AMD: R7xx Speculation

"They" are biased ati fanboys who pull benchmarks out of their hat just to screw with peoples minds
Ill be first in line for a 4870x2 if it blows the 280 to hell (which i hope happens so nvidia will move to gddr5 among other things otherwise nvidia will continue to be uninnovative and stomp all over the consumer with overpriced, slightly increased performance crap every six months just look at how the 1k series destroyed the "7" series and forced nv to pull the unified shader card next round and look what that brought us

 

hey,razor1 do you argee with him?

 

But is the GTX 280 limited by memory bandwidth?

 

no but why not have the extra umph

 

i don't need his approval anyway i was joking

 

Because adding more memory bandwidth to something that isn't limited by memory bandwidth won't do much, if anything.

 

they could shorten the bus to 256bit and that would increase yields which in turn lowers cost

 

Yeah, but that's not what you implied originally.

 

Yes i know sometimes the way i think doesn't get translated to my posts 512bit gddr5 would be overkill i agree its just gddr5 is such a huge leap so efficient uses less power while churning out twice the clock speeds of gddr3/4 it makes sense to use it.
Btw how is the latency on gddr5?, like how would 4ghz gddr5 actually compare to 2 ghz gddr3
would the latency really affect it?, or would it end up being about twice as fast.

 

What latency? Are you mistaking the higher latency of GDDR4 for GDDR5?

 

If G92 is, GT200 is limited by bandwidth at ~2xG92. Of course, it will perform better than 2x in cases where the framebuffer is the limitation (über resilutions with AA). It would be interesting to compare it to a SLI of 1 Gbyte 9800GTXs .

 

Possibly i imagined that gddr5 would have a higher latency than gddr3 really i don't know that much about it....

 

New GPU-Z is out now: http://www.techpowerup.com/downloads/1104/TechPowerUp_GPU-Z_v0.2.4.html

 

What part of G92 is limited by bandwidth? TF? Z? B/c not everything has been doubled from G80, and I was under the impression that ALUs aren't the main consumers of bandwidth.

 

It depends obviously by the application , but if we take a look at some comparative scores regarding 9600GT, 8800GT, 9800GTX, we see that on a +100% scaling in the shader/TMU department, we observe only a 15-30% tipically more performance. This happens also at resolutions where fill rate should be not a limiting factor, neither CPU should be. And I don't think it makes sense to speak about " a part" of the chip being limited, as the chip works as a sum of all parts.
And I said "if"

TMU are not doubled in GT200 AFAIK , but the texturing power of G92 was already overwhelming.

And to have double the power of i.e. a G92 part, you have to throw in double of all (shaders, bandwidth, ROPS). Now, GT200 is NOT 2x G92 in term of texture fill rate (+25%), pixel fill rate (+78%) whereas is +180% in shader power if we don't count the MUL in G92 (otherwise it will be a 67% more). It could be better in Z fill rate, but we don't know. Now, if this gets results in the 2x G92 range, it will be clearly mean that G92 is bandwidth limited (more than 2X should only indicate a framebuffer limitation in G92, as we can see in 9800GX2 beaten in some particular tests by a single 8800 Ultra).

 

True. I didn't mean to nitpick, just to point out that while GT200 has 2x the bandwidth of G92, it generally has less than 2x as much of everything else, so it may be more accurate to say that if (when) G92 is b/w limited, GT200 may also be, but to a lesser extent.

But I don't know how a GPU functions at a low level (specifically, texture caching), so I'm probably wrong or oversimplifying when thinking that if G92's TF units are bandwidth limited and b/c GT200 gains +25% TFs but +155% b/w, then its TF units aren't (as b/w limited). I'm probably forgetting/ignoring that ROPs are the main b/w consumers, and that GT200's increased ROP count tracks closer (than TF) to its increased b/w. So, if ROPs are the main b/w consumer, then GT200 is pretty close to ~2x G92, as you said.

For instance, G92's texture power is indeed "overwhelming," but does it overwhelm available bandwidth or are you just saying it's wasted relative to the power/speed of the rest of the chip? I was guessing that ALUs don't require too much bandwidth, and are therefore not bandwidth "limited," b/c something like 3DMark's Perlin Noise test shows the 8800GT is faster than the 9600GT by the exact percentage of its theoretical FLOPS advantage: 62%. And if GT200 is scoring 300 while G92GT's scoring 155, then we're also seeing an improvement that tracks about 1:1 with the increase in ALUs (if not with FLOPS, but I don't know if the GT200 is better at flogging that extra MUL, or if this specific test will even give it the chance).

Yeah, a GPU is the sum of its parts, but the word bottleneck exists for a reason, doesn't it?

 

Fud is saying that the 4870 is sampling this week. Is that even possible to release a card so soon if it is currently sampling?
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7908&Itemid=1

Also they are saying that it isn't much slower then the GTX280, but how can you believe that when they keep misquoting everything?
http://www.fudzilla.com/index.php?option=com_content&task=view&id=7916&Itemid=1
They say its a GX2 9800 when it clearly says in the link that it is a 9800gtx oc

 

The SP number of RV770 is not 800

We learnt from AMD that the number of SP inside RV770 isn't 800. Actually, this number is more than 800sp, but AMD hasn't told us the exact number.

But you can image 850 or 840 sp are actually put inside the RV770 chip, AMD only opens 800sp to consumers, the left sp have been shielded.

We hope someone or some software could open the shielded sp for us.

Anyway AMD will shown a live demo of RV770 with a frog as the protagonist.

source
http://www.pczilla.net/en/post/45.html

 

So Arun was wrong! No cookie!

<edit>and here's a new GPU-Z shot

 

If they are counting fine grained redundancy.... 850 . Someone needs to inform the author that learnt is archaic.