AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread

Are those prices competitive with nVidia's offerings?

At a first glance, they seem a bit ridiculous to me... One could get a pair of R9 280X for less than a R9 290 and get superior performance than a 290X. Heck, I bet even a crossfire of those ~200€ HD7950 Boost we're seeing around could go over the 290 for 275€ less..

I was kinda hoping to see the R9 290 pushing the prices for the high-end market down in 2014, but I guess I was wrong.
2 cards always gets better performance than 1 for less. the tradeoff is needing 2 cards. 7850 crossfire will beat the 280x too.
 
that benchmark posted by lanek what does uber mode mean ?


I was going to explain what is it , but like i have only something reported by another guy on xs, i dont want give false explanation.

QuietMode: Low fan speed profile, core speed could throttle lightly between stock and Boost clock. ( max clock )
UberMode: More noise, higher fan speed, the core speed is trying to stay at max boost clock ( maybe higher tdp limit ? ).

In both case the max clock speed is allways the same.

I think this was the purpose of the slides we have seen, showing there's only a really little difference on performance between both ( 1-2fps ) ...

Dont blame me if it is absolutely not that on the end.
 
Last edited by a moderator:
I was going to explain what is it , but like i have only something reported by another guy on xs, i dont want give false explanation.

QuietMode: Low fan speed profile, core speed could throttle lightly between stock and Boost clock. ( max clock )
UberMode: More noise, higher fan speed, the core speed is trying to stay at max boost clock ( maybe higher tdp limit ? ).

In both case the max clock speed is allways the same.

I think this was the purpose of the slides we have seen, showing there's only a really little difference on performance between both ( 1-2fps ) ...

Dont blame me if it is absolutely not that on the end.

Well, that depends on the test stand. In open stand, little difference - in a badly ventilated case noticeable difference.
 
Well, that depends on the test stand. In open stand, little difference - in a badly ventilated case noticeable difference.

Thats the case with every card.. take a GPU like the 780 with a temp limit at 80°C ( dont remember the exact number ).. if you have 45°C as Tambient in your case.. good luck for maintain max boost of your gpu every second.
 
Regarding the new bridge-less Crossfire:

- If there's no dedicated connector anymore, does it mean that any card compatible with the new Crossfire setup can have an "unlimited" number of cards?
Could I have for example 4x R7 260X as long as I have them connected to the same PCI-Express host (i.e. all in the CPU's PCI Express channels, not some in the northbridge and others in the CPU).

- Any difference in the way VRAM is handled? Is it still AFR so both cards must have an exact copy of whatever is filling the RAM?
 
Regarding the new bridge-less Crossfire:

- If there's no dedicated connector anymore, does it mean that any card compatible with the new Crossfire setup can have an "unlimited" number of cards?
at least there is no hardware limit. The connector was also rather needed for rendering, you could have previously more cards in your machine and use them individually via opencl.

Could I have for example 4x R7 260X as long as I have them connected to the same PCI-Express host (i.e. all in the CPU's PCI Express channels, not some in the northbridge and others in the CPU).
as long as they are in the same address space you probably can. 4x shouldn't be a problem either, there are 4x crossfire machines already. but going beyond that, it might not be supported (or at least never tested).

- Any difference in the way VRAM is handled? Is it still AFR so both cards must have an exact copy of whatever is filling the RAM?
does not need to be the exact copy, but yeah, you need on one card what is needed for that frame. From HW perspective you could utilize resources from other cards or even main memory, but that way you might be just PCIe limited, two cards being slower than one.

And that's probably also the limiting factor, if you'd need to feed all data into 8 or 16 cards, you'd probably stall feeding them.
 
Embargo lifts 12 hours from now according to this. (13 hours, I keep forgetting I'm in a different timezone just now....)

http://hexus.net/tech/news/graphics/61513-amd-account-manager-reveals-radeon-r9-290x-launch-dates/

dfd482e0-9b2a-4051-8497-4de3e20e3aee.jpg
 
As they have remove the link here's the bench ( some result look strange, so i dont know if they are true , the left colum is only low / medium quality, no AA, not really the interesting one, Bioshock provide really too high number )

The Bioshock numbers for Titan are wrong for sure, the perf difference to R9 290X could of course be explainable by whatever, but the difference to GTX 780 shows it has to be wrong for Titan.
 
Back
Top