New 460 GTX 768mb is stuttering. Need Help!

[|V|ez]

My old card was an EVGA 8800GT which never died on me and performed well for a couple of year.

I replaced it with this card:
http://www.newegg.com/Product/Produc...82E16814162058

Review of this card:
http://www.kitguru.net/components/graphic-cards/steve-ruxton/galaxy-gtx-460-768mb-review/

But when I play games I definately notice more stuttering during certain points as compared to my 8800GT. Source engine games seem to yield the worst stuttering, but Starcraft 2 (my most played game right now) is definately not running as smooth either. I think it might more accurately be called microstuttering according to some things I read around the internet. Google "460 GTX stutter" to see that I am not the only one having this issue across various hardware/OS setups--even fresh install of windows (no driver contamination).

My install process:
1. Uninstall Nvidia drivers
2. Uninstall 8800GT and install 460 GTX
3. Boot into safe mode and use driver sweeper to remove Nvidia remnants
4. Re-boot into windows and install 260.99 Nvidia drivers
5. Re-boot into windows

Since then I have used different drivers from Nvidia to see if I have improved results, including a set of modded Quadro drivers to install for the 460 GTX which I found in a forum that claimed reduced stutter. I found a program that measured the change in FPS (using FRAPS's framerate files) to give a stutter index % (the Source engine has this inherently called "variability" when doing a Timedemo).

Benchmarks (Driver Ver.): Avg. FPS---Min---Max---Variability (Stutter %)
(used max settings for both games)
Left 4 Dead 2
8800 GT (260.99)-----------56.36----34----77------6.85%
460 GTX (260.99)-----------71.92----24---113-----20.61%
460 GTX (259.81 Quadro)----68.71----21---113-----19.43%

Starcraft 2 (forgot to disable vsynch)
8800 GT (260.99)------------53.78----22----64-----4.91%
460 GTX (260.99)------------46.41---19---61-------21.05%
460 GTX (259.81 Quadro)-----46.70---26---61-------21.57%

Furmark Results: Score---Max Temp-Avg. FPS---Min---Max
8800 GT----------1490----87c(!)----25---------21---36
460 GTX (260.99)-3251----73c------54---------43---88

Aquamark3-Score---GPU Score---CPU-----Triangles---Avg. FPS
8800 GT----141k---31,695------16,134---48.1 mil---159.33
460 GTX-----93k---32,460-------6,594(?)-28.3 mil---93.92

The Starcraft 2 and Aquamark3 results are the worst for my 460 GTX. I am not sure what is going on. I don't know why I get a horrible CPU test in aquamark3 with the GTX 460--maybe its the PhysX. The 460 GTX generally runs cooler than my 8800 GT with its highest temperature during gaming reaching 72c that I have seen.

My specs:
Phenom II 550 BE X3, Asrock MA3770DE, 2x2GB Gskill ripjaws, Galaxy GTX 460 768mb, 850 Watt Xion PSU, Windows XP x64

Most likely problem is the OS, but I tested this in my brothers Pentium 8400 (?) core duo/ Windows 7 x 64 box and it stuttered less I think (subjective), but still didn't seem as smooth as my 8800 GT in L4D2 and Team Fortress 2.

Considering RMAing it, but I rather keep it if drivers or tweaking can fix this. Plus on an RMA I get less back unless it is defective and I don't think this is defective (not sure though).

 

[|V|ez]

Any love for the stuttering GTX 460? (bump!)

 

[toyota]

sorry I cant help with the stuttering but a Phenom 550 is an X2 cpu not an X3. and an E8400 is Core 2 duo not a Pentium. also there is no Physx in Aquamark3.

 

[Skurge]

What resolution is that btw?

 

[Seero]

Use GPU-Z can check the clock speed, idle and load. Use CPU-Z to check PCIe lines.

 

[Ben90]

That CPU score on Aquamark looks pretty fishy.

 

[nyker96]

I'm using a asus top 460/768, played quite a few games using 260.99 driver, not a hickup. What overclock is your CPU running at and what resolution are we talking about here?

 

[|V|ez]

Responses in order:
-It really is a Phenom II 555 BE with one extra core unlocked to become a X3 instead of X2 (I have run L4D2 with only 2 cores with similar results). Sorry, I am just so use to Intel's CPUs being called Pentiums.

-I have used GPU-Z while I was in windowed mode for L4D2 and a couple of times it seemed to be running smoother with about ~85%-90% load on GPU. At other times (always full screen and I would have to alt-tab to read GPU-Z) the load seemed to stagger more. Doesn't always have this same result I think.

-All game test were run in 1680x1050, yet still the 8800GT was more stable. The GTX 460 had a bettter fps avg. in L4D2 but was definately more jumpy--measured in variability which actually is a fps unit not a % (not sure how it is measured).

-That was my score in Aquamark3--I swear. It was nice to see how well my 8800gt pushed it, blazing fast. My GTX 460 has got to be bugged somehow (in OS compatibility, drivers, hardware(?) card bios not sure.) Sometimes I think maybe Nvidia sold some bad chips that should have failed QA. Furmark ran as expected though.

-No overclocks all stock speeds. 3.2 ghz for the originally mislabeled Phenom II 555 BE. The GTX 460 clocks are found on product page of posted url.

 

[bryanW1995]

what settings were you using in those games? were they going to a dx11 path b/c you had a dx11 card installed?

best way to truly test them is ensure identical settings, then test in your rig (if they weren't identical before), then test in another rig.

 

[mlc]

Normally I would think the drivers are the issue, but it sounds like you've tried multiple versions...
Then I would question the cooling on the card, make sure everything is mounted fine with the heatsink, and thermal paste is properly applied etc... Don't assume they did it right ..

If that fails.. then it could be an issue with the hard drive/fragmentation, etc, or the settings for the game being maximized for the older card? Try reinstalling one of the games, and see if that changes anything.....

Also.. make sure no background tasks are running that could disrupt the gameplay.. (e.g. antivirus, spyware detectors, etc...) just to rule that out....

Good Luck ... Mike

 

[zagood]

Go back to 2 cores on the CPU.

 

[Seero]

Responses in order:
-It really is a Phenom II 555 BE with one extra core unlocked to become a X3 instead of X2 (I have run L4D2 with only 2 cores with similar results). Sorry, I am just so use to Intel's CPUs being called Pentiums.

-I have used GPU-Z while I was in windowed mode for L4D2 and a couple of times it seemed to be running smoother with about ~85%-90% load on GPU. At other times (always full screen and I would have to alt-tab to read GPU-Z) the load seemed to stagger more. Doesn't always have this same result I think.

-All game test were run in 1680x1050, yet still the 8800GT was more stable. The GTX 460 had a bettter fps avg. in L4D2 but was definately more jumpy--measured in variability which actually is a fps unit not a % (not sure how it is measured).

-That was my score in Aquamark3--I swear. It was nice to see how well my 8800gt pushed it, blazing fast. My GTX 460 has got to be bugged somehow (in OS compatibility, drivers, hardware(?) card bios not sure.) Sometimes I think maybe Nvidia sold some bad chips that should have failed QA. Furmark ran as expected though.

-No overclocks all stock speeds. 3.2 ghz for the originally mislabeled Phenom II 555 BE. The GTX 460 clocks are found on product page of posted url.

I will like to know the clock speed of the video card. GPU-Z tells you was clock/mem speed it is running at.

Also, Ben90 has a point, CPU score shouldn't drop like that after swapping GPU. Check with CPU-Z on the FSB, clock, memory speed too.

 

[nenforcer]

It looks like you've got quite a few variables but like you said I'm convinced it may be the DirectX9 OS your running.

I don't know why you upgraded to a DirectX11 card when you can't even take advantage of it even more so when your games are all still DirectX 9 based.

My guess is that the Windows XP x64 drivers for the GTX 460 cards are just not as heavily optimized as the Windows Vista / Windows 7 x64 code base it is intended to replace.

 

[taltamir]

My install process:
1. Uninstall Nvidia drivers
2. Uninstall 8800GT and install 460 GTX
3. Boot into safe mode and use driver sweeper to remove Nvidia remnants
4. Re-boot into windows and install 260.99 Nvidia drivers
5. Re-boot into windows

with current nvidia drivers a clean install is as simple as running the installer and checking the box that says to do a clean install.

In regards to micro stutter, it has to do with time to render of individual frames.
if a frame takes 5ms to render (5/1000 seconds) then it is running at 200fps (1000/5), if it takes 60ms to render then you are running at 16.7FPS. FPS is typically calculated on a per second basis (aka, count how many frames were rendered that second, rather then calculating the FPS from the ms taken to render each frame), thus at 22fps means that for that second 22 frames total were rendered and the average frame took 45ms to render.
The "min fps" is the minimum value of that "averaged over a second", which means it completely ignores micro-stutter. (if you have a bunch of rames take 5ms then one take 60ms you will notice it in the form of micro stutter, your FPS counter will not)
The "variability" figure in fraps is new, I need to look it up to see exactly what it measures.

People have raised a lot of good questions already, so check their suggestions, one suggestion i hadn't seen yet is... Is it possible that the GTX460 requires more CPU power and that you are CPU limited in those games? have you tried ocing your CPU to see if things get better?
It is a bit of a stretch, but maybe the GTX460 is simply expected to have a much beefier CPU (hence why it was better in your brother's computer, his CPU is better)

 

[|V|ez]

It was my red headed step child of an OS. I installed 32bit XP on a different partition and ran some games and things ran smoothly. The drivers for the GTX 460 just don't perform well with XP x64--no wonder because I might be the last person on Earth who still uses it.

Nenforce and I are the winners because we figured it out independently. And I get to keep my new card and it doesn't suck.

My Windows 7 Ultimate is now on its way to my house. MS owns me again

The "variability" measurement was from Source engine timedemo, not from FRAPS. But I found a little program that uses FRAPS framerate files to measure the same thing, I am guessing, because its author said it measured microstutter and its values closely reflected the "variablity" from Source. Thanks for the info on microstutter Taltamir I understand it now.
I think:
Variability= avg of differences of slowest frame vs fastest frame for each second

Here are some peeps that had same problem in Aquamark3 with different hardware:
http://www.overclockers.com/forums/showthread.php?t=606894
It was a driver issue. Aquamark3 still gives me a stupid result in 32bit XP with the GTX 460 btw.

 

[Ben90]

Ahh, that makes sense. XP64 had the opportunity to be such a great OS, but it has a lot of driver problems. I used it until the switch to 7 and it seemed just about every piece of my hardware needed an official hotfix from microsoft.

 

[nenforcer]

Yeah people talk about how Vista was bad with drivers initially but getting Windows XP X64 to fully work on a system in 2005 / 2006 was really a chore!

I think you will be really impressed with Windows 7 X64 and that card! I'm jealous!