Help diagnose stability problems with games - is it heat?

[spidey07]

I can run this machine for days on end with prime95 and seti. That seems to tell me that the processor is stable.

But for some reason games freeze after about 20-30 minutes of play. If the games run this long and then fail is my only chance for stability better cooling? The processor rarely tops 38c under full load.

I'm sure you fellas have run across this scenario, should I just back down to 1 gig and be happy or am i missing some kind of AGP bios setting? Ram CAS 2-2-2 and 3-3-3 don't make a difference.

thanks bunches,
spiderman....swingin from grapples and railin the baddies

 

[Chuffmaster2k]

I'd say back down to a gig. Seriously though, what kind of performance increase do you see from 1000 Mhz to 1050 Mhz? Back it down if it is more stable. Good Luck.

 

[spidey07]

Your right about no noticable performance increase. However i'm pushing it for the same reason I supe up my car, cause it can go faster!

Seriously though, is this a sign of needing to run the CPU cooler? I'm confused because all other stress tests succeed except for games and 3dmark. I thought the Creative Labs Geforce2 could handle high bus speeds, even up to 99-100. (bx board)

I'm gonna put a 866 processor on her and run some more tests at 150 FSB just to isolate the processor as the cause.

 

[Brig]

Since/if it only freezes in games, my guess would be that the graphics card is the problem, not the cpu. With the BX133 Raid, that agp bus is at 99 mhz. What graphics card are you using ?

Also, on a side note, what HS/F are you using on the processor ?



And just as I was typing the graphics card question, you were already typing the answer.
Do your games run ok at 133 FSB ?

 

[spidey07]

thanks for the help brig...even at 133 FSB/89 AGP everything is flawless (literally, can't make it crash). This being my fifth overclock I'm looking for the finer art of stability. Believe it or not I've got a BS in CS/EE from Purdue and I can't for the life of me figure out why a processor would behave differently between floating point operation during prime95/seti and games.

My only conclusion leads me to belive that games use the full rage on instructions...hence hitting just about every transitor in the proc. Cache, memory, RAW I/O (agp/mem), and FPU. What gives? I would really love to hear a full explanation of the phenonemon as it occurs on my buddies' CUSL2/700e as well. Are we simply asking too much from the procs or are there other things like AGP aperature size that I need to take into account?

My question still stands...Is this symptom most likely do to cooling. Again my proc does not top 38c.

Thanks again friends,
spidey
ps - long live BX, I just love the fact that every single component in my 'puter is OC'd

 

[spidey07]

Forgot one more thing BRIG...using alpha pal35.

DON'T LET ANYBODY TELL YOU DIFFERENT - ALPHAs COOL BETTER THAN ANY OTHER HSF OUT THERE. Period. Point Blank. Ask anybody that has actually tried dozens of heat sinks.

Examine the physics. Newton's thermodynamics teach us that more suface area = more cooling, actually linearly (somebody correct me please, I've been out of school for 8 years).

Text

 

[Brig]

Spidey07

Thanks for the strong recommendation on the PAL35. I need to get my cpu temps where yours are, on the same mobo, and there's no better recommendation than successful experience.

I agree, stability is everything.

I'm not able to intelligently participate in a discussion of theory when it comes to these chips, busses, etc. I tend to think more in terms of a 'Block Diagram' than a 'Schematic'. However, I am very pragmatic when it comes to this stuff.

Nothing pushes a pc to it's limits more than a complex, great looking game. A lot of people think that business apps force a cpu to work hard. They are, however, NOTHING in comparison to what's going on in a pc when a good game is running.

Given: (1) your cpu runs fine at 1050, (2) freezes only occur during a game (most graphics intensive app), after a period of time, at 150 mhz FSB, (3) whole system (including games) runs flawlessly at 133 mhz FSB over an extended period of time

From the above, I would think that heat is probable as the cause of the problem but.....heat in the graphics card, not the cpu. That cpu can comfortably handle considerably more heat than your current temps and, it supposedly can also run about a hundred mhz faster than what you're pushing it now. I think that video card is just running so far out of design specs at 99 mhz that it can't handle the strain.

If you could somehow put a BIG cooler of some kind on the graphics chips (don't know the best way), then step it up a notch at a time, you 'might' hit 150 FSB with stability. However, you might also kill your video card in the process. Is it worth it for 17 mhz FSB ? I dunno, that's your call. ( I once thought it was worth it to take a motorcycle over 100 mph, just to see if I could do it. Would I do it again ? NO WAY. )

Good luck and, if you don't mind, I'm really interested to know how you make out with this.

Also agree that this is a nice board. Just too bad about BX chips with no 1/2 divider on the AGP. Oh well, it's certainly worth the trade-off.

BTW- I would also like to get a good explanation of what exactly the AGP Aperture Size in the bios does, and what it's effects are. Haven't been able to find much about it.

 

[jr59984]

If you are only using 1.7 volts and can get it to 1050, you got a sweet processor. I run mine at 1002 MHZ 1.85 volts with an alpha. It has been running this way since I bought it no problems. Bump it up to 1.75 or 1.85v then see what she can do. I guarantee if you are running it at 1050 1.7v it will make it stable at 1.75 or 1.8. Hell, you still have a lot of headroom. If it was me, I would crank it to 1.85 and see what it really has in it. Everyone I have talked to agrees 1.85 is a good voltage and if it burns out in three years it'll be worth $40 anyway. Or I'll take a trade for my chip if you only want a gig (

The only other thing is it could be the vid card set the processor to 100 or 105 x7 and agp divider to 1/1 and run some test. If it handles it, you know it is not the vid card you need more volts on the cpu or it is the ram. Also check your I/0 voltage in softmenu3 I think the default is 3.5v . Increasing this to 3.6 or 3.7 has enabled some people to gain stability at higher speeds since this feeds a little more juice to the ram and the vid card might want to try that also.

 

[NoNicK]

I have a similar problem with my setup.
PIII-700 with a PEP66 on a Abit Slocket and Abit BF6

I've had it going at 1015Mhz running seti and prime95 all night (bout 8 hours) without a hitch, but the second I get into any game it locks hard. I thought it was the vid card at first not likeing the high AGP speed, but I tested the card all the way upto 125Mhz AGP (1/1 ratio) and it works flawlessly upto 125 so thats not the problem. I'm at a loss for what the problem is. the CPU never gets above 45c and the case temperature never gets above 30.

Any thoughts?

 

[nyati]

Hey,

The AGP Aperture setting in the BIOS tells the BIOS how much of the system RAM it can give to
the AGP card. (example: If you set it to 64MB, then the video card will be allowed to use up to 64MB of the system RAM for texturing, but not more.) It is a good way to limit how much RAM your
AGP hogs. I've always heard that the best setting is 1/2 of the total system RAM.

Have Fun.

 

[Brig]

Thanks nyati
Hmmm..."amount of system ram the video card 'can' use". Wonder if that means it reserves it exclusively for the video card to use, or if it's just available anytime the video card needs it.

Also wonder if this somehow plays into Spidey07's original question.....just thinking out loud.

 

[spidey07]

Well thanks for your help guys. I think the processor is the limiting factor here. I tried a 866 (6.5x133) and bumped the bus up to 150 CAS2. The box ran super stable even in games at 975 (6.5x150)

Boy did that proc run hot, 49c under load!

I'll back down the speed of my 700 to 1000 or 1015, but at least I know my system can handle a 150 front side bus I don't see the need to start doing esoteric chipset/clock cooling if I can run fine at 150?

thanks again
spidey

ps BRIG - I believe the AGP aperature size is used to hold textures in memory, one of the new features of AGP. This way textures can go directly from memory to the agp bus.

 

[spidey07]

Double post..whooops

 

[lostnthenet]

Spidey, Your saying that 49 degrees is too hot? I am running my Duron 650 @ 800 MHz and it is running rock solid at 50 degrees idle and arounf 55 under pressure.

 

[cabojoe]

Sounds like u have mastered the fine art of stability....stability occurs when ya don't lock up...backing it down from some ridiculus dream u have is one good way to obtain stability.

Athlon 850 @ 1007 1/3 L2 1.9 volts
Abit KA7
256 meg PC-133
Ninja Micros FreeSpeed Pro goldfinger
GlobalWin FKK32 Heatsink/Fans
Creative Annihilator 2 GeForce GTS
Sound Blaster Live
MS Digital Sound System
HP 8x4x32 Burner
48x CD-ROM
Kingston Nic
WireSpeed DSL Modem

 

[JawaJedi]

This may be a long shot on your setup, but since it is only your 3d programs crashing, could it be that your gf2 doesn't have the proper cooling?

-jawa

 

[eia430]

Spidey, I've got two things that might be worth checking into.... First off if you have a memory diagnostics program use it to check to see if your ram is good. If you don't have one a free d/l is here (http://reality.sgi.com/cbrady/memtest86) Second, if your ram checks good go into your system monitor (start,prgrmfiles,accesories,systools,sysmon) and track free physical ram. Run the game (while sysmon is running)then quit the game, see if the game is not letting go of memory (memory leak) Of course this is assuming that you have the latest drivers (vidcard,direct X,sound card,etc.) Good luck with finding the bug.

 

[eia430]

Duh!! why didn't I think of that? Jawa Jedi is right, it might be a good idea to see if your vidcard fan is running well and see if the problem isn't the vidcard chipset overheating when 3d accel goes into overtime.

 

[spidey07]

Thanks a bunch. I'll mount a thermistor on the Gforce2 chip and monitor. I thought that you would at least see some tearing or visual artifacts if the core was overheating, but I don't. Game just freezes.

Like I said in my last post the system is perfectly stable at 150 fsb.

spidey

 

[eia430]

Oh yeah, a few other suggestions.... I've diagnosed lockups before to the most unusual things. One was from a faulty internet keyboard(defective) and another was to a loose power connection to the harddrive. Another was because the computers I was fixing was located in an auto shop and they had (electric)air compressors turning on and off giving all sorts of power spikes. Just start eliminating causes one by one till you find the culprit. Good luck... I'll post if more comes to mind...