Official HD 2900 Discussion Thread

Quote:

---

Originally Posted by Tim

OH my goodness, how can you even say that. This is driver HELL. Read HardOCP's comments on the card please! Working perfectly hahaha :lol2:

I'm under the table here man, please.

Just ordered a GTS one minute ago.

---

HardOCP is the only test where HD 2900 lost ALL the test versus GTS, THE ONLY TEST OF THE WHOLE INTERNET. This test it's a joke.

Quote:

---

Originally Posted by AbelJemka

HardOCP is the only test where HD 2900 lost ALL the test versus GTS, THE ONLY TEST OF THE WHOLE INTERNET. This test it's a joke.

---

The only review? Not really.

Quote:

---

Why everyone compares HD 2900 with GTX in order to bash it? Is HD 2900 cost 500$.

---

Nobody is, the 640mb GTS runs over it in the majority of benches. But 2900XT is their high end, so IMO it SHOULD be compared to the GTX. That's the way it always goes.

Quote:

---

Originally Posted by OBR

Poor performance is not in early drivers! ATi has long time to prepare good drivers, this card is just slower than GTX and better drivers cannot to solve this.

---

You just don't have a clue. Results have already been posted proving otherwise on this forum. There have been significant performance gains going from 8.36-->8.37-->8.38. As much as 20%.

Quote:

---

Originally Posted by Tim

OH my goodness, how can you even say that. This is driver HELL. Read HardOCP's comments on the card please! Working perfectly hahaha :lol2:
I'm under the table here man, please.

---

Working perfectly in context of what I said, yes. As in, works with DX10 without problems.

Performance issues do not relate to the stability/functionality of a driver.

And that HardOCP review was the worst. Guru3D was better. Heck even ocworkbench was better.

kinc proved 2900XT is very close to GTX when using Catalyst 7.5


expect 8800GTX price drop till the end of May

Quote:

---

Originally Posted by Tim

The only review? Not really.



Nobody is, the 640mb GTS runs over it in the majority of benches. But 2900XT is their high end, so IMO it SHOULD be compared to the GTX. That's the way it always goes.

---

Give me the review, i already read so much.
Let's go comparing 1000$ HDTV from Sony to 1500$ TV from Panasonic coz it's their High End.
ATI say they compete with GTS. Nvidia they GTX have no competition. Both say true. That's the way it goes.

@Techreport
http://www.techreport.com/reviews/20...t/index.x?pg=1

I think it can be summarised that DX9 performance is disappointing for a top model, high end part. The fact that ATi have priced this around the cost of a GTS make it considerable again, but power consumption knocks it back into the disappointing category.

Overclockability is great, but loud or aftermarket cooling is definitely needed. And again, power.

Me? I'm waiting. X1950XT is fine for me the moment, and I just got a nice boost from overclocking to 3.6Ghz. Firstly, I'm waiting for some big title DX10 designed games (e.g not a patched Call of Juarez) to really see how the two side's differing implementations fare then. Secondly, I'm waiting for subsequent driver releases to see what perfrormance gains can be had, becuase it's obvious that the driver is still very new.

Maybe by that time ATi will have brought a 65nm version of the card with better power and heat figures and/or a XTX version as well.

http://forums.overclockers.co.uk/sho...38#post9318238

Quote:

---

Originally Posted by Gibbo

Hi there

Well this might have to be removed as this product is still under NDA until midday today. However as everyone else is posting reviews I shall post the findings I had with the early test sample I had about a month ago on 8.36 drivers.


System Specification
To begin with we started testing on an Intel based system to see how well the card managed with Intel fastest processor today, the QX6800. The system specification was the following:-
Intel QX6700 @ 3.00GHz (1333FSB)
DFI ATI RD600 Motherboard
Corsair PC2-8888 (1150MHz) CAS4 DDR2 Memory
WD Raptor 150GB 10,000rpm HDD
Creative SB Fatal1ty Sound Card
Antec P900 Case
Enermax 1000W PSU
Windows XP Pro
Windows Vista Premium
BFG GeForce 8800 GTX OC 768MB GDDR3 (600MHz Core / 1800MHz Memory)
ATI Radeon HD 2900XT 512MB GDDR3 (750MHz Core / 1600MHz Memory)


Drivers
Installation with the NVIDIA card was easy under both operating systems. In testing all applications ran but Farcry had minor corruptions under both Windows XP and Vista which must be an issue with NVIDIA’s drivers. Also the Farcry benchmark programme would not run with the NVIDIA card under windows Vista. Apart from that the BFG card gave good results under Windows XP but it was noticeably slower under Windows Vista which clearly says NVIDIA still have a lot of room to make improvements on their Vista drivers.

Moving on the ATI HD 2900XT was also an easy installation under both Windows XP and Vista. All I will say is make sure that on release the cards also have the dotnet software as part of the installation too as to prevent any installation issues. Under Vista the ATI installation was very polished and superior to NVIDIA’s attempt.
Image quality was on par with NVIDIA for 2D and general windows applications. The noise levels of the card was quieter both at idle and underload, the card also ran cooler than an 8800GTX does as well, these are great points. Performance the card was every so slightly slower than the 8800GTX under windows XP, but this was less than 5%. However under Windows Vista the ATI card was quicker in both Direct3D and OpenGL application plus the card was more stable too and had no issues running my Farcry benchmarks unlike the NVIDIA card. Plus the ATI also displayed NO visual corruptions at all in Farcry wheras the NVIDIA card did.


Benchmark / Game results

NVIDIA BFG 8800 GTX OC (Overclocked) vs ATI R600 (HD 2900) at stock speeds (Overdrive ATI Speeds = 850MHz / 1800MHz)

Windows Vista

NVIDIA DOOM3 1600x1200, settings max, 8x = 57fps
ATI DOOM3 1600x1200, settings max, 8x = 68fps (71fps OC)

NVIDIA Farcry (1920x1200) = Crashes on benchmark programme
ATI Farcry (1920x1200) = Looks similar speed with better image quality and does not crash (54fps)

NVIDIA AquaMark Default Test = 154.28fps
ATI AquaMark Default Test = 155.36fps (159.54fps OC)

NVIDIA Fear 1920x1200, settings max = 33/64/140
ATI Fear 1920x1200, settings max = 22/54/108 (Possible driver issue here?)

NVIDIA 3D Mark 2005 Default test = 16,097
ATI 3D Mark 2005 Default test = 17,025 (17,457 OC)

NVIDIA 3D Mark 2006 Default test = 11,504
ATI 3D Mark 2006 Default test = 11,500 (12,504 OC)


Windows XP

NVIDIA DOOM3 1600x1200, settings max, 8x = 71fps
ATI DOOM3 1600x1200, settings max, 8x = 70fps

NVIDIA Farcry 1920x1200, 8x AF & 8x AA = 51fps
ATI Farcry 1920x1200, 8x AF & 8x AA = 55fps (ATI better IQ)

NVIDIA AquaMark Default Test = 163.43fps
ATI AquaMark Default Test = 158.37fps

NVIDIA Fear 1920x1200, settings max = 34/67/158
ATI Fear 1920x1200, Settings max = 23/57/142

NVIDIA 3D Mark 2005 Default test = 17,250
ATI 3D Mark 2005 Default test = 17,110 (17,554 OC)

NVIDIA 3D Mark 2006 Default test = 11,950
ATI 3D Mark 2006 Default test = 11,788 (12,617 OC)


As you can see from the results in brackets I overclocked the ATI card using the overdrive utility to compare against the BFG overclocked card. I ran the ATI overdrive utility which maxed out both sliders. The new clock speed were 850MHz Core and 1800MHz memory, even at these clock speeds the card still remained both cool and quiet. Now onto the results well I was very impressed as now the ATI card was pulling ahead of the NVIDIA card under Windows XP and under Windows Vista the R600 was considerably ahead of NVIDIA’s offering.
So in 3D Mark 2005, 3D Mark 2006, AquaMark, DOOM3, FarCry the ATI card won on both performance and image quality. It was only Fear where NVIDIA had quite a good lead which is no doubt due to the fact it’s a “best played on NVIDIA” game possibly or drivers?


Summary
Overall from my early testing the new ATI HD 2900XT 512MB looks like a serious contender to the GTX and GTS 640MB for gamers and benchmarkers. The only disappointing results were in the game called Fear but everything else was faster on the ATI product. The ATI card was also quieter and cooler running with fantastic overclocking potential. The product is a little more expensive than a GTS 640MB but cost considerably less than a GTX. So considering the card already has fantastic Vista drivers and far superior features to NVIDIA's offerings such as HDMI along with better video editing and DVD abilities does indeed make the HD 2900 XT worth while considering.

If your thinking of buying a new high-end card then I would suggest you don't make your decision based just on my review because my game testing is limited and as such I would recommend you check reviews done by professionals on a wider range of games before making your decision. If I was buying a graphics card and £300 was my limit then yes the HD 2900 XT would be my choice not only because its about the best performer in this price region but due to the fact it has great features, has little to no performance hit under Vista and is very overclockable. It appears most HD 2900 XT cards hit 850MHz-900MHz core and upto 2000MHz memory and beyond. However if you want the best then it does seem that the GTX does offer a performance edge on most games under Windows XP, but under Windows Vista the R600 based card is keeping pace with the 8800 GTX and for a lot less money in the programmes I tested. :)


Everybody already knows what the card looks like but here is one I took when I had the card a few weeks ago:-
http://img170.imageshack.us/img170/4061/atir600dx7.jpg


I know my review does not go into great detail and the game testing is rather poor but at least you now know a little more about the card. We have 200+ R600 cards in stock at this moment in time and you can buy them after midday today. :)

---

That Gibbo guy doesnt seem to match what everyone else gets.

Quote:

---

Originally Posted by Shintai

That Gibbo guy doesnt seem to match what everyone else gets.

---

Yeah read this.

Quote:

---

Originally Posted by Gibbo

Hi there

That was my testing a month ago. So the ATI drivers were 8.36 and the NV drivers were whatever was the latest a month ago, now remember NV have greatly improved their drivers since.

Also I ask you guys not to rely on my in-game benchmarks as I was pretty clueless with what I was doing so would advice that nobody buys a card based on my game results because I was not 100% confident of all the settings.

I advice you guys to wait until some reliable results come out from places like Anandtech etc.

One thing for sure is that OcUK is I believe the only company in the UK right now with stock and pricing will be under £300 Including VAT and you can have one in your hands tomorrow if you order from OcUK around midday today. :)

What did impress me with the R600 is how overclockable it was but the fact the fan never span fast enough to be audible, even when overclocked and under extreme gaming conditions. :)

---

The nVidia driver improvement is more or less BS.

And looking on Techreports noise graphs, he is full of BS with the fan too.

Quote:

---

Originally Posted by Baron

Yeah read this.

---

I just bought a 640mb GTS from Scan.co.uk, for £214 INC VAT! Match that Gibbo! £50 under the cheapest 2900XT on scan. :p:

Quote:

---

Originally Posted by Shintai

The nVidia driver improvement is more or less BS.

And looking on Techreports noise graphs, he is full of BS with the fan too.

---

It's gibbo, enough said! :p:

lol umm the only review i trust is vr zone .. shamino dont bs liek others do ...

r600 has massive problems with drivers....

look at the results in games bouncing around like nuts....

you can pick a game, a resolution and a quality setting according to what result you want to get......


what am i talking about?
(all numbers taken from vr-zone)

in quake 4 at @1600x1200, ultra quality, 16x af + 16x aa

8800 GTX 54 fps,
2900 XT 52 fps,
8800 GTS 640 MB 37 fps,
8800 GTS 320 MB 36 fps,

in quake 4 at @1600x1200, ultra quality, 16x af

8800 GTX 108 fps,
8800 GTS 640 MB 104 fps,
8800 GTS 320 MB 104 fps,
2900 XT 98 fps....

and pretty much the same way it continues throughout all the games... and the same goes for most other reviews.....

you can make one card do the happy dance on the competitors corpse just by picking the right quality setting, game and resolution.....

Quote:

---

Originally Posted by DilTech

To those who waited for this card, my condolences.

Even I'm shocked by how bad this card really is.

Even in DX10(thank you guru3d) the 8800GTX wins.

I hope AMD doesn't lose too much due to this, as the R600 is now AMD's GeForceFX mistake.

---

:wierd:

One DX_10 game were at 1280 without filters the performance is arround 27Fps average looks like bullsheat.
Even in DX_9 the drivers of HD 2900 XT are very bad, so in DX_10 they should be worse.....

Bring some decent DX_10 games to say something about DX_10 performance......

Quote:

---

There are many things we don't yet know about the GeForce 8800 and Radeon HD 2900 GPUs, not least of which is how they will perform in DirectX 10 games. I don't think our single DX10 benchmark with a pre-release game tell us much, so we'll probably just have to wait and see. Things could look very different six months from now, even if the chips themselves haven't changed.

---

@Bjorn3D
http://www.bjorn3d.com/read.php?cID=1099&pageID=3433

Quote:

---

Originally Posted by Shintai

The nVidia driver improvement is more or less BS.

And looking on Techreports noise graphs, he is full of BS with the fan too.

---

Noise graphs do not tell you what HZ sound the fan is producing. The human ear perceives sound at a certain HZ (say X) as much worse that at other HZ (say Y), even if the dB meter says X and Y are of the same dB, a human will think X is louder Y.

You humans suck! :stick:

Quote:

---

Originally Posted by MaxxxRacer

well in theory you could do a diff in motion by caputing all of the frames in BMP format and then going through and doing diff for every frame.

doing it by hand would take a year or so, but if there was some way to automate it, in theory it could be done. FYI though, it would require a seriously fast raid to do the capturing in lossless BMP

---

It is fairly straightforward to write this in software. It is likely that on the Internet sourcecode is available already. Only small parts would need changing.

Let's see....
I believe 3D Mark 2006 allows outputting sequences.
To prevent poor performance, a RAM disk is needed, but with the amount of RAM we have these days it shouldn't be much of a problem. 1 GB RAM-disk is easy now, so you can store at least 10 seconds worth of 30 FPS on it. Flushing large data to disk is straightforward too.
AutoIT can script the mouse/keyboard input, so the entire process can be repeated (although this can be done by hand).
The picture files need to be merged into a pure AVI.
Afterwards a VC-1 encoder should be used to create a crispy clear movie.

Actually the difficult part is transferring the files given the file sizes.

Wow Crossfire results are awesome (see Techreport)! ATI have done a good job there atleast -- scaling better than SLI.

Quote:

---

Originally Posted by Shintai

The nVidia driver improvement is more or less BS.

---

I just installed forceware 158.19 replacing 97.94 a few hours ago and I saw fps improvement in rthdribl. Max fps jumped from ~120 to 140 fps... I know it's not a game but I just shared an observation regarding driver improvements.

http://www.pcgameshardware.de/screen..._DX10_Snow.JPG

http://www.pcgameshardware.de/screen..._DX10_Cave.JPG

http://www.pcgameshardware.de/?article_id=601352

OOOOOUCH!!!!!! :slap: :slap:

Quote:

---

Originally Posted by insurgent

I just installed forceware 158.19 replacing 97.94 a few hours ago and I saw fps improvement in rthdribl. Max fps jumped from ~120 to 140 fps... I know it's not a game but I just shared an observation regarding driver improvements.

---

It was mainly just 3Dmark and only 3Dmark it improved.

And some more DX10

http://translate.google.com/translat...language_tools

Damn that DX10 game looks nasty for x2900XT :(

So much for lower prices I guess.