The NEXT LAST R600 Rumours & Speculation Thread

Why is no one committing about the IQ?

 

Cause its hard to see thru smoke and mirrors?!?

Those that might be putting up legit reviews will wait til the NDA is up and put up full fledged reviews they can take credit for.

 

http://bp2.blogger.com/_BabjUDZIqPw/RjIc5GGObEI/AAAAAAAAACM/NsWoRGaAG-0/s1600-h/013.jpg


16 texture units = Grown-Man trying to ride /accelerate using pedals on baby bicycle
Oh! come on....

-Or-
16 texture units = You have a huge engine but trying to fit in a small car. Good Luck with that

I hope ATI/AMD do realize that in order to boost FPS you need 32+ texture units and 20 or 24+ ROP's "Just hope that 65nm R650 will be a big surprise for me"

320 stream processor units running on lower MHz Frequency, I rather have 128 stream processor units but running at high Frequency. It will be very hard to utilize all 320 stream running low MHz frequency.
But I still like ATI.

 

Care to explain why it's harder?

 

If you have a very big pipe but the water flow is low or not enough in order to push the water flow to the max, creating pressure that pores the water out in/at the great speed that it will wash you away if you stand in its way.

Big pipe = 320 stream processor units
Water = MHz

 

hmm doesn't really work that way

 

But 3dmark03 is known to be very little influenced by CPU.

 

But you're still moving the same volumn of water!

Do not think of the speed of the water, think in terms of the volumn.

 

Different comparison: but: comparison R600 = 7 lanes on freeway traffic driving at 50MPH Vs. G80 3 lanes traffic on freeway driving at 120MPH.

 

work done is what you want to see at the end at a given frequency at a given time.

 

But the original question was to utilize 100% 320 stream processors units, which will never happened!

Just like R580 w/ 48 pixel shader processors.

 

You've broken the secret code of chip design!

 

Think your confusing a bottleneck with ultilization.

 

Please don't make fun of me, I'm trying to find/use the best English words to my knowledge.

 

It is not your english that is amusing, trust me.
Your analogy about freeway lanes doesn't quite cut it either. What you have described is more like the difference in latency for a single instruction, and even then, you are making the big assumption that the pipeline depth (the length of the freeway) is the same for both.

 

Which modern games are heavily texture dependent? These would probably be the best to show any defenencies due to having only 16 texture units. Hopefully, Rys will have that under control though.

In addition, is there any analysis that shows at what resolutions 16 ROPs may become a bottleneck?

 

Disappointed, huh?

The bizarre thing is that in the same time it can do a vec4 MAD (vector or scalar makes no difference) it can do any of the following:

• 2 scalar MULs
• 2 vec3 MULs
• 1.6 Vec4 MULs
• 4 scalar ADDs
• 3.2 vec3 ADDs
• 2.5 vec4 ADDs

But not knowing how many texture fetches there are and other such quibbly details, I dunno if these are worth much investigation. The GPUBench guys had to keep fixing their code in order to get meaningful results...

Jawed

 

Granted, I don't know how the "320 Stream Processors" relate to the 128 Nvidia utilizes, but I'd assume it would require more transistors to schedule and keep more parallel stream processors full (since ATI probably doesn't use scalar ALUs like Nvidia does it's difficult to make a correct comparison and there is a bunch of hand waving and assuming here ). Thus, less parallel processors could require less transistor overhead and allow that space to be utilized for other more productive uses.

I'd also assume though, that ATI engineers would do a complete analysis on such usage before starting, so I'm not sure how relevent this is in reality.

 

I would like to say sorry to everyone on this forum for my comparison, it is well possibly my message is not accurate and correct because of my lack understanding of the GPU design.

 

lol, It's ok, at least you are brave enough to try. I have no idea what some of you guys are talking about half the time.