• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD To Launch RV770 On June 18th

Has anyone noticed how well the 4850 seems to scale when clocked?

at XS jimmy raised the core clock only 75mhz (625->700) and the memory just 107mhz (993-1100) and got over a 1000pt increase in 3dmark06.

Gives us a good idea of how the 4870 will perform as it seems that it's the same specs but running at higher speeds.

Sorry if this was posted already.

54053dc3-8fed-4435-9bb4-ac970385ecfb.jpg
 
Yeah, seems the unlinked shader clock was a load.

Also, if PCZilla's ramblings about more than 800 shaders are correct, and the earlier claims of the 4800 series being 'clocked with power consumption in mind'... Could it be possible that AMD might release a 4890 'RV770 XTX' at some point?

Edit: also, it seems quite likely that the 4850 is quite memory bandwidth limited, if that were the case, the increase in memory bandwidth could almost linearly scale with performance until it starts providing enough bandwidth for the RV770 core to really stretch its legs. The bad news is that it means the 4850 may never quite reach 4870 real-world speeds, the good news is that means the 4870 could be a fair bit faster.
 
Last edited:
Also, if PCZilla's ramblings about more than 800 shaders are correct, and the earlier claims of the 4800 series being 'clocked with power consumption in mind'... Could it be possible that AMD might release a 4890 'RV770 XTX' at some point?

Not until yields significantly improve, and the 55nm process matures.

The 'more than 800 shaders with some disabled' is to allow higher yields. For example: Nvidia's GT200 GPU has 256 stream processors, arranged in 16 groups of 16. The 240-SP (ie 15*16) layout allows for one block of 16 to be faulty without affecting the ability of the GPU to run as a GTX280. Up to 4 can be faulty whilst still running as a GTX260 (192 SP = 12*16).

The same is true with the 4800-series. If a 100% success rate was required for operation, with these 1Bn+ transistor GPUs, then yeilds would be poor indeed. While AMDs core is smaller (talking in terms of transistor count here), it is also produced on a smaller and less reliable process (55nm).

But yes - maybe in 6 months or so when perfect SP-block yields can approach 50% or so of the production runs we may see an 'XTX' type card, with 5% or so extra shader pipes. Similarly I'm expecting a 256-SP 'Ultra' from Nvidia at some point.



edit - regarding the limitations of the 4850 and 4870: Yes, the memory bandwidth will hurt the 4850 and I'm expecting the 4870 to be notably faster (similar in proportion to the difference between the 260 and 280). However I still feel that it's fill-rate that's holding back both cards, and the architecture in general.
 
Last edited:
Although it can be foolish to base opinions on rumours, for the first time in a long while I'm considering an upgrade followed by another shortly after - 4870 to 4870X2 - if things scale as well as expected.
 
The ruby demo looks amazing, pity we cant have games looking like that and running well.
R700 ? has to be a 4870 X2, if it is a single GPU card this could get interesing :)
 
Back
Top Bottom