• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

ATITOOL & artifact scanning... alternatives?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

D749

New Member
Joined
Jan 7, 2008
I bought a PNY 8800 GTS G92 512mb card a little while back and soon after began playing around with overclocking it. I didn't have high hopes but was pleasantly surprised. Using the stock heatsink and fan (@ 50%) I was able to push the card quite far (~ +25% core/shader) with temps < 80C at load.

Of course I was concerned about stability and what not so I decided to test my settings with ATITOOL (the latest beta). A week or so later I've almost given up on ATITOOL. It's ability to detect artifacts seems completely inconsistent (have to love that slider). I'm just wondering if anyone else feels this way and what else is out there for artifact scanning.

I have no problems (lock ups, crashes, or visual distortion that I can see) running my card 24/7 at my overclocked settings with apps or games. I run 3DMark06, Crysis demo benchmark (med/high @ 1920x1200), Hellgate London (DX9, max settings @ 1920x1200), TF2 (max settings @ 19200x1200), etc. without ANY problems. However, according to ATITOOL one run it might tell me that I have an error and then one run report no errors but still show a single yellow pixel here and there (which seems to prompt people to cry instability).

Thanks.
 
Instability, like beauty, is in the eye of the beholder. Certain games/apps stress the card differently, and what can seem stable in game/app-X might cause all kinds of issues w/ game/app-Y. To me, what you're describing isn't stable. I hardly sit there and watch ATI Tool, though, so if a yellow dot pops up, but isn't detect, I've yet to see it. If the 2 timers are the same after any given amount of time then I trust it hasn't had any artifacts.

I like to see what each slider will do in turn. That way I have a better idea of where the problem is. 815/2000/1915 (which one is the memory & which is the shader?) is a possible, but highly unlikely stable OC on a stock GTS. I have mine at 837/2106/1107 (I do the memory at it's actual clock rate instead of the DDR rate to avoid confusion), but I have also volt-modded it to 1.35v; +0.2v above stock, and have the HR03GT aftermarket cooler installed. Stock w/ fan at 100% I can only get 771/1944/1062.

What do you mean by 1 run? ATI Tool 'Scan for Artifacts' will run forever if you let it, so are you referring to a certain period of time? I usually shoot for 15mins when doing the sliders one at a time, and then go for 25mins once I stress all 3 maxes I was able to obtain independently. To find the weakness in the memroy, 2 tests have been shown to be most effective. Run 8 loops of 'Battle of Proxycon' 3DMark03, and then do the same w/ 'Canyon Flight' 3DMark05. If you can make it through all that, then you'll be 99% certain you are stable. This seems to work very well for me.
 
Instability, like beauty, is in the eye of the beholder. Certain games/apps stress the card differently, and what can seem stable in game/app-X might cause all kinds of issues w/ game/app-Y. To me, what you're describing isn't stable. I hardly sit there and watch ATI Tool, though, so if a yellow dot pops up, but isn't detect, I've yet to see it. If the 2 timers are the same after any given amount of time then I trust it hasn't had any artifacts.

I like to see what each slider will do in turn. That way I have a better idea of where the problem is. 815/2000/1915 (which one is the memory & which is the shader?) is a possible, but highly unlikely stable OC on a stock GTS. I have mine at 837/2106/1107 (I do the memory at it's actual clock rate instead of the DDR rate to avoid confusion), but I have also volt-modded it to 1.35v; +0.2v above stock, and have the HR03GT aftermarket cooler installed. Stock w/ fan at 100% I can only get 771/1944/1062.

What do you mean by 1 run? ATI Tool 'Scan for Artifacts' will run forever if you let it, so are you referring to a certain period of time? I usually shoot for 15mins when doing the sliders one at a time, and then go for 25mins once I stress all 3 maxes I was able to obtain independently. To find the weakness in the memroy, 2 tests have been shown to be most effective. Run 8 loops of 'Battle of Proxycon' 3DMark03, and then do the same w/ 'Canyon Flight' 3DMark05. If you can make it through all that, then you'll be 99% certain you are stable. This seems to work very well for me.

Core: 815; Memory: 1000 (1000x2=2000); Shader: 1915.

From reading various thread in regard to ATITOOL and the display of a single yellow pixel the consensus seems to be that any artifact (e.g., the display of a single yellow pixel) implies instability regardless of whether ATITOOL reset/dings to notify you of the single pixel (which it typically does not). This is one reason I'm trying to get away from ATITOOL - inconsistency.

What I mean by one run is opening up ATITOOL and running it for a period of time before I close it. For example, running it one day I might get X results, later that day Y results and the next day Z results. Typically when I run an instance of ATITOOL I let the artifact scanner run for at least 10 minutes.

You mention that having stable gameplay with the games listed running at the settings listed is not indicative of a stable card. I ask you then what determines stability? ATITOOL? My point is how can one say a card is unstable due to ATITOOLs seemingly random nonsense but yet have no apparent instability when running, at least somewhat, taxing 3D games/benchmarks. I agree that a game on my system would run different on your system but I suppose my point is that if I can run all of those games at 1920x1200 w/ max options and get > 60 FPS and make it through 3DMark06 over and over again how can I still say my setup is unstable?

I've been wrong plenty of times before though. :beer: I'll have to look into the 3DMarks. I don't believe the demos I have allow you to specify specific tests.

FYI I'm no longer running a stock setup in terms of heatsink and fan. I'm now using custom RAM sinks (pic below) and a VF900. Under load my temps have dropped ~15-20C. :D

Thanks.
 

Attachments

  • IMG_1892.jpg
    IMG_1892.jpg
    194.1 KB · Views: 1,895
Last edited:
Like I said..in the eye of the beholder. A lot of guys on here post super high clocks, and scores, but they aren't even stable in games. To me that is kind of like cheating, but I guess if you're a bencher then that is what you live for. I'm starting to cheat a bit myself by upping my CPU freq for 3DMark06 runs.

Anyway, if your happy w/ your clocks, and your games play great then just enjoy them. I'm anal about my OC. I know when that card comes from the factory it should be able to run all the benchmarks w/o any issues, and I like to keep it that way. For me I think stability goes beyond just barely being stable enough to play games or run a bench; I want there to be a margin of error built in, and that's where my definition of stability comes from. Just a personal preference, though.

A little anecdote...my previous GT was running great and even passing ATI Tool w/ flying colors at 944MHz on the memory (stock was 900). Whenever I tried to play F.E.A.R., though I would get a memory avalanche & lock-up almost immediately...even at stock clocks. After some investigation and some help from Viper John I learned about the loops to run in 3DMark to find this weakness. After running the loops I discovered my memory wasn't stable over 885! F.E.A.R. was finally stable. An RMA fixed that problem. The next GT did 1026 on the memory.

You need the pro version to run loops in 3DMark.
 
You made those sinks D? Pretty trick :D

Don't trust ATI Tool, W1zard (the developer) has said for some time it's not working exactly as it should be. That means do all your stability testing using things you actually use.

Hey jason, you're right 'stability' is is in the eye of the beholder. For me I don't play games but I do like to bench, that means everything is on the bare edge. When it comes to my posting rig, it's basically running stock ;)
 
I will run ATItool just to make sure my OC doesn't crash the card. Or if i see some serious artifacting I'll back the OC down. Other than that, I will fire up 3dmark, and sit back and watch, and adjust the clocks down as necessary.

I had to push my memory clocks down to 924 on My Gt in order to play Crysis. But 999 passes 3dmark just fine. Liek it was sai dabove, some things will stress the card differently.
 
You made those sinks D? Pretty trick :D

Don't trust ATI Tool, W1zard (the developer) has said for some time it's not working exactly as it should be. That means do all your stability testing using things you actually use.

Hey jason, you're right 'stability' is is in the eye of the beholder. For me I don't play games but I do like to bench, that means everything is on the bare edge. When it comes to my posting rig, it's basically running stock ;)

Hehe. I didn't make those sinks. They're made by a guy named Ian on the XtremeSystems forums. Petra Shop sells them but they're hard to keep in stock. I had them paired with a Zalman VF900 but I'm not happy with the performance of the VF900 after running oZone3D fur at stock clocks - temps rose to 90C+ and crashing commenced. I never had this problem with my stock 8800 GTS G92.

My new definition of load is running oZone3D fur @ 1280x1024 MSAA0 stability mode windowed with the ATITOOL artifact scanner running (for heat generation, I could care less about the random pixel nonesence). :D
 
Back