[PCPER]Frame-rating, what it is, what is does, and why it's important.

[KainXS]

I kinda guess something was going on since I had my old HD 3870 setup(but still had 4890 setup too) but it did not seem too bad, maybe thats why ATI never prioritized it, does anyone feel the 2900XT had this problem.

I can't see the old X800's and X1900's having had this problem, I had a X1950 setup and it ran beautifully back then.

 

[cadaveca]

Can you speculate how long this issue has been going on? Is it all the way back to the X800 days?

Nope. I suppose anyone can now build this test setup, get the software, and test other configurations. Might need to test every card with every driver version.



Kind of a pain in the ass, really.

I kinda guess something was going on since I had my old HD 3870 setup(but still had 4890 setup too) but it did not seem too bad, maybe thats why ATI never prioritized it, does anyone feel the 2900XT had this problem.

I can't see the old X800's and X1900's having had this problem, I had a X1950 setup and it ran beautifully back then.

Yeah, I never knew how good 2900XT was, since I had used always in CFX, never single card.


Then one card died.


And it seemed better. I just assumed it was the bad card, until I got my replacement.

 

[PatoRodrigues]

Really, really nice.

Sites like Guru3D are already mentioning that they'll be changing their methods to measure frame latency.

FCAT seems like a great tool, good to know that nVidia actually dedicated time and resources for this.

Indeed, fascinating.

 

[cadaveca]

they'll be changing their methods to measure frame latency.

Useless. You need a dedicated system that captures what's sent over the cable to the monitor, anything inside the system is suspect for not showing actual performance. Frame latency matters, sure, but not in this specific context.

 

[PatoRodrigues]

Looking to contribute with this thread, i'll post another recent article about it...

http://www.tomshardware.com/reviews/graphics-card-benchmarking-frame-rate,3466.html

Not PCPER, but look... this is a serious subject, and any information is welcome. I guess.

 

[erocker]

So... Fix around July? Read something the other day about that...

 

[cadaveca]

So... Fix around July? Read something the other day about that...

Nope, that's outside mentioned when Crossfire will START to be fixed, not 100%. No way to tell when it'll be working right, could be earlier, even.

 

[Xzibit]

If only AMD had a forum.

They could pull an Nvidia and have hacker hack it. Wait 7 months. Minimize Micro-stutters and Throttling in drivers and then bring back the forums after.

Quick AMD make a forum

 

[DRDNA]

I guess my only real worry is the end result will actually end up lowering performance of already released chips, but resulting in truer better feeling (maybe).

 

[techtard]

I haven't used Crossfire in a long time, and ended up selling because it was too hot and loud to be worth it.

Good to see they admitted they have a problem and are fixing it.

If only AMD had a forum.

They could pull an Nvidia and have hacker hack it. Wait 7 months. Minimize Micro-stutters and Throttling in drivers and then bring back the forums after.

Quick AMD make a forum

They do have forums, they just don't use them!
I used to lurk in their forums years ago, but there never seemed tro be dev presence. Maybe they shut down.

 

[erocker]

If only AMD had a forum.

http://forums.amd.com/game/

 

[DRDNA]

http://forums.amd.com/game/

Just skimmed threw their forum and now I remember why I fell in love with TPU!

 

[Xzibit]

Found it.

Just makes it sound bad and makes it very suspecious. That someone working with Nvidia brought the problem up. They tried to write the software himself but couldnt and then goes to Nvidia for help to write it.

starts at 24:00 mark


It would be better to get an un-bias method to grabbing the screens all along the process. That way you know whats the issue in the pipeline and if either AMD or Nvidia are doing things different along the way for a more comprehensive analysis.

 

[cadaveca]

It would be better to get an un-bias method to grabbing the screens all along the process. That way you know whats the issue in the pipeline and if either AMD or Nvidia are doing things different along the way for a more comprehensive analysis.

Or, have another programmer verify the NVidia software, which AMD has done already, it seems, since they admitted the problem is real, and are working to fix it.


I mean, it's not a question of whether this is real or not. AMD said yes, it is.

 

[Xzibit]

Or, have another programmer verify the NVidia software, which AMD has done already, it seems, since they admitted the problem is real, and are working to fix it.


I mean, it's not a question of whether this is real or not. AMD said yes, it is.

Im not saying there isnt an issue.

I want to see everything from when it leaves the game engine up to the when its displayed.

Right now its only a CF/SLI

I'm thinking beyond that to comparing single gpu chips on not just windows platforms. There more players in mobile so it be interesting to disect the differences all along the pipeline.

 

[cadaveca]

Im not saying there isnt an issue.

I want to see everything from when it leaves the game engine up to the when its displayed.

Right now its only a CF/SLI

I'm thinking beyond that to comparing single gpu chips on not just windows platforms. There more players in mobile so it be interesting to disect the differences all along the pipeline.

Sure, I guess in that aspect, I understand, but I'm pretty sure that this is so basic, really, that there is little reason to be wary of the results.

I'd kind of like more info into what's going on, honestly, but the fact of the matter is that this is something for developers and AMD programmers to deal with, not the end user.


And the only really "new" thing about what's being done is that the video output over the cable is being captured, and HDCP isn't an issue.

 

[DRDNA]

And the only really "new" thing about what's being done is that the video output over the cable is being captured, and HDCP isn't an issue.

I really wonder too how they kept the handshake with HDCP

 

[PatoRodrigues]

Those are some really scary numbers to say the truth.

And i even got suspicious, because 680 SLI always seems to have absolutely ZERO drops or runts.

 

[Xzibit]

Those are some really scary numbers to say the truth.

And i even got suspicious, because 680 SLI always seems to have absolutely ZERO drops or runts.

Thats why Nvidia is buying/sending reviewers one of those setups. Its mentioned at the end of the Podcast. They see a PR opportunity why not take it.

 

[cadaveca]

Thats why Nvidia is buying/sending reviewers one of those setups.

I guess we could say that the real cause of the "issue" needs to be identified before you'll be happy, which is acceptable.


I mean, we could go full conspiracy here and say that perhaps the problem is due to Nnidia's influence at developers. Perhaps games have coding that in intentionally placed to make this a problem in the first place. Who knows.




And Nvidia is not sending me one of these setups. So it's not like they are just jumping on marketing opportunities here either, since I don't have a single Nvidia product in my house, at all. Nothing but AMD GPUs in MY reviews.

 

[PatoRodrigues]

Now i'm really, REALLY curious to see how the reference HD7990 performs.

I think reviewers will have some massive work to do in the next few weeks.

 

[cadaveca]

Now i'm really, REALLY curious to see how the reference HD7990 performs.

I think reviewers will have some massive work to do in the next few weeks.

Nope.



Just Ryan @ PCPER does. His new job is verifiying that drivers work properly(whatever "properly" means).





And as long as they are, every other reviewer can just keep on doing what they do.

 

[Xzibit]

I guess we could say that the real cause of the "issue" needs to be identified before you'll be happy, which is acceptable.


I mean, we could go full conspiracy here and say that perhaps the problem is due to Nnidia's influence at developers. Perhaps games have coding that in intentionally placed to make this a problem in the first place. Who knows.




And Nvidia is not sending me one of these setups. So it's not like they are just jumping on marketing opportunities here either, since I don't have a single Nvidia product in my house, at all. Nothing but AMD GPUs in MY reviews.

This doesnt effect me cause I dont Crossfire/SLI

And if I did I'd just turn on V-Sync in Crossfire and Adaptive V-Sync in SLI

Maybe games with Frame limiters would help, I dont know i'm just reading graphs.

One thing is for sure I would sleep well at night

 

[cadaveca]

Maybe games with Frame limiters would help, I dont know i'm just reading graphs.

This doesn't happen in every game though

So there's more than one issue here.


Naturally.

 

[LAN_deRf_HA]

AMD made it sound like all these issues exist because they didn't know they were there.