8800GT Final Specs Confirmed !! Information Arrives on Retailer Websites !!

[theholmboy]

http://www.buy.com/prod/geforce-xlr8...206166489.html

Personally, I'll wait until the 'egg has it.

[RAW-Raptor22]

You're wrong. Texture filter units = TMUs.

Texture Mapping Units (TMU) or Texture Filtering (TF) units.
Source : http://www.hardwarezone.com/articles...?cid=3&id=2263

2900XT has 16 TMUs OK ?




Source : http://www.xbitlabs.com/articles/vid...tecture_5.html


Source : http://www.hothardware.com/articles/...houses/?page=2

Texture Fill Rate = TMUs x Core Clock Do you agree with me ?

8800GTX : 36.8GT/s : 575MHz = 64 TMUs.

I can easily match those stock GTX numbers with my GTS...

[LordEC911]

You're wrong. Texture filter units = TMUs.

Texture Mapping Units (TMU) or Texture Filtering (TF) units.
Source : http://www.hardwarezone.com/articles...?cid=3&id=2263

2900XT has 16 TMUs OK ?




Source : http://www.xbitlabs.com/articles/vid...tecture_5.html


Source : http://www.hothardware.com/articles/...houses/?page=2

Texture Fill Rate = TMUs x Core Clock Do you agree with me ?

8800GTX : 36.8GT/s : 575MHz = 64 TMUs.

Then why does the GTX only have 32TMUs and NOT 64 like you are trying to tell everyone?
Maybe because TMUs |= TFUs?

Theorectically the texture fillrate for the GTX should be 18400MTexels/sec.

Each cluster also has 4 pixels per clock of address ability, but 8 INT8 (FP16 half speed) bilerps per clock of fetch and filtering, in order to get data out of VRAM or cache into the core (for any thread type of course) for working on. Therefore the hardware will give you two bilerps per pixel address, for 'free' (and proven with in-house codes up to and including FP32 per channel), rather than the usual one. So that's 32 pixels per clock (ppc) of INT8 bilinear 2xAF, or 32ppc of FP16 bilinear, or 16ppc of FP16 2xAF, or 16ppc of FP32 bilinear per cycle, out of the combined texture hardware across the chip, giving you a few rates to ponder as we talk about the rest. We'll get back to fetch and filter performance later in the article.

http://www.beyond3d.com/content/reviews/1/8

[AnarchX]

Theorectically the texture fillrate for the GTX should be 18400MTexels/sec.

That is right, but the speciality is here, that this fill-rate is preserved when you activate trilinear filtering or 2x bi-AF and not like on the most other chips, who halve in this cases.

TMU = Textur-Adresser + Textur-Filterer

[bakalu]

I can easily match those stock GTX numbers with my GTS...

Thanks you.

8800GTS has 48 TMUs or 48 texture filter units

GPUz 0.0.8 8800GTS OC

GPU Clock = 692MHz, and Texture Fill Rate = 33.2 GTexel/s

TMUs of 8800 GTS = 33.2 : 692 = 48 TMUs.

Then why does the GTX only have 32TMUs and NOT 64 like you are trying to tell everyone?
Maybe because TMUs |= TFUs?

Theorectically the texture fillrate for the GTX should be 18400MTexels/sec.


http://www.beyond3d.com/content/reviews/1/8

You're wrong. Texture filter units = TMUs.

Radeon X1650 XT has 8 TMUs OK ?

The Radeon HD 2600 has only 8 texture filter units: more than in the Radeon X1600, but just as many as in the Radeon X1650 XT.

Source : http://www.xbitlabs.com/articles/vid...tecture_5.html

The texture fillrate for the GTX is 36.800 MTexels/sec.


Source : http://www.nvidia.com/page/geforce8.html

Texture Fill Rate = TMUs x Core Clock

8800GTX, 36.8 : 575 = 64 TMUs

[AnarchX]

Guys, just run a multi-texture fill-rate-benchmark and you will see that your G80 does not have 64 or 48 TMUs.

The maximum output is limited by the 32/24 Textur-Adresses!

So better speak about, 32/24 TMUs with free trilinear/2xbi-AF, that better describes it.

btw.
This is also a reason, why NV added on G84/86 Texture-Adresses, because in this segment bilinear/optimized-tri Performance is more demanded, than free HQ-filtering.

[bakalu]

Guys, just run a multi-texture fill-rate-benchmark and you will see that your G80 does not have 64 or 48 TMUs.

The maximum output is limited by the 32/24 Textur-Adresses!

So better speak about, 32/24 TMUs with free trilinear/2xbi-AF, that better describes it.

btw.
This is also a reason, why NV added on G84/86 Texture-Adresses, because in this segment bilinear/optimized-tri Performance is more demanded, than free HQ-filtering.

Can I ask you some questions ?

- How many Texture address Units does Radeon HD 2900 XT have ?

- How many Texture filter Units does Radeon HD 2900 XT have ?

- How many TMUs does Radeon HD 2900 XT have ?

[Shintai]

2900XT is 16 TMU, 16 ROP.

8800GTX is 32 TMU, 24 ROP.

[LordEC911]

Thanks you.

8800GTS has 48 TMUs or 48 texture filter units

GPUz 0.0.8 8800GTS OC

GPU Clock = 692MHz, and Texture Fill Rate = 33.2 GTexel/s

TMUs of 8800 GTS = 33.2 : 692 = 48 TMUs.

You're wrong. Texture filter units = TMUs.

Radeon X1650 XT has 8 TMUs OK ?


Source : http://www.xbitlabs.com/articles/vid...tecture_5.html

The texture fillrate for the GTX is 36.800 MTexels/sec.


Source : http://www.nvidia.com/page/geforce8.html

Texture Fill Rate = TMUs x Core Clock

8800GTX, 36.8 : 575 = 64 TMUs

If you don't understand my Beyond3d quote you shouldn't be posting again...

PS- G80 architecture TMUs CANNOT be compared to other architecture's TMUs.

2900XT is 16 TMU, 16 ROP.
8800GTX is 32 TMU, 24 ROP.

Might be a first but even Shintai agrees with me...

[bakalu]

If you don't understand my Beyond3d quote you shouldn't be posting again...

PS- G80 architecture TMUs CANNOT be compared to other architecture's TMUs.

I can't understand your Beyond3d quote.

But you don't understand my xbitabs, hothardware and nvidia quote, do you ?

So you shouldn't be posting again.

[AnarchX]

Can I ask you some questions ?

- How many Texture address Units does Radeon HD 2900 XT have ?

32, you see that is has more than TFs.
But is a feature of R6xx, maybe a bit useless, to accerlate vertex-fetches, for which simple point-samplers avaiable in the R6xx "Texture-Units" (4x20).

- How many Texture filter Units does Radeon HD 2900 XT have ?

16

- How many TMUs does Radeon HD 2900 XT have ?

16

Sure you could speak in G80's case about 64 practical TMUs, because nobody would play without tri or 2xBi-AF filterung on such a chip or?
But the better notation I said above.

[Eddie3dfx]

Why are people jumping the gun and buying the gt. I mean you could at least wait until the day they come out to see what the true benchmarks are lol!
All the benches and info I have seen are all from chinese sites.
Isn't it just a little bit odd that Not one single western review site has benchmarked this card??????
I mean a turkish review company outted the msi card..
Turkish!!!

[Shadowmage]

I can't understand your Beyond3d quote.

But you don't understand my xbitabs, hothardware and nvidia quote, do you ?

So you shouldn't be posting again.

I don't think it's as simple as you're trying to make it out to be.

[Start]

Why are people jumping the gun and buying the gt. I mean you could at least wait until the day they come out to see what the true benchmarks are lol!
All the benches and info I have seen are all from chinese sites.
Isn't it just a little bit odd that Not one single western review site has benchmarked this card??????
I mean a turkish review company outted the msi card..
Turkish!!!

Because the West fails And you are forgetting this is XS...

[perkam]

Why are people jumping the gun and buying the gt. I mean you could at least wait until the day they come out to see what the true benchmarks are lol!
All the benches and info I have seen are all from chinese sites.
Isn't it just a little bit odd that Not one single western review site has benchmarked this card??????
I mean a turkish review company outted the msi card..
Turkish!!!

A. It's Nvidia.
B. It's a 65nm G80.
C. It runs at 600 core with a single slot cooler.

If I see a tsunami coming my way, I'm not going to wait for it to land before taking action

Perkam

[RPGWiZaRD]

If I see a tsunami coming my way, I'm not going to wait for it to land before taking action

Perkam

Well summed up in poetical way.

For me it's never been this clear that we're soon getting a midrange beast released. And my own arguments in short (could easily write a full A4 page or two argumenting about this so I try to stick to a single sentence or two here) on this would be:

1) G92, based on an highend architecture we know works well => G80.

2) Due to being a G80 based and not many changes at all but have a BIG die-shrink (90nm to 65nm!), it speaks very well for that we're most likely going to see a huge overclocking boost here compared to the older brothers. If there's no huge changes in architecture it means none of those "changes" in architecture can become some plausible bottleneck, now the biggest concern is only how far the components used on these cards can go or how much they can take. Also the weak looking single card slot solution NVIDIA designed to use speaks well for overclockability too, indirectly due to lower manufacturing process => lower power consumption => lower heat dissipation.

3) It's focused on shader performance (taking price into account). You all got the specs for 8800GTX GTS and GT, so I won't list the difference here but you quickly realize the focused shader performance comparing these cards to each other taking price into account as well. Future games will only be more and more shader dependant, ie the focus 8800GT has is the right thing to focus on. We're spending money worth spending on.

[iadstudio]

2x GT here I come!!! I can't wait!!!

[Eagleclaw]

I know I am grabbing one ASAP, I hope for EVGA or Foxconn.

Sure glad I waited I was just a few weeks away from getting a 8800gts 320

woot!

edit: was just looking at Foxconns warranty on a 8800GTS on newegg, 1 year warranty only?? Ugg

[mascaras]

From FUD

D8P, Geforce 8800GT doesn’t support DX 10.1

Missed a spot


We already wrote that Nvidia’s latest and greatest D8P/G92 won’t support Tri SLI. It doesn’t have the two SLI connectors required at the top of the card, but more importantly the Geforce 8800GT won’t support DirectX 10.1. We have finally confirmed the original story.

ATI’s contender, the RV670, supports DirectX 10.1 and Shader model 4.1 and we strongly believe that ATI's marketing chaps will use this heavily to differentiate the two products, as well as to gain some advantage points.

Nvidia has already been heavily criticized for its 8800GT and the lack of Tri SLI support and this will definitely limit sales. We are sure that the Nvidia partners won’t be happy to advertise that its cards are DX 10.1 non-supporting.

Nvidia will launch the 8800GT in a week's time, on the 29th of October.


Source

Geforce 8800 GT doesn't do DirectX 10.1

No support for Shader model 4.1

Documents seen by Fudzilla indicates that G92/D8P aka the Geforce 8800 GT is not Shader model 4.1 compatible. It can mean one of two things, one, that Nvidia doesn't want release the information or two, simply that this chip doesn't have support for Shader model 4.1 and DirectX 10.1.


This comes as an interesting surprise as we know that the RV670 aka the Radeon HD 2950 series will support Shader model 4.1 and DirectX 10.1.


We will ask around and try to find out if this is the case, but this would be a big setback for Nvidia, at least when it comes to feature tick boxes on upcoming games.


Source

regards

[gunit]

well if all these leaked benchmarks are even close to being accurate, then the GT with its confirmed 112 SP has in fact retained the setup of the G80 core and will have 28 TMUs (one qaurter of its number of SP, just like G80. GTX:128SP/4=32TMU and GTS 96SP/4=24TMU)

[GoldenTiger]

well if all these leaked benchmarks are even close to being accurate, then the GT with its confirmed 112 SP has in fact retained the setup of the G80 core and will have 28 TMUs (one qaurter of its number of SP, just like G80. GTX:128SP/4=32TMU and GTS 96SP/4=24TMU)

Yea... .

[bakalu]

@AnarchX : One more question

8800 GTS has 96 SP and 20 ROPs.

8800 GT has 112 SP but it has 16 ROPs. Why ?

[AnarchX]

Because ROPs and MC are coupled, they are arranged in ROP-Partitions, each consists of 4 ROPs and a 64Bit Memory-Channel.

SPs are coupled with the TMUs in the Clusters: 4 TMUs and 16 SPs.

But this all relates to G80, maybe in G9x there are some changes...