R600: Finally DX10 Hardware from ATI - Page 2

Last message on previous page:

I don't see how you can claim a driver issue "from looking at the benchmarks."

That, and drivers have 0 to do w/the fact that you need half of the nuclear power from soviet russia to power these things.

ATi is using superscalars (ironically like Geforce FX...), which thereotically gives them more power. Unfortunately it requires much more programming to keep them busy.

nV is using normal scalars, which is a much simpler setup. They only need to keep 2 shaders busy per unit vs. 5.

This is why drivers will make much more of a difference with 2900 than 8800, and also why the 2900 performance is so inconsistant.

I agree with the power. I'm tired of buying new PSUs and having a noisy machine.

I know this is a hard question to answer given that there are no DX10 games out there, but...

Reading the Tom's review, it appears the 2900 is geared toward DX10 architecture-wise (I know, 8800 is a DX10 card too, but the 2900 seems, on paper anyway, that it should be able to do it better). Is it possible that the 2900 will perform better than the 8800 in DX10 games?

I only ask because, like others, I've been sitting on the fence with an itchy trigger finger waiting for ATI/AMD to show us what they have in store for us before purchasing a new rig.

I would hate, based on these reviews/early drivers, to be disappointed with ATI and purchase the GTX now... then when DX10 games (Crysis/Alan Wake) hit the shelves find out that the 2900, with it's architecture, actually performs better.

Someone tell me what to do.

Only time will tell. Very good article, you guys went into major depths, i feel like printing it out, good stuff . I'm looking forward to the upcoming months to see both AMD/ATI battle it out with Nvidia.. I'll pick one out like one person has mentioned already when both companies put out there aces. And were talking about early on DX10 cards, theres sure to be more better ones and of course thats when it never ends and you can never really decide because a new card comes out so quick, hehehe.

Welcome to 4 years ago. http://www.extremetech.com/article [...] 535,00.asp


It's widely known that, in the past, companies have "cheated" at benchmarking software by figuring out the algorithms used to compute the scores and subbing in driver software that catered to those tests.


But again, either way you look at it, synthetics are worthless. If they were real numbers based off hardware, they're worthless. If they were fake numbers based off driver 'cheats,' they're worthless.

I have no clue why anyone would put stock into benchmarking software.

The same problem is with both Oblivion and F.E.A.R, the 2900xt Crossfire is not on the benches. Considering the 8800GTS is shown in SLI on them, and that in Doom3 they show Crossfire, that means that they intended to have crossfire will all there game test, but the drivers probably didn't allow it, or there were incapability issues. This leads me to believe that the drivers aren't ready, and there should have been more work.

I agree. If you designed an architecture around 3DMark, that really wouldn't be cheating either. Might be stupid , but its not cheating.

Hey MartenKL,

I don't think you were quite unintelligent, insulting and aggravating enough in that post. Could you please step it up a notch?

Oh, and HDMI is useful but most people wouldn't combine media center with gaming PC. So, that's more of an added value for ATIs lower end cards.

Thanks and have a great day,

-Dean

Well the official reference hardware used for development of DX10 itself is nVidia's G80 so my bet is no.

you guys have to understands is that this card isn't supposed to compete with the 8800GTX, only with the GTS. the 2900XTX is for the GTX. PLUS the cards that are just coming out are on 80nm cores, that means higher power and temps. wait till the September refresh(if u can wait) and you'll see the real battle.

OH did u guys for get that NEITHER nVida or ATI have fixed all the problems and glitches with the drivers? there is lots of hidden power left in the 8800s and the 2600. drivers are like wine, the get better with time

That's what I'm saying about the people talking about drivers... someone said that there was a 5-10% difference in the new 2900 drivers. Of course there is!! That happens just about EVERY time a new card is released! Then, as time goes on, each company makes better drivers... and each card increases in performance. Other than the FEAR demo screwup, there's nothing that is inherently *wrong* with the drivers though, so I don't really think that "better drivers" are the hopes that ATI fanboys should hold on to.

Also cited was the fact that nVidia squeezed out a driver from where the sun doesn't shine when they heard that ATI dropped improved drivers.. just so that their benchmarks would look better as well. If ATI had drastically improved drivers in the future, I wouldn't be surprised if nVidia does the same thing again...

I double that.. Hot, Powerhungry and loud monster (with crippled legs). I lost 2 Nvidia cards to high temps and switched to ATI. Now it's back to Nvidia again

Many say that this card must compete VS GTS 640 but GTS 640 can rarely (if ever) beat GTS 320 witch is a lot cheaper

http://it-review.net/index.php?opt [...] &Itemid=91

And even if you manage to get to resolution where GTS640 tops 320 version your FPS will be limited by GPU's core clock so it's a loss anyway.

In my perhaps n00bish opinion:

- 2900XT fails to compete with GTS 320 because of the price/preformance ratio

- 2900XT fails to compete with GTX couse if you compare the two 2900 plainly sucks (beeing hot/loud/slower in practice)

- 2900XT pwns GTS 640 but what's the use since you could say the same for GTS 320 pwning it's older brother.

Maybe ATI has a wonder driver up it's sleeve to unleache 2900's full power but I (and I think many of us) are tired of waiting. I simply see no reason for ones who waited for ATI's release going for 2900 at this point and since I am included in that category will be buying GTX with a full C2D setup in a week

Hot is right...

Look what I found in the release notes for the current driver

"Hardware thermal protection is now activated when the temperature rises above 120 degrees"

8O

http://support.ati.com/ics/support [...] onID=27535

First of all, comments are already too numerous for me to read them all. I just spent what feels like an hour reading the entire article, so I'm not going to torture myself reading the comments too (I enjoyed reading the article, but there are only about 1 of 3 comments that are actually useful Having said that...

...Cool architecture DOES matter, and the article explains that in several places. Architecture is what means the newer R600 replacements will be much faster or not much faster, and architecture is the only way to compare the card itself to other cards due to the "windows" factor (OS version, driver version, etc.) NVIDIA may out-benchmark this card today, but if ATI can get a driver to fix those last-minute addendums or glitches that may or may not exist, then ATI could dash ahead.

Example is the initial 8800GTS benchmarks from websites in Vista 32-bit. Look at those. Now benchmark it with the nvidia drivers from May 2nd. See a slight difference? (Hint: The word 'slight' is a huge understatement!)

So who's better? Truth is, who knows? Only the hardware itself can tell us, but sadley even with superior hardware drivers could suck and make it appear to be worse even if it's better. Will ATI best Nvidia with DX10 now? Depends on nvidia's and ATI's ability to write drivers on one hand, while the hardware could be vastly superior from one manufacturer, have poor drivers, but still best the competitions.

Nothing about this debate is simple, yet everyone and their mother will try to tell you something incredibly stupid at this time and say "AMD is dead, even the 8800gts beats ATI's newest card" or "the R600 will beat NVIDIA's 8800GTS". They're giving opinions at this point so anyone who tells you that during this week is NOT telling you the truth.

/me agrees with shark dude.

You're the lucky one who happened to have the comment on top of page three, so I read it. Yeah, you're thrilled to death that some moron on a hardware forum read yours I know (try the lottery, today may be your luckey day )

As I stated in my last comment, I personally believe (that means that the following is my opinion, and not an infallible absolute truth...saying for any fanboys who may be reading this) that it is simply too early to tell. Wait for the next ATI card first, and see what the numbers say then. If you need a video card today--at this instant in time, 4:04 PM Central Time May 14th, 2007--it appears that with the drivers used in these tests that the R600 does not best the 8800GTS, or at least trades blows with it.

So based on the information I have right now, nvidia appears to have the better deal for the money--but that could change very quickley.

Looking at the current benchmarks from different sites I can conlclude that the 2900XT competes with the 8800gts 320 and outdoes the 8800gts 640. Anyone that has said otherwise doesn't know how to read graphs and analyze data. And this complaining about drivers is rubbish. Drivers with new video cards are always wishy washy. The drivers will only get better.

I think people are just complaining because the 2900XT wasn't the monster release that the 8800GTX was. But it's obvious it wasn't released to compete with the GTX. Right now at newegg the 2900XT is selling for $429. Being that this card is brand new, it is very fair and competitive with the $399, 8800GTS.

Overall I think this was a good release for AMD/ATI. In my eyes it didn't disappoint. Rather it shows promise for a good competitive future.

First post here.

The review is ok but I wanted to see benchmarks in D3d10. The good news for everyone is Capcom is releasing a D3d10 benchmark for LostPlanet in only 27hours. You can all download from http://www.lostplanetcommunity.com/ Then we can benchmark the cards ourselves in D3d10 the way they were designed for.

Thanks to Darren for his review. I am sure we will be getting some lots of D3d10 benchies in the next few weeks anyway but who wants to wait, right?

Kindest Regards,

Mr Slartybartfast.

ummm... actually, the 320mb's are going for $310 or so.

It WOULD have been a good release IF it would have been out 6 months ago when it should have been in order to keep things "competitive". With all the known issues with these right now (not like the 8800's didnt have theirs), the card was delayed for what? Not one good reason I can see yet... yes, I said YET.
Unless these cards do something miraculous under dx10, ATI has shot themselves in the foot with their delays and horrible market timing with a "cutting edge" card that hardly can claim to be that.
Ati does show promise, but I'm disappointed as hell that ATI cant even keep up on the high end. Makes me wonder if I should hold my breath on the new AMD chips. Probably not, because the way they are running things, I'll be holding it for 6-8 months after the initial launch dates....

Just wanted to throw this out, if you check www.anandtech.com and check out their 2900 XT test, it says that the problem with the 2900 XT was due to bad drivers and faulty manufacturing.

So what was good about it then? If the hardware is bad and the software is bad, what's good about waiting 8 months for them to "get it right" as some ATI engineer put it?

i was looking forward to more crossfire benchmarks

I think that's a pretty stupid comment to make. Of course I should expect nothing less coming from you. We've already seen improvements from the first drivers. Improving the drivers most certainly can help.

Looks like the Alpha drivers did help some situations, so what you're saying isn't reflected by all tests (BF2142 and Oblivion with AA);

http://www.techreport.com/reviews/ [...] ex.x?pg=12

http://www.techreport.com/reviews/ [...] ex.x?pg=13

I don't know aht future drivers will bring and I'm not one for relying on 'magic drivers', but it's obvious that there was a difference in the alphas that The Tech Report had a chance to play with.

So the assumption new drivers will do nothing is ignorant especially consideirng how new drivers helped the G80 series at launch and even recently. I don't expect it to do much 'dramatic' but I wouldn't be surprised if it does something positive, contrary to what you state.

Interesting read. I hadn't read that TechReport article yet.

ummm... actually, the 320mb's are going for $310 or so.

http://www.newegg.com/Product/Prod [...] 6814122022
$269.99 after MIR.

OpenGL is getting two major specification changes this year. The first will streamline GL as well as give it full to all of the features DX9 hardware has to offer. That specification is called "Long's Peak" while the second is called "Mount Evans." The second will bring the functionality of DX10 hardware to OpenGL. Here is an article by Barthold Lichtenbelt at NVIDIA who also is the Khronos OpenGL ARB Steering Group chairman.
(http://www.opengl.org/pipeline/article/vol003_1/) Hopefully this will spur some more OpenGL development but for now it is heavily D3D for most games for Vista.