Video memory tester?

[Poci]

I just unlocked the shaders on my 6950 and am OCing now. OCCT along with a GPU core tester also had a video memory tester, but in latest beta seems to not be there anymore.

What are you guys using for testing video memory?

[BeepBeep2]

I just unlocked the shaders on my 6950 and am OCing now. OCCT along with a GPU core tester also had a video memory tester, but in latest beta seems to not be there anymore.

What are you guys using for testing video memory?

Facepalm. Just run furmark, or don't test it. Game on it for a while. Unless you are bitcoin mining or running F@H GPU I doubt you will need the stability testing.
GDDR5 is self-error correcting, so it skips clocks as a fail-safe. You end up with lower performance. Once it can't correct you will start to get artifacts on screen, possibly a lock up with grey screen. I actually usually run 3DMark06 to test this.

[sepol]

for testing video memory I use GPU Tool. For GPU clock, OCCT and Crysis 2 with DX 11 patch.

[Poci]

Facepalm. Just run furmark, or don't test it. Game on it for a while. Unless you are bitcoin mining or running F@H GPU I doubt you will need the stability testing.
GDDR5 is self-error correcting, so it skips clocks as a fail-safe. You end up with lower performance. Once it can't correct you will start to get artifacts on screen, possibly a lock up with grey screen. I actually usually run 3DMark06 to test this.

Facepalm eh?

Furmark? Nah, I found it didnt catch very many errors and is only good for giving you a good indication of how well your cooling is doing since it does heat up GPU a lot.
Anyways I think I have found the perfect tool. MemtestCL (by the F@H team) , this thing even detected errors on GPU that OCCT GPU test couldnt find. You can set it to also test video memory by a command line to use all avaiable video memory. I find leaving MemtestCL overnight much more reliable. It found errors even after long sessions of different game engines that seemed stable.

And btw, can you plz show a little more respect, and stop assuming I just got a computer yesterday.....
I have been OCing for a loooong time, I think I would know by now what artifacts and hard locks are......
In any case both programs furmark and 3dmark 06 I find let a lot of stuff slide. Heck even valves source engine is a better stability tester then those 2.
As of right now like I said, for video card im finding memtestcl the best. It may be new, but thank god the folding creators made this, its even better then OCCT.

[BeepBeep2]

Facepalm eh?

Furmark? Nah, I found it didnt catch very many errors and is only good for giving you a good indication of how well your cooling is doing since it does heat up GPU a lot.
Anyways I think I have found the perfect tool. MemtestCL (by the F@H team) , this thing even detected errors on GPU that OCCT GPU test couldnt find. You can set it to also test video memory by a command line to use all avaiable video memory. I find leaving MemtestCL overnight much more reliable. It found errors even after long sessions of different game engines that seemed stable.

And btw, can you plz show a little more respect, and stop assuming I just got a computer yesterday.....
I have been OCing for a loooong time, I think I would know by now what artifacts and hard locks are......
In any case both programs furmark and 3dmark 06 I find let a lot of stuff slide. Heck even valves source engine is a better stability tester then those 2.
As of right now like I said, for video card im finding memtestcl the best. It may be new, but thank god the folding creators made this, its even better then OCCT.

I'm pretty sure you didn't know that GDDR5 is error correcting, therefore stable speeds may be slower than a theoretically slower speed.

I run 3DMark06 to see the perf. difference, not to see artifacts like you think I am.

[Poci]

I'm pretty sure you didn't know that GDDR5 is error correcting, therefore stable speeds may be slower than a theoretically slower speed.

Dont be so sure then, cause I already knew that....... like I said, think before you go on assuming, just cause I finally been able to post here at xtreme doesnt mean im a n00b. And sorry to burst your ego bubble that your not the only enthusiast on this planet....

lol u use furmark for "stability testing" and your gonna question my knowledge of GDDR5 memory?

I run 3DMark06 to see the perf. difference, not to see artifacts like you think I am.

All I said was that 06 is NOT a reliable stability tester. I have seen unstable systems pass 3d marks with no problem at all.

[BeepBeep2]

Dont be so sure then, cause I already knew that....... like I said, think before you go on assuming, just cause I finally been able to post here at xtreme doesnt mean im a n00b. And sorry to burst your ego bubble that your not the only enthusiast on this planet....

lol u use furmark for "stability testing" and your gonna question my knowledge of GDDR5 memory?



All I said was that 06 is NOT a reliable stability tester. I have seen unstable systems pass 3d marks with no problem at all.

Of course unstable systems can pass 3DMark06 with no problems at all.
Error correction causes a dip in the score, if you are able to lower memory clock and your score increases then obviously you are going in the right direction.

I never called you a "n00b", I assumed due to the nature of your post. I never assumed you got a computer yesterday, however you did post in a DDR3 overclocking thread being amazed by the fact that someone could run 1T on AMD and posted in a thread asking if 5970's were still good saying "5970's all the way" after the OP chimed in to tell everyone that he went with 69x0 series already.

You harassed me in PM.

[Poci]

Of course unstable systems can pass 3DMark06 with no problems at all.
Error correction causes a dip in the score, if you are able to lower memory clock and your score increases then obviously you are going in the right direction.

I never called you a "n00b", I assumed due to the nature of your post. I never assumed you got a computer yesterday, however you did post in a DDR3 overclocking thread being amazed by the fact that someone could run 1T on AMD and posted in a thread asking if 5970's were still good saying "5970's all the way" after the OP chimed in to tell everyone that he went with 69x0 series already.

You harassed me in PM.

As for the PM its to get you to quit "harrasing" me....

As for the DDr3 thread, are you going to deny that finding RATED 1T command rate modules is pretty hard?
As for the 5970s uhh, the guys if I remember correctly was going to put two of these into system vs two 6970s, and the 5970s are dual gpus on EACH card. Only reason someone recommended to not get the 5970s is due to heat, noise and power consumption. And since you love to bench so much wanna show benchies of dual 5970s vs dual 6970s? Or do I have to post them? Two dual gpu 5970s perform better then dual single gpu 6970s......

[BeepBeep2]

As for the PM its to get you to quit "harrasing" me....

As for the DDr3 thread, are you going to deny that finding RATED 1T command rate modules is pretty hard?

I apologize for the behavior in this thread. However calling my childish is merely flamebait.

As for the DDR3 thead, 98% of performance DDR3 does 1T at rated timings, I can pull DDR3-2200 1T bandwidth numbers out if you'd like. However you are changing subject from being amazed that someone can post results at 1T vs 2T, as you commented on someone doing ~1850-1900 7-7-7 at 1T which is hardly impressive at all.

As for the 5970s uhh, the guys if I remember correctly was going to put two of these into system vs two 6970s, and the 5970s are dual gpus on EACH card. Only reason someone recommended to not get the 5970s is due to heat, noise and power consumption. And since you love to bench so much wanna show benchies of dual 5970s vs dual 6970s? Or do I have to post them? Two dual gpu 5970s perform better then dual single gpu 6970s......

You obviously did not read the thread, the OP posted that he already bought 6970's, you entered to state "5970's all the way" ...doesn't make much sense does it?

You also obviously own two 6970 and a single 5970 so lets see those numbers. I love to bench, but I don't have the highest end hardware. Sorry I'm not rich, though I do my fair share of reading and research.

[Poci]

I apologize for the behavior in this thread. However calling my childish is merely flamebait.

As for the DDR3 thead, 98% of performance DDR3 does 1T at rated timings, I can pull DDR3-2200 1T bandwidth numbers out if you'd like. However you are changing subject from being amazed that someone can post results at 1T vs 2T, as you commented on someone doing ~1850-1900 7-7-7 at 1T which is hardly impressive at all.


You obviously did not read the thread, the OP posted that he already bought 6970's, you entered to state "5970's all the way" ...doesn't make much sense does it?

You also obviously own two 6970 and a single 5970 so lets see those numbers. I love to bench, but I don't have the highest end hardware. Sorry I'm not rich, though I do my fair share of reading and research.

As for the DDR3 im not talking about if you can manage to get 2T rated memory running at 1T STABLE... Im talking about ratings. We used to get them now I dont see cmd rate ratings anywhere.

And as for the 6970s, yes I didnt realize he already had the 6970s. But as far as two of them vs two 5970s....




http://www.techpowerup.com/reviews/H...HD_6970/1.html

I rest my case... 5970 performs better in practically everything, not by a considerable amount but still

[BeepBeep2]

I really shouldn't post at 1:30 in the morning after two days lack of sleep...

Didn't realize it was two 5970 vs two 6970, of course the 5970 wins hands down. >.<

[Poci]

Two days no sleep??? Are you on speed or something dude?

[PatRaceTin]

3DMark Vantage

also good for unstable overclocking

[Poci]

Im just saying 3d mark is not good for stability testing. They are great for making sure your hardware is performing correctly and then once you find a stable OC comparing that to stock.

In fact im finding a higher percentage increase in points by OCing video with 3dmark11 then 06.
06 must we way to CPU sensitive.