AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

I'm looking at the Skyrim graph, and I'm 100% sure there aren't enough ROPs. Reports are that later this year well see 4k monitors @ 30-36", and 2880 @ 17". Giving Tahiti only 32ROPs is a colossal mistake. With the rumoured price, the people on 1080p res probably won't go for this card, and we're already seeing graphs where it's no faster than a 580 @ 2560. Failure all round.

But it's OK! WINZIP is accelerated. AMD, pull your head out your arse plz.

No one will be gaming at that resolution with just one card. Especially people with budgets for that kind of set up. As for 2560, I have not seen anything leaked that suggest it's no faster than the 580 at that res? I also don't get the "people on 1080p" won't go for this card... That is pretty loony.
 
So.....how long do you guys think it will take for AMD to release 1.5GB versions of these babies (if ever)?

What price drop could we expect from a 7970 3GB to a 7970 1.5GB?

Also will there be 7950s at 1.5GB from the beginning, since it is said that AMD unleashes the AIB partners choices with this one?

thanks

GTX 580 1.5GB 490$ vs 3.0GB 590$
since 7970/50 use faster memories, price difference may be higher..
 
Also, pixel fillrate is a direct proportion of core clocks.
Of course, it is.
These sort of numbers are half useless. They just refer to individual blocks. They don't tell you how these units are wired or how/when they are used or how efficient these units are. And these are the kind of things that make an architecture.

So, correct me if I'm wrong, but it looks like they bought no performance improvement with GCN (if you look at a pure TFLOPS metric which isn't a direct correlation to game performance I know) because the improvement in FLOPS is a direct relationship between the SP increase and clock increase.
That's how these numbers are calculated in the first place.

So, they could have achieved the same thing by shrinking Cayman and scaling the architecture up theoretically.
The real number to look out for is bandwidth. Tahiti has 50% more bw than Cayman, and if the performance increases by >50% vs Cayman, you can be sure that a dumb scaling of Cayman wouldn't have cut it.

I guess the takeaway is that GCN in itself isn't revolutionary in terms of game performance and this iteration is more about the die shrink, added features, lower idle power and directx 11.1 compliance?
It's a pretty good evolution. These days, you are not going to get >10% improvement unless you shrink.
 
GTX 580 1.5GB 490$ vs 3.0GB 590$
since 7970/50 use faster memories, price difference may be higher..

That's some good thinking! Thanks to you and the rest of the guys as well!

So a 7950 1.5GB could be very well priced and with some OC it could hit 7970 performance, with significantly less money. That is for people like me, that will not need the 3GB versions anyway.

Things just got interesting regarding the 79XX for me! :)
 
Yeah, 4k monitors are ludicrously expensive, and will continue to be so for years. Nothing to get panties in a bind over.
 
The real number to look out for is bandwidth. Tahiti has 50% more bw than Cayman, and if the performance increases by >50% vs Cayman, you can be sure that a dumb scaling of Cayman wouldn't have cut it.

That's a very good point. And some of the leaked slides show +60% over the GTX 580, so probably at least as much over the HD 6970.
 
Of course, it is.
These sort of numbers are half useless. They just refer to individual blocks. They don't tell you how these units are wired or how/when they are used or how efficient these units are. And these are the kind of things that make an architecture.

That's how these numbers are calculated in the first place.

The real number to look out for is bandwidth. Tahiti has 50% more bw than Cayman, and if the performance increases by >50% vs Cayman, you can be sure that a dumb scaling of Cayman wouldn't have cut it.

It's a pretty good evolution. These days, you are not going to get >10% improvement unless you shrink.

Yeah, you can tell I'm rather GPU architecture ignorant. :oops: Thanks to you guys, I'm learning. I knew if I put those numbers out there you guys would give me context and help me understand how GCN is better.
 
GTX 580 1.5GB 490$ vs 3.0GB 590$
since 7970/50 use faster memories, price difference may be higher..
Yes, but OTOH the 3GB GTX 580 is more of a "special" part whereas the 3GB 7950 is a "normal" part, which could make the difference smaller. If you look at instance for the difference between 1 and 2 GB HD 6950 it is quite small. Based on that I'd guess the difference won't be more than 50$.
 
That's a very good point. And some of the leaked slides show +60% over the GTX 580, so probably at least as much over the HD 6970.

I would wager that it is about 50% faster than Cayman, unless independent benchmarks show up. (Leaked slides are showing 60%, so it's more like 40% in reality and about 10% between Cayman and gf110).
 
They don't tell you how these units are wired or how/when they are used or how efficient these units are. And these are the kind of things that make an architecture.

I agree. FLOPS for comparison are one aspect, another would be branch-efficiency which we know went up. No problem being twice as fast with the same FLOPS. As such I can't really understand the appeal in the number-orgy the previous poster made. We're not in the age of embarassing parallel fruits anymore.
 
That's some good thinking! Thanks to you and the rest of the guys as well!

So a 7950 1.5GB could be very well priced and with some OC it could hit 7970 performance, with significantly less money. That is for people like me, that will not need the 3GB versions anyway.

Things just got interesting regarding the 79XX for me! :)

Why would you not need 3GB? BF3 already uses 1.4GB of VRAM according to Dice at highest settings 1080P (they blamed the stuttering problems some saw on VRAM shortage). Crysis 2 hi res pack recommends 2GB card. For the future of course this will only increase no matter what resolution you run.

Too me that's one of the Nvidia lineup's biggest problems. GTX570's with 2.5GB VRAM are prohibitively expensive, I've seen reviews love the 560 Ti 448 with it's 1280 MB, too me a $300 graphics card needs to be more future proofed than 1280 MB...
 
Why would you not need 3GB? BF3 already uses 1.4GB of VRAM according to Dice at highest settings 1080P (they blamed the stuttering problems some saw on VRAM shortage). Crysis 2 hi res pack recommends 2GB card. For the future of course this will only increase no matter what resolution you run.

Too me that's one of the Nvidia lineup's biggest problems. GTX570's with 2.5GB VRAM are prohibitively expensive, I've seen reviews love the 560 Ti 448 with it's 1280 MB, too me a $300 graphics card needs to be more future proofed than 1280 MB...

I still see people recommending cards with 1GB VRAM only. It usually takes some convincing for them to realize 2GB is the sweet spot right now.
 
I would wager that it is about 50% faster than Cayman, unless independent benchmarks show up. (Leaked slides are showing 60%, so it's more like 40% in reality and about 10% between Cayman and gf110).

Oh I'm not expecting the average to be anywhere near 60%, but I don't think AMD would make things up either, so these cases probably do exist. And they're clear evidence of efficiency improvements, at least for these specific games.
 
Will having 3Gigs of vram increase Hardware Reserve on 32bit OS?

yea i think so, although at least in win 7 you can unclick the maximum memory checkbox in msconfig and it dynamically reserves ram as necessary.

but im pretty sure whatevers on the graphics memory has to be duplicated in ram. it would be interesting to see what happens anyway, i guess everything would end up in the page file?
 
Oh I'm not expecting the average to be anywhere near 60%, but I don't think AMD would make things up either, so these cases probably do exist. And they're clear evidence of efficiency improvements, at least for these specific games.
But they can always cherry pick them. :???:
 
No one will be gaming at that resolution with just one card. Especially people with budgets for that kind of set up. As for 2560, I have not seen anything leaked that suggest it's no faster than the 580 at that res? I also don't get the "people on 1080p" won't go for this card... That is pretty loony.

I'm perfectly fine with 1080p and 1 card.
 
72018098.jpg
 
Back
Top