Asking Ryan Smith of AT if a special examination could be done?

[Keysplayr]

Which "accredited" review site did you get this information from?

I made it all up.

 

[BrightCandle]

I can't be the only person that interprets that diplomatic not now but it is coming soon as a flat no. I also consider the rejection of frame times as unreliable as a dig at the tech report and its methods. Seems inconsistent with Anand's comments. I just didn't see a yes in Ryan's comments I saw a diplomatic no and a rejection of the very premise that they might have missed this in their own review.

Maybe they will prove me wrong and release something with tooling no one else has used and give us something no other site has. But I am also still waiting on an email back about an investigation into this very issue from 5 months ago.

Pretty soon it won't matter, the next crop of cards will be out and we will see fps reported everywhere again and likely more microstutter discussions. I will be looking at the graphs and wondering what problems they didn't notice that I will notice.

Those 7970s were a major buyers remorse for me, the 680s were merely adequate (in the sense they worked but still don't let me play at the settings I want). I don't want to play the useless performance measure game anymore.

 

[Hitman928]

I must be the only person that interprets that diplomatic not now but it is coming soon

The way I read it was,

We're not going to investigate it until we feel like we have the proper tools and proper methodology to do it. We're hoping this happens soon, it could not happen for a while, it may never happen.

Kyle from [H] essentially said, we've looked into it, the tools aren't there, we're not going to try to do this for the foreseeable future.

[TR] said, the future is now

That was how I read it anyway.

 

[Rikard]

What sweclockers said in yesterdays podcast was something along the line that "We studied microstuttering on a lot of cards, we saw some small effect but nothing as large as TR are claiming. But then we did not publish the article since the new drivers came out that completely change our conclusions, and we did not have time to redo it all again. But it is coming sometime during the Spring."
(The quote is just me typing from memory, not word by word correct.)
So they are sitting on information on lots of cards apparently, and information that something changed when changing drivers, but what changed, and what drivers???

 

[thilanliyan]

The way I read it was,

We're not going to investigate it until we feel like we have the proper tools and proper methodology to do it. We're hoping this happens soon, it could not happen for a while, it may never happen.

Kyle from [H] essentially said, we've looked into it, the tools aren't there, we're not going to try to do this for the foreseeable future.

[TR] said, the future is now

That was how I read it anyway.

That is how I interpreted Ryan's comments as well. No point in a scientific investigation if they don't think they have the tools to do it properly.

 

[SirPauly]

Personally did take Ryan's and Anand's comments as they did write them. Seem pretty clear to me.

Nice to see some potential improvements with smoothness in FarCry3:

Following 310.70 WHQL's creation we updated our Far Cry 3 SLI profile to incorporate new 'bits' that prevent frame rate stuttering observed on some systems with the previous profile. To force an immediate download of the new profile, install 310.70 WHQL, then right click the NVIDIA tray icon and select "Check for updates." If you have GeForce Experience installed, head to the 'Updates' section of the 'Preferences' tab and click 'Check Now', ensuring the 'Other Software' checkboxes are ticked.

http://www.geforce.com/whats-new/articles/nvidia-geforce-310-70-whql-drivers-released

 

[f1sherman]

What sweclockers said in yesterdays podcast was something along the line that "We studied microstuttering on a lot of cards, we saw some small effect but nothing as large as TR are claiming.

TR is not claiming anything. They are measuring.

Sweclockers are claiming that they studied MS... which they won't publish... ehh

 

[Keysplayr]

Which "accredited" review site did you get this information from?

I made it all up.
Seriously though ive already stated in that post that all these questions are answered in that podcast.

 

[Rikard]

which they won't publish... ehh

Patience, my young padawan

 

[KingFatty]

The overwhelming majority of review sites must concur.

Building a consensus is one way to show results.

Another way is to have a methodology that gives you reproducible results.

I think any reasonable person would be justified in questioning something that lacks a consensus (e.g., a from a single review site) and lacks the ability to reproduce the results (e.g., getting different results each time you run the so-called 'test' and/or relying on eye-ball subjective impressions; a proper test provides predictable and reproducible results that can be demonstrated).

 

[BrightCandle]

Frame times might not be reliable, but fps measures are just as bad, potentially worse. What remains is that none of the reviews we "trust" can review card performance objectively, they are not worth the pixels they are shown on. With nothing but flawed performance measures we may as well just roll a dice and pick that way. 60 fps doesn't mean anything useful in terms of motion unless the frames come evenly.

So far all reviews assumed they were smooth, but we all know they are not(high speed cameras confirm this). What remains is the only useful review is the subjective one....and that sucks as we know its bad science.

 

[SirPauly]

Imho,

Subjective may be based on different tolerance levels; where a gaming example may be smooth for one may not be for another.

Subjective makes some sense if there is at least some data to back it.

 

[KCfromNC]

I made it all up.
Seriously though ive already stated in that post that all these questions are answered in that podcast.

That's awfully vague. At what point in the recording do they discuss their results with drivers prior to 12.8?

 

[Keysplayr]

That's awfully vague. At what point in the recording do they discuss their results with drivers prior to 12.8?

Give it a listen. I never said prior to 12.8. I said 12.8. Unless your taking another comment of mine out of context. Just let me know. You can start at about 38 minutes in. Another thing. You need to stop the baiting. Like now.

 

[KingFatty]

Well, calling it baiting could be a bit hard on him. There was a claim that all questions were answered. Then he said it's vague, presumably questioning the "all" claim. It's hard to say it's taking something out of context, when the original claim was very specific to answering "all" questions. So I wouldn't say he was baiting, at least it doesn't seem like he was intentionally baiting, because he was just holding up an example that disproved the rule, which I think should be encouraged as fair game.

 

[Keysplayr]

Well, calling it baiting could be a bit hard on him. There was a claim that all questions were answered. Then he said it's vague, presumably questioning the "all" claim. It's hard to say it's taking something out of context, when the original claim was very specific to answering "all" questions. So I wouldn't say he was baiting, at least it doesn't seem like he was intentionally baiting, because he was just holding up an example that disproved the rule, which I think should be encouraged as fair game.

Yeah it would be nice if he just listened to the podcast. I mean, if he puts as much effort into that as he does posting spiteful comments and little digs, he'd be alright. And I see you'll hang on every word. I'm referring to the "all". Last time I'm saying it. Listen to the Podcast. At least from 38 minutes on.
We good? Great!

 

[KCfromNC]

Give it a listen. I never said prior to 12.8. I said 12.8. Unless your taking another comment of mine out of context. Just let me know. You can start at about 38 minutes in.

AMD first started the fps campaign with possibly the 12.8s, introducing higher framerates over previous drivers but also introducing more severe hitching and stuttering/latency.

You're speculating that 12.8s introduced higher latency compared to earlier drivers, which implies you have data from a driver before that to compare to. If you're admitting you don't have that data, it's cool, just come clean and let us know that it's just speculation on your part.

And no, I'm not going to listen to 45 minutes of talking to do your work for you. I don't think that asking for something a bit more specific is out of line.

Another thing. You need to stop the baiting. Like now.

Yawn.

 

[blastingcap]

Frame times might not be reliable, but fps measures are just as bad, potentially worse. What remains is that none of the reviews we "trust" can review card performance objectively, they are not worth the pixels they are shown on. With nothing but flawed performance measures we may as well just roll a dice and pick that way. 60 fps doesn't mean anything useful in terms of motion unless the frames come evenly.

So far all reviews assumed they were smooth, but we all know they are not(high speed cameras confirm this). What remains is the only useful review is the subjective one....and that sucks as we know its bad science.

This. Especially the bolded parts. I think we need to have high-speed cameras recording more than one game at one setting. Make it many games at many settings. From all that data, the truth will emerge.

 

[VulgarDisplay]

This. Especially the bolded parts. I think we need to have high-speed cameras recording more than one game at one setting. Make it many games at many settings. From all that data, the truth will emerge.

I think they need to use 120hz monitors as well, or at least compare a 120hz monitor to a 60hz monitor. Maybe even break out a CRT.

 

[SirPauly]

I think frame-times hold some relevance based on one can discern this is a major contributor to the differentiation between AFR and single GPU smoothness; One can also discern that three or four GPU's offer smoother gaming than two; One can discern the frame metering with Sli -- latency may play a hand in smoothness, one may imagine.

 

[Keysplayr]

You're speculating that 12.8s introduced higher latency compared to earlier drivers, which implies you have data from a driver before that to compare to. If you're admitting you don't have that data, it's cool, just come clean and let us know that it's just speculation on your part.

And no, I'm not going to listen to 45 minutes of talking to do your work for you. I don't think that asking for something a bit more specific is out of line.



Yawn.

So you're saying AMD drivers always had this higher latency and was not introduced with the introduction of the 12.8 drivers and my speculation was wrong? Well, I've been wrong before. Thanks for clearing that up.
And if you won't put in the effort to listen to that podcast, I know you're not truly interested in this topic. I think you're more interested in me.

 

[SirPauly]

If one goes back to the first GTX 660ti, where there was no apparent standing out by Techreport with the HD 7950 from a latency point-of-view -- they were using 12.7 beta.

Methods:

http://techreport.com/review/23419/nvidia-geforce-gtx-660-ti-graphics-card-reviewed/3


99th percentile:

http://techreport.com/review/23419/nvidia-geforce-gtx-660-ti-graphics-card-reviewed/11

This is why the latest testing was surprising for some and may be worthy of additional investigations.

 

[bononos]

Give it a listen. I never said prior to 12.8. I said 12.8. Unless your taking another comment of mine out of context. Just let me know. You can start at about 38 minutes in. Another thing. You need to stop the baiting. Like now.

I listened to the podcast from the 38 min mark and there didn't seem to be any mention of 12.8. All that was mentioned was 12.11 beta drivers.

 

[Keysplayr]

I listened to the podcast from the 38 min mark and there didn't seem to be any mention of 12.8. All that was mentioned was 12.11 beta drivers.

Nope. He mentions 12.8s are slower than 12.11s but still exhibited the same latency problems.

 

[Final8ty]

WHOOHOO!!!! I have just finished a set of benchmarks. I'll draw a conclusion at the end.
I am comparing two cards: Sapphire HD 7870 OC vs. eVGA GTX 660 Ti FTW Signature 2. Drivers used were the Cat 12.11 Beta11 for the HD 7870 and the newly released Forceware 310.70 WHQL drivers.
HD 7870: http://www.newegg.com/Product/Produc...82E16814102983
GTX 660 Ti: http://www.newegg.com/Product/Produc...82E16814130837
For comparison, I paid $209.99 for the HD 7870 on newegg, while I paid a whopping $299.99 for the GTX 660 Ti. Granted, the 7870 was on sale and marked down about $40, but that's what it cost to get it.

Now for the benchmarks. I ran both cards at stock for every test. The resolution was exactly the same on all the games (1920x1200), and the details were left unchanged so that there could be an accurate comparison. Straight off the line, I was pretty sure that the 660 Ti was going to be faster on framerates, but lets see what happened.
The games I benchmarked were Battlefield 3 (multiplayer), Crysis, Far Cry 3, and Skyrim. Each benchmark was run for 2 minutes using FRAPS, and the resulting file was put into the FRAPS bench viewer utility.
Let's get started with Battlefield 3:
NOTE: All pictures submitted are done with the 660 Ti as the top graph, and the 7870 on the bottom. This can be proven because the time stamps on the files. The 7870 benchmarks were done first, so the time stamps on those files are earlier (by about 45 minutes or so) than the 660 Ti benchmarks.

The benchmark was done on a 64-man Caspian Border server which was full for both runs. I ran around the map, shooting and taking points as if I was playing a normal game.
You can draw your own conclusion, but both cards were set to the same detail settings (all Ultra with no MSAA), and the results were not even close. The 7870 wins this one by a landslide. The 7870 was smoother and much more enjoyable. There was a constant frame stutter exhibited by the 660 Ti while the 7870 exhibited very smooth gameplay. This REALLY surprised me, but the graphs back up exactly what I felt in game.
Before anyone goes crying foul, I have to believe that this is either a driver issue or possibly an issue with my CPU. However, the 7870 ran just fine on my Phenom II X4, and the more powerful 660 ti didn't, so I don't believe my CPU is the culprit here. Software is more than likely to blame.
BF3 Winner: HD 7870
Next, let's look at Crysis:

Everything was set to very high with no aa on for this test. I started from the beginning of mission 2 on Crysis (not Warhead). I ran down to the small outpost, ran around the building two times, ran down the road, cloaked and shot the passing humvee in the gas tank (big explosion) then ran into the forest until the end.
This one was a little tougher, but again the 7870 pulled ahead. The average framerate was exactly the same, but the 660 Ti had a lower minimum framerate. Also, you can see by the graph that the 660 ti had quite a few more frame spikes throughout the run. I also noticed several areas where the 660 ti's framerate dipped that did not occur on the 7870. This means that the 7870's framerate and gameplay experience felt more consistent, and thus more enjoyable.
This could be a CPU issue, but I doubt it. Crysis only uses 2 threads, and even then, it ran perfectly on a stock Core 2 Duo. This probably has to do with the Kepler architecture and the driver as well.
Crysis Winner: HD 7870
Next let's look at Far Cry 3:

The settings on this were very high, HDAO, and no MSAA. I started from the same point on the island and ran for 2 minutes. Unfortunately, this was a little more difficult because the baddies spawned differently during both runs, so I couldn't even get close to reproducing the runs. However, I tried to stay in the same area for both.
This time things are quite a bit closer. First, the 660 Ti did maintain an average 5 fps higher than the 7870, and that could be felt (barely) in game. It did feel like the 660 ti maintained its framerate better than the HD 7870 this time, although based on the graphs, it's sort of hard to tell. The 660 ti has more frame spikes towards the end while the HD 7870 has more towards the beginning. However, since the 660 Ti maintained a higher framerate, it did provide a smoother gameplay experience, and thus the 660 Ti takes this one.
Far Cry 3 Winner: GTX 660 Ti
Finally, lets talk about Skyrim:

Settings were maxed except for MSAA, which was set to 4x for both cards. For this test, I spawned at a quick save point near White run and ran up the road for two minutes, both runs following the same path.
I'm not sure how to approach this one. The 660 Ti most definitely had a higher framerate for this game BUT it also had a more inconsistent one. It wasn't much more inconsistent, but when you turn off vsync, it's noticeable. There were a couple of times where the 7870 exhibited some bigger stutters, but they didn't last for that long. That could even be contributed to a background task on the computer. However, you will see on the 7870's graph (lower) thant at about 50 seconds, the frame times become almost a tight line, whereas the 660 Ti is unable to achieve that at any time.
Long story short, I'm confused. My gut is telling me that the 7870 was smoother for most of it, but the 660 ti's framerate was more consistent. Needless to say, if you turn on vsync on either card, that eliminates virtually every issue period. Because of this, I'm going to call it a draw. I simply do not have a preference in this instance.
Skyrim Winner: DRAW
Conclusion: Based on these 4 games, I have to give the nod to the 7870. The performance wasn't higher, but the framerates were more consistent. This lead to the gameplay experience being smoother overall. Since BF3 is my favorite game, I choose the 7870 without a shadow of a doubt. The fact that the 7870 is far cheaper than the 660 Ti makes this an easy decision for me... I choose the HD 7870.
If anyone else has requests for games, let me know. I'm tight on cash, but I'll do the best I can here. Let me know if you guys need any clarification on these tests or would like to see something done differently. Thanks. :thumb:

http://www.overclock.net/t/1337206/...d-7950-stumble-in-windows-8/360#post_18856460