OC'd E8400s Out There (aka, That dude with the 4.2GHZ) 30 replies

Please wait...

Sgt. D. Pilla

Uber Geek

50 XP

23rd October 2007

0 Uploads

3,473 Posts

0 Threads

#21 11 years ago

Nah, its not, The CPU is responsible for alot of the particle calculations and model calculations, if any game slows down in high particle areas or highly detailed models, is almost always the CPU. Every particle has to have collision detection done with every other particle on screen, hence why any area with lots of particles (Smoke, Dust, Fog, Snow, Kick back dust effects etc) has a large FPS drop over other areas. Having higher clock speeds means all these calculations can happen in a shorter space of time, meaning a shorter period of FPS drop, or even no FPS drop. In addition, Physics will run much faster, especially in games such as Company of Heroes, Crysis and other physic's intensive games. --EDIT-- Sorry C#, I have no idea wtf this computer is doing at the moment with my posts, it only happens on this PC...




Sgt. D. Pilla

Uber Geek

50 XP

23rd October 2007

0 Uploads

3,473 Posts

0 Threads

#22 11 years ago

Nah, its not




*Daedalus

A Phoenix from the ashes

50 XP

18th April 2006

0 Uploads

3,091 Posts

0 Threads

#23 11 years ago

Actually, depending on what driver version you have (177.39 and up I believe), the GPU now handles all physics calculations.




Sgt. D. Pilla

Uber Geek

50 XP

23rd October 2007

0 Uploads

3,473 Posts

0 Threads

#24 11 years ago

Yeah, but the keyword in that is 'Now' Older driver versions this wasn't so, but there is still thousands upon thousands of collision detections to perform with each and every pixel onscreen in a particle intensive area. Its not just the collisions, but also the other operations required in that scene at the time, smoke+gun effects+waves+model detail = lots of CPU lag Anyways, I've been able to get my CPU upto 4GHz so far, however its overvoltaged, Going by the link somebody posted before, It was determinded to best leave it at auto and work down from there. 1.55v is its auto setting, and so far seems stable as all hell. I tried to tighten my memory timings but it wouldn't have it, 6-6-6-18 is the best I can do :( My memory is also down to 888 now, using "2.00B" some setting in the bios that reduces the FSB of the Ram. Currently is on 200 or something, I'm going to try to increase it slightly, o so slowly. I'll let you know how I go :) PS. So far the OC is with Stepping enabled, To reduce CPU temps and power usage even if its just a little. --EDIT-- So far, 4GHz, No Stepping @ 1.424v (9x444.1) Memory @ 888MHz I've had memory at 932MHz at one stage, shame I can't get it there again... 1066Mhz, the lowest, next step up I can go reverts my BIOS. Even with voltage increases. Pft, My memory fails.




*Daedalus

A Phoenix from the ashes

50 XP

18th April 2006

0 Uploads

3,091 Posts

0 Threads

#25 11 years ago

Please tell me your voltage is no longer on auto?

NEVER do that!

The BIOS will almost always fuck that up and end up frying your CPU on you. ALWAYS set voltages manually. If you haven't already, go into your voltage settings, and change all the auto settings to manual. Set them at whatever their current values are. It'll mean a more stable system when it comes to stress testing.




Sgt. D. Pilla

Uber Geek

50 XP

23rd October 2007

0 Uploads

3,473 Posts

0 Threads

#26 11 years ago

I have already. I followed the Overclocking Clubs guide on OCing, they suggest leaving it on Auto to check if firstly your OC even worked, if so, then you go back and begin lowering it :P Im at 1.454v at the moment and stable. However just having some massive heating issues. Had my threshold reached before at 60C, god damn I hate the bios beep, then I had it hit 70C. I've found documentation from intel that says the TJMax is 100, and Thermal Temp is 72.4. I've now set RealTemp to use that TJMax which now gives me what I believe (When compared to other programs like CoreTemp and Speedfan) to be accurate readings... 37 - 39 idle @ 4.0GHz 50-55 load @ 4.0GHz under Prime95s "Small FFTs" Real Temp has been setup with an alarm to warn me when my temps get to 70C, Which is just short of Intels documentated safe maximum Thermal Junction temp. However Reducing Voltage more will hopefully knock another few degrees off those temps.




*Daedalus

A Phoenix from the ashes

50 XP

18th April 2006

0 Uploads

3,091 Posts

0 Threads

#27 11 years ago

How long is your processor stable under Prime95, out of interest?

People say six hours is the minimum. I prefer ten.




Sgt. D. Pilla

Uber Geek

50 XP

23rd October 2007

0 Uploads

3,473 Posts

0 Threads

#28 11 years ago

At the moment I'm just doing one pass. From majority of what I've read one pass is enough to reveal most errors. One pass is taking 20 minutes to complete.

Yes, I know its not enough testing, but as soon as I hit a vcore that is either not booting with (Bluescreen at Windows Boot, No Boot at all) or failing a pass, Then I'll bump it up one by one till I get an appropriate vcore. I'm at 1.35240 vcore at the moment, and the pass is half way done with no issues. I'm contemplating leaving it at that vcore though, Most 4GHz+ OC'd E8400s have a vcore of 1.3#### So far, load temps are at 55 (RealTemp @ 100 TJMax) and 59 (SpeedFan TJMax Unknown, 105 assumed)




*Daedalus

A Phoenix from the ashes

50 XP

18th April 2006

0 Uploads

3,091 Posts

0 Threads

#29 11 years ago

Happy gaming. :)




C38368

...burning angel wings to dust

50 XP

14th February 2004

0 Uploads

5,013 Posts

0 Threads

#30 11 years ago

Sgt. D. Pilla;4620955I'm at 1.35240 vcore at the moment, and the pass is half way done with no issues.

I'm contemplating leaving it at that vcore though, Most 4GHz+ OC'd E8400s have a vcore of 1.3####

Most E8400s will have a Vcore between 1.3V and 1.4V in the BIOS at load. It's only the remaining digits that really count when you're talking about maximum performance. Bumping Vcore by 0.025V can be the difference between no POST and Prime stability, but bumping Vcore by 0.1V can the difference between a CPU and an expensive piece of silicon under a copper lid.

On the issue of Prime stability, it's great for bragging rights and any situation where you must have bit-perfect calculations, always. But if you're like the majority of us, however, it means very little, and a computer can be far from Prime stable without ever endangering data or end-user stability.

I mention this because you may run into a point where you're looking at needing 0.05V+ just to get 1 or 2MHz more out the the FSB to be Prime stable, at which point the heat-performance tradeoff really ceases to make any sense at all. Food for thought.