Printable View
saying we should focus on dx9 because there's no dx10 game out yet is stupid. crysis, alan wake, et al are coming, IF g80 can't compare with the r600 in dx10 that would be very disappointing. buyers expect the g80 to perform well on dx10 not only on dx9, they are high-end after all.
Exactly, the FX 5900 was very disappointing too on Pixel Shader 3.0, so the same could happen with G80.Quote:
Originally Posted by insurgent
Buy a 8600 GT passive and touch the heatsink, you'll get hurt too unless you put a quiet active fan on or very near to it.Quote:
Originally Posted by Tim
Clearly this isn't finalized yet and contains many stupid errors.Quote:
In this seventy pages document you can read about the power connectors and pay attention on the page 6 as AMD suggests you to upgrade the AGP drivers before you install this AGP card. Well done.
Lighten up, it's just a joke, not bashing the XT. :)Quote:
Originally Posted by Noobie
Some more hilarious stuff. I mean, wow, I can do that in paint!
Yea it's clearly not finalised, probably Fuad has an old copy, haha.
Ey? :confused:
(from the pdf)Quote:
GPU Physics
GPU Physics are only active when Crossfire is inactive. This ensures the highest possible performance.
Its no secret the GPUs cant "multitask". The original concept was to use an old GPU or minor GPU for physics. Like x1600 series.Quote:
Originally Posted by Tim
Hm, oh yea I remember, hence 3 PCIe slots. Shame.Quote:
Originally Posted by Shintai
Sorry, my joke detector is a bit off :stick:Quote:
Originally Posted by Tim
I have the original bitmaps, put them here, hopefully the XT owner will do the same.Quote:
Originally Posted by MAS
http://www.rage3d.com/board/showthre...3890816&page=3
http://www.sharebigfile.com/file/167947/gtx1-bmp.html
http://www.sharebigfile.com/file/167952/gtx2-bmp.html
Or png
http://img50.imageshack.us/img50/4821/gtx1wt7.th.png
http://img520.imageshack.us/img520/8040/gtx2sp4.th.png
Anybody know a better host?
try imageshackQuote:
Originally Posted by fornowagain
Doesn't work, just resizes them, file size is too large.Quote:
Originally Posted by jimmyz
From the 2900XT users guide.
http://img153.imageshack.us/img153/2116/image1vl7.jpg
That explains the HDMI, well sort of.
here you go:
[IMG]http://img153.imageshack.us/img153/9948/1dfpg7zc7.jpg[/IMG]
and the other.
[IMG]http://img153.imageshack.us/img153/6772/2cfgq1ap1.jpg[/IMG]
Thanks, but I think you've misunderstood. They're way more compressed than the Jpegs I posted a page back. I'm trying to compare the original bitmaps.
http://www.xtremesystems.org/forums/...&postcount=978
Out of curiosity, what settings are you using (both in 3D Mark & nVidia config)?Quote:
Originally Posted by fornowagain
Application controlled in the NV panel, no optimizations. Highest in 3dmark06 and the same as the other guys.Quote:
Originally Posted by Noobie
http://img291.imageshack.us/img291/9798/image1bu1.jpg
Who's buying?
In the first set of pics the boxes on the HD have more detail. The telephone wire to the left have a bit more detail to them as well on the HD. As for the second pic the pointed spear on the front of the balloon is more visible in the HD then in the G80. From what I read in that forum those pics were setup by 2 different persons. And, the G80 pics have less compression.
i am.
as soon as it hits retail I'm buying.
Ill be buying.
i'll wait. maybe it's true that it performs better in dx10 but the 8800gts are cheaper now so there's a lot to consider.
Ya I noticed that as well. Not bad for a highly compressed image file.Quote:
Originally Posted by Eastcoasthandle
BTW I bought already should be here Wednesday :( If lucky Tuesday but doubt it.
Definatly will be trying the Lost Planet demo both in DX9 and DX10 though, DX10 first ;)
Do you have it? If you do you mind providing a link?Quote:
Originally Posted by deathman20
I wanna see some retails OC results before i make a decision ;)
The Demo is coming out the 15th so you can get it then.Quote:
Originally Posted by Eastcoasthandle
The benches of the DX10 of Lost Planet on the GTX are already done.Quote:
Originally Posted by deathman20
Translation:Quote:
Die Testwerte (siehe unten) zeigen deutlich den großen Hunger nach Grafikkartenleistung. Die Geforce 8800 Ultra befeuert das Spiel im D3D9-Modus ohne Probleme. Schaltet man jedoch in den D3D10-Modus samt Soft-Shadows, bricht die Framerate über 50 Prozent ein. Dieses Phänomen kennen Shooter-Freunde schon von F.E.A.R. (dt.) und Riddick (dt.), welche mit Soft-Shadows zwar grandios aussehen, aber ähnlich einbrechen. Erfreulich ist jedoch, dass der D3D10-Modus mit den Detail-Einstellungen des D3D9-Pfads etwa zehn Prozent schneller läuft.
The test-values (see below) clearly show the hunger for graphic-performance. The 8800 Ultra runs the game in DX9 without problems. However, if the DX10 mode, along with soft-shadows, is used, the framerate drops more than 50%. This phenomena are known to FPS-friends cos whilst FEAR & Riddick look great with soft-shadows (Noobie's note: not it doesn't, FEAR soft-shadows quality was tested a long time ago by a site and it was found to be very ugly), performance drops similarly. However, it is pleasing that the DX10 mode with similar detail-setting as the DX9 mode is 10% faster.
I could be wrong but I think that most reviews tomorrow will use 8.36 nstead of 8.38 drivers. If this turns out to be true that's going to suck.Quote:
Originally Posted by LowRun
Ouch. Thats a heavy hit.Quote:
Originally Posted by Noobie
In the MSI and Asus kits, both are included but the 6-pin is 1x6 to 1x8. The molex connector is 2 molex to 1 8-pin.Quote:
Originally Posted by fornowagain
the review is available in vr-zone now but it loads slowly, i don't know if it's a test or an error
The majority of reviews should be using the 8.37.4 drivers, the 8.38s (early beta 2) still have a lot of issues, especially under Vista with CrossFire and OpenGL. Also, the 8.38s do not have all of the IQ settings optimized yet, so a lot of those reported increases are not exactly fair comparisons. ;) However, the final 8.38 drivers will improve performance over the 8.37.4 drivers, AMD is expecting 5~12% improvements.Quote:
Originally Posted by Eastcoasthandle
Quote:
http://i112.photobucket.com/albums/n...der/site_2.jpg
http://i86.photobucket.com/albums/k1.../HD2900XTs.jpg
http://img242.imageshack.us/img242/6...ossfireui3.gif
Qualidade Antialising:
AA guys ...
I run Doom3 under Windows Vista 32bit .... ups, old game that I love
Simply press F5 and F9, and then shot a pic.
(1) Box l 8xAA l 8xSamples + 16xHQAF
http://www.gamexeon.com/forum/imageh...6516bb6e62.jpg
http://www.gamexeon.com/forum/imageh...6516bd1813.jpg
(2) Narrow-tent l 8xAA l 12xSamples + 16xHQAF
http://www.gamexeon.com/forum/imageh...6516c1d507.jpg
http://www.gamexeon.com/forum/imageh...6516c2713e.jpg
(3) Wide-tent l 8xAA l 16xSamples + 16xHAQF
http://www.gamexeon.com/forum/imageh...6516c67073.jpg
http://www.gamexeon.com/forum/imageh...65265a32c4.jpg
:thumbsup:
First real preview!
http://www.vr-zone.com/?i=4946
Nice! :toast:Quote:
Originally Posted by OBR
It's getting hammered though. :mad:
Yeah and it getting hammered to death.Quote:
Originally Posted by OBR
I'm in!!!!
Sounds like the re-introduction of hardware Truform to me, which the 8500 has (later series only had software truform), which is brilliant because TruForm allowed Morrowind to look well beyond it's time. Screenshots of Truform on morrowind can be found on http://tes.lanior.ru/tweaker/images/root.htmlQuote:
Among some of the new features of the 2000 series include a programmable Tessellation Unit. This feature enables the GPUs to apply image enhancements when support is written in games such that game developers have a much easier time doing up simpler graphics, and letting the GPU do the image enhancements. This also supposedly saves on GPU Horsepower as you can see below. You get 2.3x performance compared to a conventionally rendered scene.
Low Res with Tess:
840 triangles - 70 KB Vertex Buffer Size - 23 KB Index Buffer Size - 821.41 FPS - Resulting image has 1,008,038 triangles
High Res without Tess:
1,280,038 triangles - 31 MB Vertex Buffer Size - 14 MB Index Buffer Size - 354 FPS - Resulting image has 1,280,038 triangles
It's definitely a great feature if it works as advertised, but support on games is currently non-existent. When asked to name 3 games coming out with this support, the speaker was unable to disclose details.
link its OK but very slow , im in 1st page
Too bad they dont used the New 8.38 driver :(
If anyone's already read it, could you please post numbers for us?Quote:
Originally Posted by OBR
TIA :)
Yeah, vr-zone is terribly slow.. must be due to lots of traffic i guess:)Quote:
Originally Posted by Sumanji
download vr-zone review here: http://rapidshare.com/files/31092804...s.net.zip.html
they used the 8.36 and 8.37 driver , they didnt try with new 8.38
anyway here some pics from the review in call of juarez
http://aycu21.webshots.com/image/159...3545314_rs.jpg
http://aycu04.webshots.com/image/171...5570292_rs.jpg
Quote:
The most obvious difference is seen when looking at the much better-looking water effects, and High Dynamic Range effects.
IQuote:
It is slightly off tradition that the GPU company's flagship product sails off not to meet the flagship of it's competitor, but one target lower. Then again, the lower we go down the price pyramid, the bigger the audience, more people with the budget to spend. I'd say that there is no clear winner between the 8800 GTS and X2900XT, the GTS displayed more consistent performance behavior while the X2900XT fluctuates around due to the in-matured driver.I would say that despite the heat thrown out by the GPU, the X2900XT overclocks better than the 8800GTS by 8-10%, but that's putting out a lot more heat and drawing more power than it already consumes. So this is something potential XT buyers should take note of, the heat produced by the card is no small amount, nor is the power consumed by it - more than 60w over the GTS. What you would be investing in is a higher potential of upcoming performance boosts (including the latest pre-Alpha 8.37.4.2_47323 Catalyst just released 3 days before this review) and full HDCP support with integrated audio controller. And of course the new programmable Tessellation technology which we will probably not see support in games until much later.
Not the fastest video card in the market for sure, but definitely holds it's own at it's current price-point. We only hope that supply will be adequate and not lead to an indirect increase in prices due to short supply. We hope to see some interesting implementations from various card partners as well, be it overclocked specifications, or improved coolers.
Quake 4 performance for the HD out performs the GTS 320/640 when 4xAA or better is used.
http://i11.photobucket.com/albums/a1...dle/th_102.jpg
http://i11.photobucket.com/albums/a1...dle/th_101.jpg
There are some 8.37 driver issues with games like Lost Coast and BF2142 as it performs worst then the 1950xtx when AA is turned on.
Cool! Am reading it now.Quote:
Originally Posted by w0mbat
Since it took me so long to respond i figured that I would pm too:Quote:
Originally Posted by jagt
catalyst ai as far as i understand it does recognize applications so by renaming the exe you tricked it. It may be that it was applying to many quality short cuts for the sake of speed or that the optimizations did not work for that game and somehow canceled out the AA.
Thanks, but getting file corrupted, I will try again, anyone else ?Quote:
Originally Posted by w0mbat
EDIT**
Oh I forgot, fileshare doesnt like me .. 1 download then after that it tells me I exceeded my DL limit :(
Just double checking, any 2x HD2900XT will run crossfire, right?
Not like the old days that required a crossfire edition card (master card).
Yes.Quote:
Originally Posted by dragonhunter
From Vrzone review
Quote:
As said before, I had results from both the Catalyst 8.36 and Catalyst 8.37 Drivers. How did performance improve between this small driver jump of 0.01 version?
Taking a setting at 1600x1200 with 16xAF, I saw a major increase in performance, particularly Company Of Heroes and Quake 4. Performance went up by 11% on COH and 42% on Quake 4! This shows that the drivers is still very raw on this card, with just a minor driver revision boosting up performance that much, it gives us quite a lot of hope for a fair bit of improvement to come. Let's hope for that!
. When Anti-Aliasing is turned on, the X2900XT showed that it carries it off with great efficiency in games that the drivers are optimized for, performing significantly better than the GTS; while the AA efficiency is piss-poor in some games due to the raw driver which has not fully blossomed to take advantage of ATi's new GPU technology. Just take a look at how performance has boosted from Drivers 8.36 to 8.37, that shows the potential in performance growth... a whole lot of it to reap.
thanks! :)Quote:
Originally Posted by w0mbat
It comes close to the 8800GTX sometimes, i think with some proper drivers this card will kick some 8800gts a$$ :D or i hope so
another review ( ATI 8.361 (for HD 2900 XT) ) :(
http://www.tweaktown.com/articles/11...ves/index.html
regards
kinda off topic but when can we expect HD 2600 reviews?
just finished reading the downloaded preview and I must say, for the most part I am impressed with the performance over an X1950XTX. I was worried it wouldn't be much faster than the X1950XTX by some earlier results that were leaked. This doesn't seem to be the case so performance wise it should definitely be better than my current card.
What I am disappointed in slightly is how the GTS beats this card even in situations where I think it shouldn't. The 320mb GTS wsa beating the card however this could just be due to driver issues so I won't hold it against the card in that respect.
What really disappoints me is how the AF hasn't been improved at all since the X1900 series. Yes the AF was already extremely good, however, the 8800 bested even that and I was hoping the X2900 would at least match, if not beat the 8800 in terms of AF quality. It seems ATI felt the AF quality was 'good enough' considering the X1900 was the best card for IQ comparisons. Now that the 8800 shows better the ATI card feels 'lacking' in this department. Surely they could have done better seeing as though Nvidia did, and they were always far INFERIOR IQ wise than ATI was. This is something I can't see changing with drivers either as it seems to be how the hardware itself applies the filtering. I am glad AA is slightly better however.
I really hope they come out with the 65nm x2900XL/XT soon. I really dislike the power consumption alot. Else overall a good candidate for a card instead of a 8800GTS.
Maybe we can get rid of nVidias crappy new super price on their Ultra too.
Quote:
Originally Posted by Eastcoasthandle
Well yeah,because they use Best AA and for 8800GTS would be 16X CSAA if I am not mistaken anyway I look forward to another review with a better driver...
it seems all reviews are with 8.36 or 8.37 , i think we have to wait few more days to see a review with 8.38
regards
mascaras....relax...we can read the reviews. You almost sound like AMDs Richard...
Just because one driver made a big difference BEFORE the card was released doesnt mean the next driver will bring miracles.
You even post the same thing several times...
Quote:
Originally Posted by Shintai
im not talking of next drivers , 8.38 it was released few days ago , my friend have it for 4 days now
i dont want miracles , the actual driver its very good in 3dmarks , now i want a good driver for games , i understand its dificult and takes time ( nvidia takes 6 months to get a good performece driver (158.19/19) i just want a good game driver fast :D
Looks like tweaktown used 8.36 ;). I am not expecting much from this review site:Quote:
Originally Posted by bingo13
-no gts
-no max fps, just min and average
-no updated drivers
I am sure that the Power Supply Unit: Seventeam 500-watt is good enough PSU. :confused:
If it was so important, then im sure AMD would have given the review sites it and made an effort to tell them to use it. Or get your friend to make a full review. No NDA to worry about...Quote:
Originally Posted by mascaras
Ocworkbench got something to with 8.37 driver...
http://www.ocworkbench.com/2007/asus/EAH2900XT/g1.htm
When comparing Nvidia 's 16 x QAA mode vs Ati 's 8 x it's quite on par with GTX:p:
Even though the TweakTown test showed that Nvidia owns the top end, it would've been nice to see the 8800 GTS in there.Quote:
We tested with the latest available testing driver from AMD which is 8.361 – and this is the shipping driver for Radeon HD 2900 XT. There is a new driver floating around but that is not entirely stable yet. We would not be surprised if AMD releases its Catalyst 7.5 monthly driver soon and that should probably see some performance improvements and fixes of known bugs. The driver that we are using now is not perfect and AMD has noted some known issues. For example, in some of our tests, the older Radeon X1950 XTX is actually faster than the HD 2900 XT – this is a driver issue and we hope AMD fixes the problems very soon.
Quote:
Originally Posted by red
drivers Avaible in ATi FTP :
Cat 8.361
CaT 8.374
Car 8.38 >>http://www.xtremesystems.org/forums/...&postcount=869
Quote:
Originally Posted by red
i have a BFG 8800GTS 320mb and i will receive a X2900 nxt week then i post some results to compare .
Quote:
Originally Posted by koosjebram
nice scores in OC review
http://aycu34.webshots.com/image/167...0827099_rs.jpg
Quote:
Originally Posted by mascaras
The 8.38 drivers are an early beta 2 release, they were leaked. Those drivers have more issues with them than the 8.37.4 in several cases. CrossFire and OpenGL in Vista are messed up, IQ has been reduced in several cases also until they tune it. ATI does not want the review sites using those drivers until the final release on 5/23.
Can the red vreg (or mosfet I forget) HS from the x1900xt/xtx fit on the HD2900 XT. Because the HD 2900XT have that HS built right into the GPU HS. So if you take the GPU HS off you also take off the other HS. No Water cooling until a HS can be put on it IMO.
From looking at VR-Zone's review he was able to get 1 push down bolt on one side but had to use a blue plastic clip for the other side. That means 1 of 2 things:
-he broke the other push down bolt
-the other push down bolt doesn't fit the hole on the HD 2900XT
Ok thnks for advise :toast:Quote:
Originally Posted by bingo13
So the "good offical" drivers will be released about 10 days after launch?
http://www.ocworkbench.com/2007/asus/EAH2900XT/g7.htmQuote:
Current DVi-HDMI adapters do not carry audio. This Active DVI to HDMI adaptor does. It allows full video and audio support via standard DVI output from Radeon HD 2000 series graphics cards
:)
Oh really??Quote:
Originally Posted by Shintai
This: http://www.vr-zone.com/?i=4946&s=9 shows it consumes a "whooping" 17W than 8800GTX (not ultra!!! ).I guess it's gonna burn a hole in your wallet for a power bill if you go for a Radeon instead of a 8800GTX.
Why would you compare the power consumption of a slower part to that of a faster part? VR-Zone does have 8800 GTS numbers to compare to.
They bencht the Nvidia Cards with very old drivers.Quote:
Originally Posted by koosjebram
Lol..guess they needed 5 months old drivers and 16 vs 8 X AA too make the HD2900XT look good.Quote:
Originally Posted by Gorgul
I'm beginning to think we'll have to wait a month or so more to see what's the real deal with this card and mature drivers:(
Not for another 3 months or so I fear. I am worried about the amount of Processing Units (PU) they have put on it tough....Quote:
Originally Posted by insurgent
For easy calculations sake.....
PU SPEED - ATI
320 PU's * 742 MHz/PU => 475 GigaFlops (derived from VR-zone review)
120 PU's * 800 MHz/PU => 192 GigaFlops (calculated & matches VR-zone)
PU SPEED - NVIDIA
96 PU's * 1200 MHz/PU => 345.60 GigaFlops (derived from WikiPedia)
32 PU's * 1450 MHz/PU => 139.2 GigaFlops (calculated & matches WikiPedia)
If you compare the 2900XT to 2600XT calculation power, you see it matches nVidia's 8800GTS to 8600GTS almost spot on.
MEMORY SPEED - NVIDIA
8800 GTS - 320-bit & 1600 MHz -> 64 GB/sec
8600 GTS - 128-bit & 2000 MHz -> 32 GB/sec (again by calculation this is correct)
We see a 2.0x decrease in mem-power with a 2 1/2x decrease in shader power for the 8600 GTS over the 8800 GTS
Now image what ATi can do. They can put a 512-bit ring-bus on the 2600 XT, which is costly. Instead it is more likely they will put a 256-bit ring-bus on it. This decreases the mem-power by 1/2. If they slap faster memory on the 2600 XT, similar to what nVidia did with the 8600 GT we get the following....
We will see a 1.6x decrease in mem-power with a 2 1/2x decrease in shader power for the x2600XT.
The real question is then, how much difference faster mem will make. Luckily nVidia also spawned the 8600 GT. We again turn to maths:
139.20 / 113.28 => the 8600 GTS has ~1.25x as much shading power
32.00 / 22.40 => the 8600 GTS has ~1.43x as much memory power
675 / 540 => the 8600 GTS has 1.25x as much core power
Thus the 8600 GTS has more of a mem-increase then increase in other parts.
We turn to a review on http://www.guru3d.com/article/Videocards/426/16/ after which we run some number-crunching and we find out that : in the review the 8600 GTS is 1.1x faster than the 8600 GT. Why? Slack most likely. So we reach a dead end.
On Guru3D we find a comprehensive 8600GT/GTS roundup and we turn to a quote:
8600 GTS Galaxy 700/1000/1620/6029 (core/mem/shader/3D_2006)Quote:
Yeah, you guessed it already. BTW make note of the performance from the Galaxy GTS card. Core and memory clocks are the lowest among all the GTS cards yet the score is really good (throughout most benchamarks actually). This is because Galaxy overclocks the Shader unitdomain really high. We'll get back on that later.
8600 GTS BFG OC 710/1008/1512/5883
So an increase in 1.07 for the shader, results in 1.025 performance increase, this is not really linear.
We for now turn our attention to the GPU & Core OC test:
8600 GT Galaxy @ 600/800 - 58 FPS
8600 GT Galaxy @ 730/900 - 61 FPS
FPS increase : 1.05
mem increase : 1.125
core increase : 1.21
A mem & core increase does not scale so well either.
In short: the balance the videocard makers have made between mem, core and shader is so well thought out, that an increase in either shader, core or mem does not have a significant contribution to the end-performance.
Thus I conclude that with the ATI 2600 XT it is unlikely that ATI will use faster memory or core, but rather will keep the price low.
The question than becomes: how low will it be? With the 8600 GTS we have already seen that it sucks, cos the 8800 GTS 320 MB is far better value for money (less than 1.5x the price, but roughly 1.8x the performance). With the 8600 GT it is a little better.
How low can the price of the x2600 be? ATi will not get it low enough I fear because of the fixed costs with card manufacturing. The 65nm could help a lot tough.
Even worse the 8800 GTS has ~2.5x the shader power, whilst only having 2x the memory power of the 8600 GTS. The performance gap between 8800 GTS and 8600 GTS will improve further in future shader-intensive games. And with the 2900XT and 2600XT it is the same.
So why do the crappy cards sell? 260 euro for a 8800 GTS 320 MB (price here) is a lot of money, the 8800 is a power-hog and it is hot. Moreover the money you save from not buying a 8800 GTS 320 MB, can be put into use in your next videocard. It's a matter of whether your going to do a lot longer with the 8800 GTS 320 MB than with a 8600 GT/GTS.
ATI at least has the advantage of their x2600 being a 65 nm process. But they have hinted at a 65 nm x2800 XL which I guess is going to be a 256-bit bus with 160-240 PU's, and might be better value for money.
.....I wanted to get a x2600 XT, but now I write this it looks crap.....
Note of self-reflection: it is amazing to see the core-speed of the 8800 GTS vs the 8600 GTS. 500 vs 675, a factor 1.35. I didn't even use it in my calculations. I think the 8800 GTS has a much better core value than the 8600 GTS, which is simply needlessly fast seeing as how the heavy core OC of the 8600 GT didn't do much good.
Uhm...burn a hole? yes! Would I buy the 8800GTX or Ultra? No.Quote:
Originally Posted by informal
I dont want a card that cost more on the powerbill for a year or 2 than the card itself cost.
So relax mr. Defender. I wouldn´t buy a 2 slot cooling GFX card.
And last time I checked, the x2900XT didnt beat the GTX but is on pair with the GTS. And the GTS is 65W lower in that review.
Side note: Those power consumption numbers are useless to Vista Aero users. The 8800 can turn parts off, to save power. On one online test it was shown that in Aero the 8800 used less power than a videocard X, whilst vid-card X used less power in idle and less in full-load.Quote:
Originally Posted by red
Then again I think Aero should be removed from Vista.
Noobie...the 2600 series is 128bit...and you cant calc the flops scaling like that.
As for the performance:this card has a lot of headroom trough the driver optimizations,it's fairly obvious thing with the latest set.Read Shamino's comment at the review about the game scores that went from 11% to 42% in different scenarios(games).
As for power bill, 65W is not that much for those who have the money for the card at the first place.And most of the time,card would be idling ,and the idle power is very good compared to other Nvidia cards.
Dunno if it has been posted yet but here is VR-Zone final review:
http://www.vr-zone.com/?i=4946
That OC work bench review is the worst.
Few weeks and I'm going to pick one up. Hopefully by then some of the driver issues that are holding his card back will be fixed.
Overall I like this card for price point versus performance. I figured it would fall between gts and gtx in performance and with the OC ability I am not disappointed. However I want to see how these cards scale in CF as compared to SLI. If they scale better by 15% or better they are gonna be even better to me. I think SLI is a rip off for the cost versus extra performance.
I didn't calculate the FLOP but used ATI (trough a review) & nVidia's (trough wikipedia) information.Quote:
Originally Posted by Shintai
Here's the calculation nVidia used:
(MADD(2flops) + MUL(1flop)) * 1350MHz * 128 SPs ＝ 518.4 Gigaflops
Note that I am not comparing ATI vs nVidia based on Gigaflops (which would be stupid cos I would need to take slack, latencies etc into account).
I do not care about ATI vs nVidia, I care about the 8600-8800 relation and am using it to reflect on the possible 2600-2900 relation. 128-bit could work when they use faster memory. 512 = 3.2 * 1.25 (faster memory) * 128 (ring-bus). Card scaling would be 3.2 for mem vs 2.5 for shaders.
This would make the 2600XT-2900XTX more like a 8600GT - 8800GTS relation. That's good cos the 8600GTS:8600GT ratio is 1.25 for core, 1.25 for shader, but 1.43 for mem and the performance of the 8600 GTS is poor compared to the 8600GT (factor 1.1 mostly). But the 8600GT - 8800GTS is 2.85x, 3x for mem, shader. More shader power for the ATI compared to mem is nice, although this may be a bit out of proportion. It better be darn cheap to compete with the 2900 tough!
Note that it is really not wise to do such calculations and that results can vary greatly due to not taking a lot of things into account, but this is all I can do with the information & time I am given.
http://www.theinquirer.net/default.aspx?article=39578
Nvidia's anti ATi presentation
performance of the hd2900xt card bounces around all over....
(according to the vr-zone review),
in some games nvidia does the happy dance...
in some other games its ati...
in some others they on par....
so.....
what should i think of that card?.... well... we will see what future drivers will bring... but then i might aswell stick with my 8800 gts and wait for r650....
Yes, there was another point release to the 8.37s today to address 3DMark issues under Vista but no real changes for the drivers you can download tomorrow. The 8.38s are scheduled for 5/23, should see some very nice improvements in games and benchmarks, along with Vista compatibility improvements.Quote:
Originally Posted by deathman20
Looking at that makes you think the GTX is at least twice as fast in everything.Quote:
Originally Posted by fornowagain
It seems that it beats the 320gts in most tests, the 640gts in not as many but still more, and loses in most to the gtx.
Not bad for a card around the 640gts range...
Part of its problem against teh gtx is the amount of memory since it doesnt appear to scale too well at high res and aa/af
Quote:
Originally Posted by Ubermann
Yeah :) but it does have some sense to it....:confused:
Is anandtech going to redo the review next week with the new drivers?Quote:
Originally Posted by bingo13
I want this tit for tat resolved. Lets see which card performs best OC using stock coolers only. That will put the nail on the coffin, I bet muhahahahaahahahahah:cool:
http://homepage.ntlworld.com/gerald.marley/index.htm
Mirror of VR-Zone review.
Let's all camp out in front of the building where the Crysis development staff is and beg them to give us a room filled with hardware to test Crysis on instead.Quote:
Originally Posted by Eastcoasthandle
The price for a 2900XT looks set for around $735 NZD, with a 8800GTX less then 100 bucks more I'm sure you alllllll know what I'm going to buy :rolleyes:
PNG lossless versions.
8800GTX vs 2900XT
http://img50.imageshack.us/img50/4821/gtx1wt7.th.pnghttp://img442.imageshack.us/img442/7563/xt1hd7.th.png
http://img03.picoodle.com/img/img03/...2m_5707e55.pnghttp://img02.picoodle.com/img/img02/...2m_b16ef13.png
All four frames in a zip file here.
http://www.sharebigfile.com/file/168282/pics-zip.html
Pre-orders for the 2900XT are only 20 euro above that of the 8800 GTS 640 MB here.Quote:
Originally Posted by Timmay
:D :D :D :D :D :D :DQuote:
Originally Posted by fornowagain
The me the 2900 wins. The very slight blur in the lines on the balloon is acceptable, the overall jaggie-removal is better, but what really sets it apart from the nVidia is that the nVidia totally screws up on the front-end of the balloon where the spike is sticking out. AntiAliasing should never remove lines.
I think it looks pretty good, but I want to see performance tests in Rainbow6 Vegas (Unreal Engine 3). I'm also disappointed to hear that the shim is taller than the core, the main reason I was waiting for R600 is cause I don't like the G80's huge IHS and prefer the traditional approach. Not to mention I'd probably need a new PSU if I wanted to OC the R600.
what?? XT?Quote:
Originally Posted by Timmay
so who won ? XT?Quote:
Originally Posted by fornowagain
That's up to you. I think the difference now is tiny. Congrats to ATi.Quote:
Originally Posted by theteamaqua
The OC results show that the architecture itself has a lot of headroom when you can supply it with enough power and remove the heat. Think the 65nm die-shrink will show what the card can actually do?
Also, can somebody point me towards some info explaining what the fill-rate/pixel shader/vertex shader results mean in practical terms? (I mean performance in games, not the definition of each metric.) It seems odd to me that the results in each test swing so wildly, especially with unified shaders.