ATI 4xxx Series Thread

[Keysplayr]

So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

 

[Extelleron]

Originally posted by: keysplayr2003
So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

Yes, that's correct.

 

[BFG10K]

So adding 16 more TMUs. will add nothing to the 4000series

Sure it will. It?ll increase texturing performance which will help things like anisotropic filtering, super-sampled adaptive anti-aliasing and other texturing operations such as multiple render targets.

No NV doesn't have unified shaders.

This is absolutely false.

NV did manage to get MS to remove DX10.1 from the orginal DX10 spec. Because they didn't have the hardware.

Possibly true but it has nothing to do with unified shaders.

 

[BFG10K]

So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

It's looking that way.

 

[SickBeast]

Originally posted by: Nemesis 1

Originally posted by: SickBeast

Originally posted by: allies

Originally posted by: SickBeast

Originally posted by: taltamir
everyone is using older stuff in their architecture... rebuilding from scratch is an insane endeavor that will lead to nowhere fast.

With the R600-based stuff as borked as it is, they should be more drastically revamping the architecture if not starting over. The shader architecture is horribly inefficient, the GPU lacks TMUs, it cannot perform proper AA, plus they wasted a ton of die space making the thing 512-bit unnecessarily.

NV started from scratch after their 5900-series cards and came out with the 6800-series which was much better. The G80 is also a new architecture, and again it performs quite well.

From what I understand, any semiconductor company pretty much has to 'start from scratch' every time there is a major die shrink involved. Usually they make a part, then to a minor die shrink refresh, then make a new part altogether.

By saying they should start over, I obviously don't mean that they should throw away everything they know about GPU design. I simply mean that the parts need to be completely revamped.

The R770 already addresses most of the points you make:

• (rumor)Revamped AA
  2x TMUs
  More shaders (maybe tweaking involve)
  256mbit bus with GDDR5

What more do you want again

I want proper shaders, not 800 crappy ones! :evil:

I suppose that's my main complaint with the recent AMD architectures. Their shader design is terrible. If they're simply adding 500 more crappy and inefficient shaders, I really don't see the point.

It just seems like it has taken AMD a long time to work the bugs out of R600. IMO a completely revamped GPU with better shaders and proper AA would have been the way to go. Hopefully the 4850/70 are just that. I'm not holding my breath. :beer:

This statement has me abit concerned as to whats actually happening here . In this thread.

I suppose that's my main complaint with the recent AMD architectures. Their shader design is terrible. If they're simply adding 500 more crappy and inefficient shaders, I really don't see the point.
So what do you know about the shaders in the new 4000seies of gpu's . If nothing why are you so intent at labling ATI shaders as shiity on the 4000's.
I was almost banned for talking about 10.1 and NV not having the capability to use it.

So in a DX10.1 game would those shitty shaders work better or not. OR are you to against unified shaders. Until we get a game that runs DX10.1 correctly we won't know how bad those ATI shaders are. NV can hold back progress only so long befor it catches up to them . AND it will. But once we get that game we can go back and see how NV held back progress threw the himtbp programm.

I had to look to see if I was in the ATI thread. With all the negitive remarks about a card you know next to nothing about.

I had the 800xtpe 4 months befor almost all others . it was great card. the 4870 isn't going to disappoint anyone. except NV .

Anand himself bashed the shader design of R600 in his review of the 2900XT.

How is NV holding back progress? They've been designing the best high-end GPUs for the past few years. AMD's lack of offerings at the high-end have been holding back performance IMO. I agree that the whole Assassin's Creed DX10.1 fiasco was questionable, but really, that's just a single game out of how many? Did AMD help fund the development of the game? They have every right to do the same thing with other developers. It's just the way things work unfortunately. It's not always for the good of the industry.

I have reason to suspect that AMD will use crappy shaders if they intend to put 800 of them on such a small die space. Also, they've mentionned a whole pile of other 'fixes' they've made to the RV670 core - why not mention that they're using more powerful shaders this time?

Look at the die size of a 9600GT. It has 96 shaders. Do you really think it's feasable to produce a GPU that has 8X the number of that type of shader?

Just because this is the 'AMD thread' it doesn't mean we can't say anything negative about AMD products. I have every right to speculate on how this new card will perform.

Hopefully you're right about the 4870. What makes you think it will perform so well?

 

[ronnn]

Originally posted by: BFG10K

So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

It's looking that way.

Ha, Ha, I love pr battles. Ati says 800, but the nv says 160. Anyways at least twice as many, so lets see how this competes with the 50% better shaders.

 

[taltamir]

NV did manage to get MS to remove DX10.1 from the orginal DX10 spec. Because they didn't have the hardware.

is someone still claiming this?
I did a timeline...
Vista was released with DX10 support the same day as the first G80 DX10 part.
5 months later AMD released their first DX10 part.
6 months after THAT (11 months after DX10 came out) AMD released the first DX10.1 part. And two and a half months later MS released SP1, enabling DX10.1

so if cubic light mapping (DX10.1) was removed from DX10, it was not to help nvidia, it was to release vista 11 months earlier.
It is a BS claim. That originated with nemesis1 who PERSONALLY ADMITTED here on the forums that he has no read it anywhere, but he is GUESSING that is what happens based on his hunches...

 

[Keysplayr]

Bait away ronnn. I'm not bite'n. But I do wish you would grow out of that. It was a question. I didn't know if AMD was still using the very same type of shaders with this new series. Since they substantially increased the shaders from 64 to 160, i think we'll be looking at a well performing card here. Sheer number of shaders should bring performance where it belongs. I thought they were supposed to have 480 shaders originally.

 

[Keysplayr]

Originally posted by: taltamir

NV did manage to get MS to remove DX10.1 from the orginal DX10 spec. Because they didn't have the hardware.

is someone still claiming this?
I did a timeline...
Vista was released with DX10 support the same day as the first G80 DX10 part.
5 months later AMD released their first DX10 part.
6 months after THAT (11 months after DX10 came out) AMD released the first DX10.1 part. And two and a half months later MS released SP1, enabling DX10.1

so if cubic light mapping (DX10.1) was removed from DX10, it was not to help nvidia, it was to release vista 11 months earlier.
It is a BS claim. That originated with nemesis1 who PERSONALLY ADMITTED here on the forums that he has no read it anywhere, but he is GUESSING that is what happens based on his hunches...

Talt. Don't feed that topic again. This particular topic only escalates, and VIRGE locked the last thread about it.

 

[RussianSensation]

Originally posted by: Extelleron

Originally posted by: chewietobbacca
No one is really saying GTX280 is going to be 80-120% faster than the 9800GX2. Most leaks and sources I've heard from say on average 50-60% faster and that can easily be double if the game doesn't scale well w/ SLI/multi-GPU configs and < 30-40% if the game does scale well

People are probably mistaking the nVidia slide comparing GTX 200 performance to the 3870 X2 as comparing performance to the 9800GX2. GTX 280 showed around that much of a performane lead over the GTX 280 in most apps. Most around 80%.

Good catch Extelleron!

 

[Cookie Monster]

Its not really a vec5 shader. Think of this way. One shader packs 5 ALUs, where 4 of them are generic and 1 of them are special. So technically speaking, you cant really say that they are vec5 shaders in terms of functionality but the appearance/layout is strikingly similar to that of a vec5 shader. So people claiming that it has 800 ALUs is correct, and at the same time, those who say it has 160 shaders is also correct.

Im still baffled by the recent rumours though. Guess we will find out soon enough.

 

[SickBeast]

Originally posted by: ronnn

Originally posted by: BFG10K

So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

It's looking that way.

Ha, Ha, I love pr battles. Ati says 800, but the nv says 160. Anyways at least twice as many, so lets see how this competes with the 50% better shaders.

Well, the 280GTX has 240 shaders that are "50% better"...so...equivalent to an 8800GT with 320 shaders. That's double the number of 'proper shaders' that the 4850/70 will have (as far as we know). Time will tell, but like I keep saying, on paper the new NV cards look much, much better than what AMD is coming out with.

 

[allies]

Originally posted by: SickBeast

Originally posted by: ronnn

Originally posted by: BFG10K

So does this mean that 4850 and 4870 are still going to be using the same VEC5 Shaders as the 3 series? But instead of 64 (320) they will have 160 (800)? Is this correct?

It's looking that way.

Ha, Ha, I love pr battles. Ati says 800, but the nv says 160. Anyways at least twice as many, so lets see how this competes with the 50% better shaders.

Well, the 280GTX has 240 shaders that are "50% better"...so...equivalent to an 8800GT with 320 shaders. That's double the number of 'proper shaders' that the 4850/70 will have (as far as we know). Time will tell, but like I keep saying, on paper the new NV cards look much, much better than what AMD is coming out with.

With the new rumor of 800 shaders, AMD cards look pretty awesome. Hopefully they'll pull another 9700pro out of their hat with the R770 chips... 2 weeks.

 

[deepinya]

I love all the arguing over pure speculation. I cant wait for the cards to come out so we can be done with all these "paper wars"

 

[manko]

4850 3DMark Tests - Could be a hoax, but it doesn't seem too far off the mark.

ATI Radeon HD 4850
Q6600 @ 3GHz
3GB RAM

12494 3DMark06 1280x1024 0xAA 0xAF
11240 3DMark06 1600x1200 0xAA 0xAF
10442 3DMark06 1920x1200 0xAA 0xAF
7935 3DMark06 2560x1600 0xAA 0xAF

8487 3DMark06 1280x1024 8xAA 16xAF
6776 3DMark06 1920x1200 8xAA 16xAF
5001 3DMark06 2560x1600 8xAA 16xAF

X2789, GPU:2686, CPU:10322 - 3DMark VANTAGE Extreme PRESETS(1920x1200 4xAA 16xAF)

P6695, GPU:5989, CPU:10356 - 3DMark VANTAGE Performance PRESETS(1280x1024 0xAA 0xAF)

 

[chewietobbacca]

The 3dMark scores are pretty in line with the rumors and waht's been heard

I am wondering about the GPU-Z shot though: it says 480Sp but it also put's fillrate in the 400's and 900's Gpixel/s and Gtexel/s which is 20x greater than any card that currently exists, so either GPU-Z is reading wrong, the pic is a fake, or ATI has some serious magician work going on

 

[ronnn]

Don't think gpu-z actually reads the chip. Isn't just information put in by the gpu-z guy?

No aa or af really doesn't tell us much.

 

[manko]

Originally posted by: ronnn
Isn't just information put in by the gpu-z guy?

Yeah, I think so.

 

[chewietobbacca]

Yeah so the GPU-Z info is pretty much discarded besides the clocks at idle and at use..

 

[Kuzi]

Originally posted by: Cookie Monster
Its not really a vec5 shader. Think of this way. One shader packs 5 ALUs, where 4 of them are generic and 1 of them are special. So technically speaking, you cant really say that they are vec5 shaders in terms of functionality but the appearance/layout is strikingly similar to that of a vec5 shader. So people claiming that it has 800 ALUs is correct, and at the same time, those who say it has 160 shaders is also correct.

Taking this number of ALUs/Shaders, that is a 2.5x increase compared to 3870, and the TMUs at 32, that is a 2x increase. Plus taking higher shader speeds 1000+, and higher memory bandwidth from GDDR5. It's easy to assume in this case that a 4870 can be almost 2.5x faster than a 3870. I wish this is true

Though I still believe it will have 96 shaders (480 ALU's) and 32 TMUs. We'll have to wait and see.

 

[formulav8]

Originally posted by: SickBeast


Look at the die size of a 9600GT. It has 96 shaders. Do you really think it's feasable to produce a GPU that has 8X the number of that type of shader?

The 9600GT is 64 shaders. Its the 9600GSO (also called 8800gs) that has 96 shaders if I remember correctly.



Jason

 

[HurleyBird]

Originally posted by: SickBeast
Well, the 280GTX has 240 shaders that are "50% better"...so...equivalent to an 8800GT with 320 shaders.

It's just as likely that GT200 just has 50% higher shader power in total. I mean, if you took a G92 and gave it 240 shaders but clocked them down to 1300MHz you'd only have ~50% more shading power than a 9800GTX.

 

[chewietobbacca]

I think thats what they meant since that original 50% quote was ambiguous as to whether it was 50% better per shader or 50% improvement over the previous generation as a whole, which I think is more likely with the performance #'s ive been hearing

And ATI shaders for R600 was actually 64 x 5 because each of the 64 ALU's could theoretically perform 5 operations (4 easy 1 fat one) for 320 total (64x5) SP's, which is more marketing than anything else. Moving to actually 160x5 for 800 if 800 is true would be a huge improvement overall.

 

[BFG10K]

Ha, Ha, I love pr battles. Ati says 800, but the nv says 160.

PR or not the fact remains four out of the five units are limited compared to the fifth one, unlike nVdia's SPs that can do everything. This requires a lot more massaging from the driver unlike nVidia's solution which almost dynamically balances itself.

 

[Hauk]

Do modern ATI cards come with a Crossfire bridge? If they don't, is this what's needed?
Some reviewers were saying two are needed to CF two cards?