First - I apologize for the length of this post, but I have been doing a ton of testing as well as talking to folks from PCT (sales and Tech). I am wondering if anyone else is having this issue.BTW - this is NOT a complaint - my 9880 is working great but has an odd quirk with benchmarks running better when the GPU is hot! My system (and benchmark) is listed below.
What I found was that when running benchmarks (3dmark05, 3dmark03 and Counter Strike Video Stress test) I was getting VERY low scores (lower than my 9860) when the laptop was first booted and running cool. The specifics are below, but FIRST - before everyone suggests this, I have already checked / verified / ruled out the following:
1 - I am running on AC at all times (connector is not loose).
2 - I have the latest Nvidia driver - 76.50 - came installed.
3 - I have DirectX 9.0c.
4 - I have the latest GPU bios.
5 - I have the latest system bios.
6 - I do not have any programs running in the background aside from the bare minimums.
7 - I do not have any spyware / adware / viruses on my system.
8 - I have not done ANY registry hacks / overclocking.
9 - The fans are running fine (more detail on this below).
10 - Changing refresh rates makes no difference in results.
11 - I am running in full res (1920 x 1200)
12 - I do not have anti-aliasing / aniosotropic on
13 - I am using DEFAULT settings on 3DMARK05 (pro version).
Ok - so here is what happens. Lets start with 3dmark05 - the very first time I ran it, the laptop had been running for a bit and it was warm in the room (I dont know what temp it was since I didnt have any GPU temp monitoring utilities). I got a benchmark of 5070! GREAT! I let the laptop sit for an hour and ran some errands. Came back - ran it again, this time I got 3300??! WTH?! Nothing was changed! Now - after running it again several times - I was then able to "predict" when I would get a high (5000+) or low (<4000) benchmark by looking at the FPS during the first few seconds of Proxycon (first video benchmark scene). I would get 30+ FPS for the high bench and 8-10 FPS for the low bench scores. This was VERY repeatable.
In addition - I would always hear the fans "ramp up" during the higher benchmarks - always. Kinda seemed like the GPU was throttled down and lower temps. (more on this below).
As a verification - I also ran 3dmark03 and got low scores as well (approx 7000). I also ran the CSS video stress test on a cold boot and got 24 FPS (compared to 70 for my 9860).
Here is the interesting part - on a hunch - I decided to run back to back benchmarks (no time in between, aside from recording benchmarks). I ran the CSS video stress test on reboot (max settings) - got 24 FPS. I then re-ran it 8 (EIGHT) more times back to back. EACH time I consistently got 87-88 FPS (for those unfamiliar with that benchmark - those are fantastic numbers!!), and the GPU fans sounded like they were blasting out max RPM!! I then immediately ran 3DMARK05 and got a 5100+ score!! Ran it 2 more times and got similar scores! I then let it cool down a bit (until the GPU fans ramped down). Ran 3dmark05 again - this is the REALLY interesting part - the FPS on the proxycon during the first few seconds was 8-10 FPS (indicating a likely low score) but a few more seconds into it, I heard the fans start to ramp up and then FPS IMMEDIATELY jumped to about 30! The final benchmark was in the mid 4 thousands. A benchmark immediately after that was back up to 5100+. I was able to repeat this several times (watching the FPS jump as the GPU fans ramped).
OK - so I decided to call tech support (pct) and got a few suggestions (which I addressed in my list at the beginning of this post) including to post here about it. I also spoke to Luke who was very helpful as always. He suggested that this may be an issue with throttling (as I was thinking - though it seemed to be working OPPOSITE to what I would expect). He felt that maybe on initial bootup the GPU was getting hot early on and throttling down due to high heat, which abated as the fans ramped. This seemed very reasonable to me...UNTIL I ran a few more experiments (below) just prior to this post.
Though the utility I used is not recommended (and will remain unnamed
) - I downloaded and ran a thermal monitoring utility. I did not use it to ALTER in any way, shape or form, the factory settings. I ONLY used it to monitor temperatures as I ran benchmarks. This is what I found as far as fan % versus temperature of the GPU (BTW - the way I monitored this was to run the Counterstrike Source video stress test windowed instead of full-screen so I could still see the temps. Here is what I found:
Temp - Fan%
<64 - 30%
64 - 40%
68 - 50%
73 - 60%
77 - 70%
80 - 80%
84 - 90%
88 - 100%
I heard the fans ramp up at each of the increments above. BTW - during this benchmark - the CPU NEVER rose above 65C and the System NEVER rose above 47. Even at the high temps - the video ran perfect! No artifacts or glitches of any sort.
Even in windowed format - the stress test at anything above 40% fan gave me over 80 FPS. Anything less gave me 20-30 FPS.
I then re-ran 3dmark05 (full screen) and again got identical results. As soon as I saw the 8-10 FPS, I hit ESC to exit the bench and checked temps - they were always under 63 C or so. When I ran it again and saw the "jump" of FPS from 8 to approx 30, I escaped out and read the temps at over 64C as well as heard the fans ramp.
SOOOOO - having said all this - has ANYONE else seen anything like this? This sounds like some sort of minimum threshold (i.e. cold throttling) in the bios of the card? I dont believe this is a high-temp throttling. Keep in mind that no matter HOW long I ran the benchmarks or how hot the GPU got, the benchmarks never went down! I guess this is a good thing since it means that during loads (heavy GPU usage) the GPU will perform well. I just wonder why it would bench lower when cooler.
Opinions?
What I found was that when running benchmarks (3dmark05, 3dmark03 and Counter Strike Video Stress test) I was getting VERY low scores (lower than my 9860) when the laptop was first booted and running cool. The specifics are below, but FIRST - before everyone suggests this, I have already checked / verified / ruled out the following:
1 - I am running on AC at all times (connector is not loose).
2 - I have the latest Nvidia driver - 76.50 - came installed.
3 - I have DirectX 9.0c.
4 - I have the latest GPU bios.
5 - I have the latest system bios.
6 - I do not have any programs running in the background aside from the bare minimums.
7 - I do not have any spyware / adware / viruses on my system.
8 - I have not done ANY registry hacks / overclocking.
9 - The fans are running fine (more detail on this below).
10 - Changing refresh rates makes no difference in results.
11 - I am running in full res (1920 x 1200)
12 - I do not have anti-aliasing / aniosotropic on
13 - I am using DEFAULT settings on 3DMARK05 (pro version).
Ok - so here is what happens. Lets start with 3dmark05 - the very first time I ran it, the laptop had been running for a bit and it was warm in the room (I dont know what temp it was since I didnt have any GPU temp monitoring utilities). I got a benchmark of 5070! GREAT! I let the laptop sit for an hour and ran some errands. Came back - ran it again, this time I got 3300??! WTH?! Nothing was changed! Now - after running it again several times - I was then able to "predict" when I would get a high (5000+) or low (<4000) benchmark by looking at the FPS during the first few seconds of Proxycon (first video benchmark scene). I would get 30+ FPS for the high bench and 8-10 FPS for the low bench scores. This was VERY repeatable.
In addition - I would always hear the fans "ramp up" during the higher benchmarks - always. Kinda seemed like the GPU was throttled down and lower temps. (more on this below).
As a verification - I also ran 3dmark03 and got low scores as well (approx 7000). I also ran the CSS video stress test on a cold boot and got 24 FPS (compared to 70 for my 9860).
Here is the interesting part - on a hunch - I decided to run back to back benchmarks (no time in between, aside from recording benchmarks). I ran the CSS video stress test on reboot (max settings) - got 24 FPS. I then re-ran it 8 (EIGHT) more times back to back. EACH time I consistently got 87-88 FPS (for those unfamiliar with that benchmark - those are fantastic numbers!!), and the GPU fans sounded like they were blasting out max RPM!! I then immediately ran 3DMARK05 and got a 5100+ score!! Ran it 2 more times and got similar scores! I then let it cool down a bit (until the GPU fans ramped down). Ran 3dmark05 again - this is the REALLY interesting part - the FPS on the proxycon during the first few seconds was 8-10 FPS (indicating a likely low score) but a few more seconds into it, I heard the fans start to ramp up and then FPS IMMEDIATELY jumped to about 30! The final benchmark was in the mid 4 thousands. A benchmark immediately after that was back up to 5100+. I was able to repeat this several times (watching the FPS jump as the GPU fans ramped).
OK - so I decided to call tech support (pct) and got a few suggestions (which I addressed in my list at the beginning of this post) including to post here about it. I also spoke to Luke who was very helpful as always. He suggested that this may be an issue with throttling (as I was thinking - though it seemed to be working OPPOSITE to what I would expect). He felt that maybe on initial bootup the GPU was getting hot early on and throttling down due to high heat, which abated as the fans ramped. This seemed very reasonable to me...UNTIL I ran a few more experiments (below) just prior to this post.
Though the utility I used is not recommended (and will remain unnamed
) - I downloaded and ran a thermal monitoring utility. I did not use it to ALTER in any way, shape or form, the factory settings. I ONLY used it to monitor temperatures as I ran benchmarks. This is what I found as far as fan % versus temperature of the GPU (BTW - the way I monitored this was to run the Counterstrike Source video stress test windowed instead of full-screen so I could still see the temps. Here is what I found:Temp - Fan%
<64 - 30%
64 - 40%
68 - 50%
73 - 60%
77 - 70%
80 - 80%
84 - 90%
88 - 100%
I heard the fans ramp up at each of the increments above. BTW - during this benchmark - the CPU NEVER rose above 65C and the System NEVER rose above 47. Even at the high temps - the video ran perfect! No artifacts or glitches of any sort.
Even in windowed format - the stress test at anything above 40% fan gave me over 80 FPS. Anything less gave me 20-30 FPS.
I then re-ran 3dmark05 (full screen) and again got identical results. As soon as I saw the 8-10 FPS, I hit ESC to exit the bench and checked temps - they were always under 63 C or so. When I ran it again and saw the "jump" of FPS from 8 to approx 30, I escaped out and read the temps at over 64C as well as heard the fans ramp.
SOOOOO - having said all this - has ANYONE else seen anything like this? This sounds like some sort of minimum threshold (i.e. cold throttling) in the bios of the card? I dont believe this is a high-temp throttling. Keep in mind that no matter HOW long I ran the benchmarks or how hot the GPU got, the benchmarks never went down! I guess this is a good thing since it means that during loads (heavy GPU usage) the GPU will perform well. I just wonder why it would bench lower when cooler.
Opinions?
