HD 2900XT & Driver 8.38

Cyrilix said:
Caffeinated, I have done my best to address every comment that was reasonable, however you have simply pushed this too far.

500 is an implication, not a 100% correct clock speed. They did not break out their timer tools, loop a one-clock operation 500 million times, and time it with a highly accurate scientific tool. What they mean is "stock speed".

[sarcasm]
Well, guess what? ATi's card is supposed to be 748 MHz stock, but they only ran it at 740! What is this? Are they cheating us users of valid benchmarks or something?
[/sarcasm]

I am lucky I don't read too much into these things because just like the 8800 GTS, the implication is stock speed. I'm also amazed that you've managed to see this in the 8800 GTS, but not in the 2900 XT. Please tell me why that is. According to you, they would both be underclocked. The next thing you're going to tell me is...the 8800 GTS was underclocked 5 MHz more, right? and given the 500/513 difference vs. 740/748 difference, the relative difference is a larger, right? It's a conspiracy!

I'm waiting for these arguments from you, but I really really hope that you don't provide them.

Have fun with your speculation. You don't need to reply. All I need is for others to see this post so they can the incredulity of your own.
Click to expand...

Once again, for clarity's sake, I am pointing out that their numbers are suspicious and that the card was not set at the default speed.

Do you work for that site, and feel compelled to defend it? If so, then it is understandable that you might feel obliged to support their results. If not, then perhaps you might want to consider whether you are taking this too personally.

My only point with the results that I showed is that they are not right - because they are using a faster processor and the core is underclocked. Due to the fact that the numbers for the NVidia card are incorrect, their results for the 2900XT could also be incorrect. I can't say for the 2900XT - I don't own one, but I can make the statement regarding the 8800GTS which I do own.

I am calling the pcpop.com numbers and their testing methodology into question - not drawing conclusions based upon their results. If people are drawing conclusions based upon their results, then they also need to realize that those were obtained using dubious methods.
 
Alright, you have a certain point. As phide mentioned, overall, the scores seem to be OK, except for that 7.5% difference from your own on the SM3.0. I noticed that too, but refused to bring up a whole racket based on just that result. You will also notice that I have a strong strong distaste of people posting overall 3DMark06 scores, as part of the 06 score is a CPU component, which simply adds in another point of uncertainty. To me, 05 was the last of the great Futuremark benchmarks, since they didn't fuck around with non-GPU results.

It's also a shame that most sites only post full 3DMark06 scores, and not the components. I will try to see what other info I can get on the other SM2.0/SM3.0 scores of the 2900 XT and follow up on this post later.

Lastly, I do not work for that site, although I feel that there has been so much needless hate against this card, that I can't sit idly and simply watch as people post suspicions that to me, are unreasonable. Knowledge of facts is power.
 
FOLLOW-UP THREAD (8800 GTS 640MB 3DMark06 SM2.0/SM3.0 scores):

For this comparison, I do not look at driver versions, only that the review used the 8800 GTS 640MB card and ran the 3DMark06 benchmark. This assessment is to give a general performance evaluation of the 8800 GTS 640MB in SM2.0 and SM3.0 scores. The threshold for similarity will be +-200 points difference (approx. 5%), so 3678-4078 for SM2.0 and 3493-3893 for SM3.0.

PCPOP results:
SM2 - 3878, SM3 - 3693 (REFERENCE)

http://www.guru3d.com/article/Videocards/391/18/
SM2 - 3724, SM3 - 3727 (SIMILAR SM2.0 AND SM3.0)

http://www.legitreviews.com/article/413/14/
SM2 - 3646, SM3 - 3658 (LOWER SM2.0, SIMILAR SM3.0)

http://www.hothardware.com/articles...GTX_and_8800_GTS_Unified_Powerhouses/?page=11
SM2 - 3741, SM3 - 3763 (SIMILAR SM2.0 AND SM3.0)

These are all the results I could find in this time frame. It would seem to me like your results are probably higher, for what reason, I do not understand, but it certainly differs from all the initial 8800 GTS 640 benchmarks. Nevertheless, the PCPOP benchies are generally a tad bit higher than the initial 8800 GTS 640 benchies. For reference, PCPOP uses 158.19 on XP -- what version are you using, Caffeinated?
 
3dmark06-3-1.jpg
 
Here is some guy with a Core 2 at 2.93/GTS: Result

For convenience, here is his stuff compared to mine, but you can go there and see for yourself:

core2293gtssc4.jpg



All that I did was go to the ORB, and do a search and compare based upon a Core 2 @ speeds between 2.92 Ghz and 2.94 Ghz with a Geforce 8800 GTS. I'm sure there are results higher and lower since there is variance between boxes - I just picked the first one that came up for me.
 
For the sake of argument, here is another one pulled out of the ORB, this time a 2.93 Extreme/2900XT:

2900xtwg9.jpg
 
Cyrilix said:
Hi Caffeinated, why does his video card indicate...

256MB video memory? 108 MHz core clock? 364 MHz memory clock?

What is this?
Click to expand...

I don't know. Mine never lists the right speeds either. According to Futuremark's ORB mine is running at 312Mhz/108Mhz and my FSB is at 11Mhz according to the ORB. :p
 
Caffeinated said:
I don't know. Mine never lists the right speeds either. According to Futuremark's ORB mine is running at 312Mhz/108Mhz and my FSB is at 11Mhz according to the ORB. :p
Click to expand...


Hah, that's funny and all... but it doesn't help when I'm trying to validate your benchmark. :p
 
Cyrilix said:
Hah, that's funny and all... but it doesn't help when I'm trying to validate your benchmark. :p
Click to expand...

I have four other results, taken a few months ago when I first built the system if you want those, too. All of them are over 9000, with the highest being 9206. The one from yesterday was with the 158.22's, which is when I changed my drivers. As I had previously mentioned, I never would have thought anything of the scores if I hadn't just run 3DMark06 like an hour before seeing those. I ran a quick benchmark to make sure everything was stable and looked okay with the new drivers before I started playing a game. I didn't want to find out in the middle of having fun that something wasn't right. 3DMark is pretty good for quick stability testing.
 
3d06_10972.png


HD 2900XT at standard clock

Frame Rate............BenchMark
3650.................... Return to Proxyon
2541.................... Firefly Forest
3794.................... Canyon Flight
3910.................... Deep Freeze
Cat 8.37
 
Weather or not the drivers/scores are real will be known soon enough. I for one think that there is a lot of potential yet to be seen from the R600. Will we ever see it? Perhaps. On paper the 2900xt should be much faster than the GTS. But its not. That means you either have a hardware issue, software or combination of the two. The only thing synthetics are good at is to show us what the hardware potential is. And from looking at all of the synthetics you can see that hard ware does not look "broken". There are some areas where its faster and slower compared to the G80. So that leaves us with some areas on the drivers where we can make some improvement. Given the complexity of he new R600 design I have no doubt that AMD will make some major strides and maybe just maybe we can see some of the paper specs match what we see on games...
 
FORNOWAGAIN said:
A new review from Korean ATiMANIA

AMD Catalyst 8.38 vs NVIDIA ForceWare 158.18





Click to expand...

Heh... not only is the site called "Ati Mania" but its up against a 320mb card @ 1920x1200 and looses in one of the games... oouch!

And the 8800 series seems to have lots of problems with the test driver game, dunno if its driver or what though.
 
jmackay said:
Heh... not only is the site called "Ati Mania" but its up against a 320mb card @ 1920x1200 and looses in one of the games... oouch!
Click to expand...
With a top end/average FPS like that I'll gladly take a couple of frame hit from Max and Avg if I get a couple of extra frames on Min.

But yeah it is goofy having a comparison with the 320MB card (not familiar with those games, would they have been in any way RAM limited with 320MB?). I've seen them included in XT reviews but OC models and I can't think of anywhere where it was the sole 8800 card.
 
Dwight said:
With a top end/average FPS like that I'll gladly take a couple of frame hit from Max and Avg if I get a couple of extra frames on Min.

But yeah it is goofy having a comparison with the 320MB card (not familiar with those games, would they have been in any way RAM limited with 320MB?). I've seen them included in XT reviews but OC models and I can't think of anywhere where it was the sole 8800 card.
Click to expand...

W/o seeing a graph of their fps while gaming, we dunno how often they hit the min mark. And 56 fps min is still good (they don't show the 2900 actual fps for whatever reason).

But yes, 1920x1200 memory will be limited on the 320 model... not to mention its a whole lot cheaper than the 2900.
 
Impressive, we are seeing improvements with each new driver release. It's only a matter of time ;). Once the drivers mature we will see the full potential of the HD 2900Xt.
 
Better hope so..I don't understand how the drivers can be SO bad that the performance is as such and it sucks more power then a 8800GTX. :eek:
 
Eastcoasthandle said:
Impressive, we are seeing improvements with each new driver release. It's only a matter of time ;). Once the drivers mature we will see the full potential of the HD 2900Xt.
Click to expand...

Did you not see the images? Its not rendering correctly.

Of course the performance will be higher when the card doesn't render HDR and other special effects.
 
XSNiper said:
Better hope so..I don't understand how the drivers can be SO bad that the performance is as such and it sucks more power then a 8800GTX. :eek:
Click to expand...
Good question, but it's not the first time a scenario like this occurred in the video industry. Didn't another ATI card also suffer from poor performance initially until a subsequent driver release boosted it at least ~30%?
 
APOLLO said:
Good question, but it's not the first time a scenario like this occurred in the video industry. Didn't another ATI card also suffer from poor performance initially until a subsequent driver release boosted it at least ~30%?
Click to expand...

The 8500 suffered until ATI started cheating on image quality to give the illusion of a 30% performance boost, when in reality all the did was decrease IQ on a driver level and say "LOOK, WE FIXED IT!!!"
 
Sorry, thats false. It was a bug. Which they fixed in a driver less than a week later, with a performance increase.

The 8500 got dramatically faster with new drivers. Sadly, it took several months. Still it caught, and passed the GF3, even even the Ti series is most games. Then the GF4 came out, and blew it away.
 
8.38 RC3 already for download in several foruns ( not too much improvements)

8.38 RC7 was released yesterday ( not public)

8.38 final should be released tonight

RC= release candidate


dl is 78mb this is only the Catalyst Centre and the drivers . Performance so far seems ok havnt tried much yet, Lost Coast issues are still present, although the IQ seems to have improved and a little less blurry, bear in mind this driver set is from the 10th of May just under 2 weeks. So now I think about it these could be dated compared to the finals that should come out tonight , still these are nice to have while we wait.
Click to expand...


i Installed 8.38 RC3 , i will test them now and wait for Final release


regards
 
InorganicMatter said:
The 8500 suffered until ATI started cheating on image quality to give the illusion of a 30% performance boost, when in reality all the did was decrease IQ on a driver level and say "LOOK, WE FIXED IT!!!"
Click to expand...

Man, some people can't let go of that urban legend.
 
InorganicMatter said:
The 8500 suffered until ATI started cheating on image quality to give the illusion of a 30% performance boost, when in reality all the did was decrease IQ on a driver level and say "LOOK, WE FIXED IT!!!"
Click to expand...
Wow, here we go again with "hearsay". Do you have a source for your hearsay?
 
fallguy said:
Sorry, thats false. It was a bug. Which they fixed in a driver less than a week later, with a performance increase.

The 8500 got dramatically faster with new drivers. Sadly, it took several months. Still it caught, and passed the GF3, even even the Ti series is most games. Then the GF4 came out, and blew it away.
Click to expand...

i will back u up. that's exactly what I remembered happening. I had both gf3 ti500 and 8500, and the in the end, 8500 was faster in most games except OpenGL stuff
 
jmackay said:
If AA / HDR not working correctly on a $400 is "fine" sure...
Click to expand...

The question you have to answer here, jmackay, is whether or not the HDR worked fine in 8.37 and was broken in 8.38. If that is the case, then you truly have an argument on your hands. If not, then it's just a bug in a specific game that should be fixed in a driver update in the future.
 
Cyrilix said:
The question you have to answer here, jmackay, is whether or not the HDR worked fine in 8.37 and was broken in 8.38. If that is the case, then you truly have an argument on your hands. If not, then it's just a bug in a specific game that should be fixed in a driver update in the future.
Click to expand...

He said the image looked fine to him. And they will probably fix HDR, but you can't say that results from a benchmark using broken drivers which turn off effects like HDR proves that the 2900 is performing well for its price.
 
jmackay said:
He said the image looked fine to him. And they will probably fix HDR, but you can't say that results from a benchmark using broken drivers which turn off effects like HDR proves that the 2900 is performing well for its price.
Click to expand...

Yes, I know, but I'm trying to address the criticisms people are using of "cheating". If the effects are purposely disabled in order to increase performance, that's cheating. If the effects have never been there in the first place, then that's an incomplete/broken feature set, but can be resolved.
 
Why do people that can't even tell when something is rendered correctly or well want a high end video card? If you can't tell the difference between banding and no banding, AA or no AA, and HDR or no HDR, just get a 1600pro or something and turn details down to medium, AA off and AF off, and you'll be set. There is no sense buying high end hardware if you don't care about the image quality - that's like buying $1000 headphones and then plugging them into a $5 cassette player.
 
Cyrilix said:
Yes, I know, but I'm trying to address the criticisms people are using of "cheating". If the effects are purposely disabled in order to increase performance, that's cheating. If the effects have never been there in the first place, then that's an incomplete/broken feature set, but can be resolved.
Click to expand...

True, we don't know if its on purpose or not, but putting up benchmark numbers when one card isn't rendering correctly and calling it an "apples to apples" benchmark is very inaccurate. So while we don't know if its a bug or not, its not a valid benchmark for the card, as not having to render AA or HDR will obviously make the card run faster FPS wise.
 
You must log in or register to reply here.
Back
Top