e8400 past 4ghz

System: e8400. 1.25 vid. Motherboard: evga 750i ftw. OCZ Reaper 1066mhz 5-5-5-15.

1st attempt: 4.356ghz @ 1.376 vcore was stable for 2h49m in Prime95 until 1 core halted.

I seem to have found the sweet spot for the motherboard: 2076mhz fsb. I paired this with an 8.5x multi and ram @ 1038 (1:1). So it works out to about 4.409ghz right now, running with 1.392 vcore under load and 1.412 at idle (1.45 in BIOS). My max load temps are bordering 59 and 60 and 33 at idle with a Xigmatek HDT-S1283.

2nd Attempt: Prime95 ran for 2h28m then one of the cores got an error and halted.

I usually don't stay on my PC for more than 2 hours at a time (especially not at 100% load), so is this okay? I have also tried 519x8 which is about 4.15ghz, which I think is more stable but I haven't tested it yet. The only benefit I see from this is the need for less vcore (1.356 load) and obviously in turn cooler temps.

Also, I think this motherboard is really fussy. When I leave the fsb voltage at 1.2, it seems okay up to 4.1-4.2ghz. When you up it to 1.25v volts, however, it feels like it is overheating just by feeling the heatsink on the NB. So when you are needing 1.38+ vcore and need the 1.25 volts on the fsb to have the system working nicely, it seems to overheat after just 5-10 minutes during stress testing. What I discovered when I was looking at the NB heatsink is that the heatpipe is coming out from the middle of the fins. It's like having your CPU HSF with the heatpipes starting halfway up the fins. Pretty stupid design, and it would have cost evga just as much to put it in a good spot: on the chip.

So I now know that 1.2 volts on the fsb is max, which is purely due to the heat, but I think I need to establish the max vcore that 1.2 volts on the fsb will withstand? Am I on the right path here? In games both 4.356 and 4.409ghz are fine for just over an hour (usually the max time I play a game in one sitting, so to speak). Oh and I used Prime95's small FFT's test for both stress tests, if that makes a difference. I know my ram is good, too. Help, please

What I found that worked was to first establish a stable FSB, then work on the CPU. Once beyond about 470MHz, my Gigabyte EP45-UD3P required additional NB voltage and tweaking of the GTL reference levels.

What I would do is leave everything the same, but change the CPU mult to x6 and see if that passes P95 blend. Blend is harder on the memory than small FFT.

Once you have a stable FSB, then turn up the CPU multiplier and test with P95 small FFT.

For reference, I have the following P95 SFFT, LFFT, Blend and Memtest86 stable using an E0 E8400 on the above mentioned GigaByte board:
4250CPU/500FSB/1000 5-5-5-15 memory (8.5x cpu mult)
CPU PLL 1.5V
CPU TERM (VTT) 1.26V
CPU REF: 0.798 (63.3%)
CPU VCORE: 1.3625V

DRAM Voltage: 2.04
DRAM TERM: 0.9
CHA/CHB DRAM REF: 0.9

MCH CORE: 1.26V
MCH/DRAM REF: 0.9
MCH REF: 0.756 (60%)
ICH IO 1.5
ICH Core 1.1V

Key to getting the above working was the 1.26V VTT and the MCH REF. The system may go faster but it will require more CPU voltage and likely more MCH voltage as well.

Get your FSB established first.

-Mike

TexasOC said:

What I found that worked was to first establish a stable FSB, then work on the CPU. Once beyond about 470MHz, my Gigabyte EP45-UD3P required additional NB voltage and tweaking of the GTL reference levels.

What I would do is leave everything the same, but change the CPU mult to x6 and see if that passes P95 blend. Blend is harder on the memory than small FFT.

Once you have a stable FSB, then turn up the CPU multiplier and test with P95 small FFT.

For reference, I have the following P95 SFFT, LFFT, Blend and Memtest86 stable using an E0 E8400 on the above mentioned GigaByte board:
4250CPU/500FSB/1000 5-5-5-15 memory (8.5x cpu mult)
CPU PLL 1.5V
CPU TERM (VTT) 1.26V
CPU REF: 0.798 (63.3%)
CPU VCORE: 1.3625V

DRAM Voltage: 2.04
DRAM TERM: 0.9
CHA/CHB DRAM REF: 0.9

MCH CORE: 1.26V
MCH/DRAM REF: 0.9
MCH REF: 0.756 (60%)
ICH IO 1.5
ICH Core 1.1V

Key to getting the above working was the 1.26V VTT and the MCH REF. The system may go faster but it will require more CPU voltage and likely more MCH voltage as well.

Get your FSB established first.

-Mike

Click to expand...

I have already found that 519 fsb works. The problem I am suggesting is that for 519 fsb I need 1.2 volts. Now this isn't an issue until I start pushing the cpu and it needs more voltage. Once that happens, it needs more nb voltage, and when I give it more nb voltage, it is stable until the northbridge overheats.

Also what would cause one core to stop during the test? Overheating or lack of voltage on the motherboard/cpu? I find it weird that it is 100% fine in games for hours and then in the stress testing one core will stop but I can simply stop the test and keep using the it with no problems, not needing to restart. Once again: 1.2v fsb, 1.5v spp, 1.25v nf200, 2.2v ram, and the vcores I listed earlier. Just seems to be one last tweak of some kind to get the system stable, as it seems it is totally stable besides the stress testing after ~3 hours.

I am re-testing the fsb at 2076. So far it is good. I also changed the fsb voltage to 1.25 to match your suggestion of 1.26 (can only go up in .05v increments) and it seems to have helped. It has ran for about 6 minutes without any errors, usually myself I think after 10 minutes if nothing happens then it should be fine (for the motherboard, at least). I will bump up the multi to 8.5 if it lasts 10 minutes then blend test it again for an hour and hopefully get no errors.

This is just weird now. It is totally fine with the 6x multiplier in blend but when I set it to 8.5x it will pass cpu test but not blend. What is making my memory fail when I bump up the multiplier? It is starting to get on my nerves.

your NB voltage is probally screwing with you ... either that or lower your timings + up your VRAM by a little bit