome). After getting these parts on, I began overclocking. I did not reinstall windows for this endeavor. I got to 3.6Ghz without issue, and kept pushing to 4.0Ghz. Past 4.0, I had trouble getting any sort of stability, but that was okay with me. At 4.0Ghz, I managed to pass a 21+hour Large FFT torture test in Prime 95 with the following BIOS settings:
Multi: 9x
FSB: 445
PCI-E freq: 100
Memory multiplier: 2 (so 890Mhz)
Memory timings: 5-5-5-15 (with advanced all auto)
PCI-E: +.1v
DDR2: +.3v (=2.1v)
FSB: +.1v
MCH: +.2v
CPU: 1.3875v BIOS (CPU-Z: 1.344v idle, 1.328 load)
I began trying to minimize voltages. I dropped PCI-E down to Normal and passed Large FFTs for 3 hours before guessing this was fine. I tried lowering MCH to +.1v and decided to dual stress test for some reason. I ran P95 Blend test and Intel Burn Test for 2 hours without errors, but had a BSoD when closing one of them. I reverted MCH to +.2v and tried FSB at normal, but BSoD's out of P95 a couple minutes into a Blend test.
I decided to stick with the above settings, except with PCI-E at normal. I upped the memory multiplier to 2.4 to give me 1068Mhz and ran an overnight memtest with no errors. I really really thought I was done here.
Then over the next couple of days, I ran into some issues. Firefox would occasionally close and give me an error. Eventually the problem got a bit worse, and the BSoDs started happening, most frequently when opening pages with streaming video on them, but also occasionally during normal browsing as well. Incidentally, my case finally arrived around this point and I swapped everything into it. Nothing of any notice happened relating to this change though.
I've tried a whole lot of things to get the system stable, but nothing seems to work. I didn't log everything I tried (which I regret now), but from memory I can tell you, I tried MCH at +.3v, DDR2 overvoltage to +.4v (to 2.2v), loosened some of the advanced memory timings, ran the ram at 1:1 (890Mhz), and upped vcore to 1.4v, one change at a time for most of these of course. They all left me with the exact same issue, relative stability in stress testing programs and crashes in actual use during browsing.
I began logging my changes again last night with an attempt at PCI-E at +.1v. Initially I set my system to settings completely identical to what they were for the 21+hour Large FFTs mentioned earlier (including the 2x memory multiplier). This system seemed stable for a while, so I set the multiplier back up to 2.4x for 1068Mhz for an overnight test, but didn't get that far. I errored on core 1 about a minute into an Orthos Blend test (while simultaneously running P95 Small FFTs without error). This caught me off guard as it was the first time I've gotten an actual torture test error in a couple days, but I made some adjustments. I pushed vcore up to 1.4v in BIOS (1.36idle / 1.344load CPU-Z) and tried again at an overnight test. I watched it for 30 minutes or so of the dual stress test programs, then went to sleep.
I woke up to a restarted system. I tried lowering the memory multiplier back down to 2x (890Mhz) and upped the FSB to +.2v, but was met once again with BSoDs . I've since also tried pushing the CPU as high as 1.4125 in BIOS (1.376 CPU-Z idle) and loosening my timings including static tRead, both to no avail. The only way I've managed to prevent the BSoDs is lowering my core clock. I have no issues for instance at 3.0Ghz or even 3.6Ghz. Both are perfectly stable, in torture tests and actual use.
So where should I go next to get a 4.0Ghz system? Is there any indication that I should be noticing of where the problem is? Is it safe, or rather, advisable to push the vCore further? Should I just give up and be happy with a stable 3.6 or 3.8Ghz?
I apologize for the length of this post, but I figured the more the detail the better for those more knowledgeable at this than I am. Any help would be appreciated!
I built my computer, or rather the majority of it, about 8 months ago. Here are the specs of those initial parts.
PSU: Antec Earthwatts E650W
Mobo: Gigabyte P35-DS3L (I can't find the revision # on the board, at least not where it says in the manual)
CPU: Intel Core 2 Duo e8400 C0 stepping
RAM: Corsair Dominator 2x2GB 1066Mhz (TWIN2X4096-8500C5D)
GFX: Sapphire Radeon HD4870 512MB
OS: Vista Ultimate x64
When I built it, I didn't do much in my BIOS, only really setting boot order and PEG graphics primary. I made sure my RAM was running at 1066 (which it was), but didn't adjust the memory timings (which were apparently quite loose on auto - I think 5-8-8-24? or something similar), or voltages (all auto including DDR2, which came out to 1.8v). I installed Vista and the latest catalyst drivers. For the first six months, the only real optimization I made was overclocking the video card within the catalyst control center (using the auto-tune tool) which resulted in a 790Mhz/1100Mhz card for 3d apps.
I did add a sound card, keyboard, mouse, and headset to my system during the next couple months, but those were without incident. I really only started changing things this month. I bought a new case (Coolermaster HAF 932), a 40 mm fan for my NB (Scythe Mini ultra kaze), and a new CPU cooler (Xigmatek Dark Knight s1283v) + Articlean and OCZ Freeze. While I was waiting for these to ship, I did two things.
First I tried to push my video card a bit more. I downloaded ATI Tray Tools, and using ReadyDriver Plus to get around driver signing issues in Vista, I pushed the card to 840Mhz core clock and dropped my memory clock down to 1000Mhz for 3D apps. These settings seemed stable in benchmarking tests as well as game use and produced no noticeable artifacts.
Second, I adjusted my BIOS where I found out about my loose memory timings / voltage. I fixed this, and ran memtest86+ overnight with no errors to ensure the RAM wasn't going to be an issue. Over the next couple days I encountered no problems running at the correct timings, browsing the web and playing games that would probably BSoD me now.
I then took apart my computer and put on the new CPU cooler and NB fan. The build was still in my old case at this time (I was impatient, and the new case wasn't going to arrive for over a week), but temps were never and issue (and still aren't - the DK s1283v is awesome). After getting these parts on, I began overclocking. I did not reinstall windows for this endeavor. I got to 3.6Ghz without issue, and kept pushing to 4.0Ghz. Past 4.0, I had trouble getting any sort of stability, but that was okay with me. At 4.0Ghz, I managed to pass a 21+hour Large FFT torture test in Prime 95 with the following BIOS settings:
Multi: 9x
FSB: 445
PCI-E freq: 100
Memory multiplier: 2 (so 890Mhz)
Memory timings: 5-5-5-15 (with advanced all auto)
PCI-E: +.1v
DDR2: +.3v (=2.1v)
FSB: +.1v
MCH: +.2v
CPU: 1.3875v BIOS (CPU-Z: 1.344v idle, 1.328 load)
I began trying to minimize voltages. I dropped PCI-E down to Normal and passed Large FFTs for 3 hours before guessing this was fine. I tried lowering MCH to +.1v and decided to dual stress test for some reason. I ran P95 Blend test and Intel Burn Test for 2 hours without errors, but had a BSoD when closing one of them. I reverted MCH to +.2v and tried FSB at normal, but BSoD's out of P95 a couple minutes into a Blend test.
I decided to stick with the above settings, except with PCI-E at normal. I upped the memory multiplier to 2.4 to give me 1068Mhz and ran an overnight memtest with no errors. I really really thought I was done here.
Then over the next couple of days, I ran into some issues. Firefox would occasionally close and give me an error. Eventually the problem got a bit worse, and the BSoDs started happening, most frequently when opening pages with streaming video on them, but also occasionally during normal browsing as well. Incidentally, my case finally arrived around this point and I swapped everything into it. Nothing of any notice happened relating to this change though.
I've tried a whole lot of things to get the system stable, but nothing seems to work. I didn't log everything I tried (which I regret now), but from memory I can tell you, I tried MCH at +.3v, DDR2 overvoltage to +.4v (to 2.2v), loosened some of the advanced memory timings, ran the ram at 1:1 (890Mhz), and upped vcore to 1.4v, one change at a time for most of these of course. They all left me with the exact same issue, relative stability in stress testing programs and crashes in actual use during browsing.
I began logging my changes again last night with an attempt at PCI-E at +.1v. Initially I set my system to settings completely identical to what they were for the 21+hour Large FFTs mentioned earlier (including the 2x memory multiplier). This system seemed stable for a while, so I set the multiplier back up to 2.4x for 1068Mhz for an overnight test, but didn't get that far. I errored on core 1 about a minute into an Orthos Blend test (while simultaneously running P95 Small FFTs without error). This caught me off guard as it was the first time I've gotten an actual torture test error in a couple days, but I made some adjustments. I pushed vcore up to 1.4v in BIOS (1.36idle / 1.344load CPU-Z) and tried again at an overnight test. I watched it for 30 minutes or so of the dual stress test programs, then went to sleep.
I woke up to a restarted system. I tried lowering the memory multiplier back down to 2x (890Mhz) and upped the FSB to +.2v, but was met once again with BSoDs . I've since also tried pushing the CPU as high as 1.4125 in BIOS (1.376 CPU-Z idle) and loosening my timings including static tRead, both to no avail. The only way I've managed to prevent the BSoDs is lowering my core clock. I have no issues for instance at 3.0Ghz or even 3.6Ghz. Both are perfectly stable, in torture tests and actual use.
So where should I go next to get a 4.0Ghz system? Is there any indication that I should be noticing of where the problem is? Is it safe, or rather, advisable to push the vCore further? Should I just give up and be happy with a stable 3.6 or 3.8Ghz?
I apologize for the length of this post, but I figured the more the detail the better for those more knowledgeable at this than I am. Any help would be appreciated!