PDA

View Full Version : MSI 7800GTX throttling issues??


orbitech
09-13-05, 09:59 AM
Hi first of all I would like to greet everyone in the forums as a new member.

4 days ago I builted my rig which you can see in my sig. I would like to mention that cpu and memory and rock solid stable and checked for 1 day at least in torture tests.

2 days before I noticed some wierd issues with my gpu. Lets take it from the beginning.

When I first boot on a cold pc(meaning that is closed for hrs) and ran some benches eg Aquamark3 and 3dmark05, I get normal scores from my gpu. Temps are a little higher but not that high from normal. But after some time if I try to ran again the benches my gpu throttles by a very significant margin. (Eg. in Aquamark3 from 90fps it falls even to 45fps!! and in 3dmark2k5 from 8200 to 5000!!).

Well I tried to find why these throtling issues occured, and of course I suspected temp issues. Here I have to mention that in real gaming conditions and not in benchmarking the throttle does not appear or it's not that high by any means. I mean I have almost the same fps if I ran FRAPS in the background with cold or hot gpu.
Anyway today I randomly looked at details of gpu in 3dmark05 and I noticed some major variations and o/c clock speeds every single time I ran the program!!!
I mean one time it shows 275(default 2d core clock speed, another 450 and another 465!!) I'm not o/cing and I have Coolbits to no o/c. No other third party 3d applications like Riva tuner are installed to mess around with the drivers!!
I also have 6.66 Nforce installed , AMD driver for DC, and gpu drivers are 78.01

I'm very confused here. My questions are:

1. Why do I have throttling issues @ default speeds when idle temps are not that tragically high? XP90c might be driving hot air to the gpu but can it cause such a high throttling ? Cse temps are not that high either but I haven't recorded load temps yet.
2. Why every time in 3dmark05 the core shows different values and more importantly O/C values when I don't o/c???
3. I remember an issue with some boards and NV4x cards that mobos o/ced automatically the gpus, and this can be a reason of why this is happening to me.
You should disabled in the bios something in order for the board not to o/c the card. Can something like this be happening to my board?
4. Plz help me with anything else that you might think of happening here. This is the first time I ran into something like this and I need some help.

Thanks in advance. Hope I didn't tire you with my long post.

Dr.Nick
09-13-05, 11:07 AM
Doesn't the MSI auto OC depending on temps? I would install RivaTuner and have it monitor the clocks while you run a few benches and games.

orbitech
09-13-05, 11:19 AM
Doesn't the MSI auto OC depending on temps? I would install RivaTuner and have it monitor the clocks while you run a few benches and games.

I never installed MSI drivers that come with DOT. From scratch I'm using 78.01 Forceware

100fuegos
09-13-05, 11:33 AM
Don't know of any motherboard that overclocks any graphics Card by itself, may be wrong though. What 3DMark2005 says is useless, you don't know for sure that your GPU is throttling it may be anything even drivers. Install Rivatuner and set it up to BackgroundMonitoring and you can control then Core Temperature and Core Clock in real time. Get back to us with what you find.

OWA
09-13-05, 12:24 PM
Hi first of all I would like to greet everyone in the forums as a new member.

4 days ago I builted my rig which you can see in my sig. I would like to mention that cpu and memory and rock solid stable and checked for 1 day at least in torture tests.


I guess I'll post my standard comment for owners of X2 CPUs since I got bit by it. I didn't have this exact problem though so this might not be any use to you. If you haven't already, make sure you install the AMD X2 CPU driver. I didn't have exactly that problem but I had a lot of weird problems that almost exclusively showed up when trying to run games, 3D benchmarks, etc. I could run CPU and memory tests all day and things were fine. So, that would be the first thing I'd check (or double-check).

This might not be the issue you're seeing now either but another issue you may have to deal with is that some games don't like the X2 CPUs. Some of the symptoms may be everything running in slow-motion or everything running extremely fast or a combination of running extremely fast, slowing down for a moment or two and then running really fast again. If you're having really weird issues like that, try using the utility "runfirst". It basically sets things so that the game will run on one CPU (core). It doesn't work with all games though so sometimes you'll still need to jump out of the game and use the affinity option.

orbitech
09-13-05, 01:21 PM
Guys I know all these stuff, BUT...

I mentioned that the vga in a cold boot runs these specific applications normal.
After some time I 'm starting to have the issues of 50% throttling or so...

I checked RivaTuner core temps too, and they don't seem to go >58-59

I have Aerocool Panel too, and thermal sensor shows no greater than 65 C under heavy load. Let's add to this a max 10 c that the core might be and we reach 75 max. Should the core throttle 50% in 75-80max @defaults speeds??? :thumbdwn:

And if the card was defective shouldn't I see a sign? A BSOD, a random lockup or something else at least for one time? Nothing!! Just throttling...

Could this be a sign of something in my pc makes my gpu to throttle @ mediocre temps? (vga bios,system bios or something else)

Dunno what do u think?

orbitech
09-14-05, 01:37 PM
Finally I got Rivatuner to show me the right core tmps.And it wasn't good at all. The max core temp was 78-80 c. Yikes!! At default clocks.

Tommorow I'll try to check the pc out of the box so case temp or cpu cooler driving hot air to the gpu will not be issues to consider. If temps remain high and throttling appears this will definately mean that gpu is defective and needs RMA. Otherwise if everything is normal I'll consider a mod with a fan or something carefully or an aftermarket cooler. I'll need your opinions on this last thought. I'll give you some feedback 2morrow.

rohit
09-14-05, 02:21 PM
sounds like something is fishy with your card!

orbitech
09-16-05, 02:29 PM
Well I made some progress on what is happening but I'm unable to grasp what is happening here.

Every time I'm opening my pc for 1-2 hrs straight, I'm running benchmarks and performance is not falling no matter how much I stress the gpu. BUT if I reboot or close the pc and open it w/o leaving it the gpu for some time to cold then performance will fall from the first benchmark.

What is this thing? Why is it that no matter how I stress my gpu on a cold boot there's no way for performance to fall, but when I restart performance will fall? :confused

This is the most weird behavior I've seen from a gpu ever. I mean if it's overheating why the hell this is not depicted on the first boot? Why pc has to be restarted in order for performance to fall? I'm going nuts here... :eek:

rohit
09-16-05, 03:05 PM
RMA it, dont struggle to understand its behaviour.
Its a GPU after all.

orbitech
09-16-05, 03:37 PM
RMA it, dont struggle to understand its behaviour.
Its a GPU after all.

Yeah I think I'm gonna do that, but I just want to be sure that it is an overheating issue and not something else...
I don't want to wait a week or more and have the same issues with another 7800gtx too..

rohit
09-16-05, 04:27 PM
well, your temps reaching 80*C is again a bad sign (at stock speeds).
most of the guyz here have their GTX working at 65*C - 70*C load max.
but again 80*C is not tooo high either.

i feel in your case, its either a bad GPU or few components on the PCB are messed up.
so either ways you need to RMA (High temp, and that wierd throtling problem).


P.S: did u try re-installing windows and trying? if not just try, re-install windows, and get the latest whql drivers supporting g70-gtx.

hope this helps

orbitech
09-16-05, 05:06 PM
well, your temps reaching 80*C is again a bad sign (at stock speeds).
most of the guyz here have their GTX working at 65*C - 70*C load max.
but again 80*C is not tooo high either.

i feel in your case, its either a bad GPU or few components on the PCB are messed up.
so either ways you need to RMA (High temp, and that wierd throtling problem).


P.S: did u try re-installing windows and trying? if not just try, re-install windows, and get the latest whql drivers supporting g70-gtx.

hope this helps

Yeah , I do understand that temps might be high, but WHY ON EARTH don't I notice fall in performance no matter how I stress the gpu on the first boot? Even if I o/c @485/1300 I'll get the appropriate score, as long as I'm staying on the first boot!! Why this happens ONLY WHEN I RESTART when supposingly temps should be lower? That is beyond my comprehension... :confused:

Anyway yes I did tried a fresh install on the other HDD same thing... I'm using 78.01 RIGHT NOW, MIGHT TRY A 8x.xx driver to check out..
Thanx m8 ...

rohit
09-17-05, 02:10 PM
do keep us updated.. this is intresting

SigmaOrionis
09-17-05, 08:04 PM
If you've tried everything and your sure that your cards clockspeeds are drifting I would RMA.

It could be anything. Like maybe a cap that provides power to the clock regulating chip. After its been going awhile it starts getting flaky. Of course I'm making this up but I'm just trying to make the point that when hardware has a marginal problem its almost impossible to diagnose.

To give you an example of weirdness. I had a 6800GT that would work for 3 days in my computer. Then on the 3rd day without fail it would lock up everytime I tried to use any 3d application. If I took the card out and reseated it it would work for 3 days fine and then would lock up in any 3d application. I then tried my friends identical card in my comp and it worked flawlessly.

This literally drove me crazy and had me blaming everything else in my comp thinking maybe I had a faulty AGP slot that had a bad sodder joint and the card would sag and break the connection. I mean I didnt want to blame the card. After RMA put in new card. No more problem.

Is there anybody you know that would let you borrow there card to try in your system. That would be the quickest way to determine if its your card or something else.

Good Luck.

orbitech
09-18-05, 09:36 AM
Well I tested my vga on a friend's pc and it seems to be working flawlessly. Temps were high at his pc too, but no throttling whatsoever... Now I don't know what to look for anymore. The card is fine but I'm continuing to have issues.. Could this be related to my psu(Antec Neopower 480w)?
Plz share an idea I'm goin nuts here :thumbdwn:

rohit
09-18-05, 06:45 PM
Good, you did that. (checking on friends PC)
It could be anything, PSU, Mobo etc.
what VGA does your friend use, try His VGA in your rig, check if it has the same behaviour as your 7800gtx.
so you will be assured its either Mobo, or PSU.

Peoples-Agent
09-20-05, 05:30 AM
Hmm....

I have been having the same problem too in Riddick using the 77.77 drivers, the ones you are using are exactly the same except support for another card or something.

I wonder if that's the problem.

Will try installing different card drivers than let you know, i'm using a Gigabyte 7800GT.
All of a sudden now and then the frame rate slows right down sorta thing.
Only noticed it in this game so far.

Peoples-Agent
09-20-05, 05:56 AM
Ok, I have just installed the 78.03 drivers and it didn't do it this time, only option enabled was V Sync on, beforehand I was overclocking plus had Triple Buffering turned on using the 77.77s which are basically the same as the 78.01s.

This time I will overclock my card again but without Triple Buffering turned on this time and see if it does it again.

Drats, it just did it again but overclocked.... seriously starting to look towards a heat issue, it didn't like that when overclocked.... time to try without.

....Ok, it seems to be doing it like when a new cutscene or section of the game starts, strange.... no idea. Huge FPS drop for no apparent reason, not noticed this happen in any other games.... like wtf!

Peoples-Agent
09-20-05, 08:48 AM
Gaaah.... turns out it was mpftray.exe with McAfee that was causing a 100% cpu usage spike.