News-Official AMD R7 & R9 Naming Unveiled

[tolis626]

Something tells me the dual GPU's won't be connected in the usual fashion either.

Amen to that!If there is a dual-GPU without the issues of Crossfire (Mostly) or SLI,I'll have one the moment it's on the market.

 

[Keysplayr]

Something tells me the dual GPU's won't be connected in the usual fashion either.

Please tell me they're not using chewing gum.

 

[Face2Face]

Please tell me they're not using chewing gum.

Big League chew... and lot's of it... the grape kind

 

[Keysplayr]

So, any guesses? Big Huge die or an X2 for top end?

 

[Face2Face]

So, any guesses? Big Huge die or an X2 for top end?

I am going to guess they go with the same setup they currently have.
R9 280X X 2 for the top end and still use a small die. I will guess that GCN 2.0 will be a small improvement over GCN as far as efficiency and performance goes. They will increase the core count to the 2300-2500 range for the 7970's replacement (R9 280X). It should fall right in-between the GTX 780 and Titan, and have a core clock of around 1GHz and have a TDP of 250-275w. I am not expecting a godly card here, I think it's rumored pricing gives it away (Under $600) and the fact it's more than likely on a 28nm process.

 

[The Alias]

the 7790 had 40% more shaders and 33% more bandwidth which equated to ~28% more performance with a ~13% increase in wattage . What makes you think AMD can't do that to the 7970ghz going from not only gcn 10 to gcn 1.1, but gcn 2.0 when they have had almost 2 whole years to design this chip ?

Note this as well, the 7790 had significantly higher clocked memory; so it's actually highly possible that the Bonaire core may have been consuming less energy than the Cape Verde one .

Amd should absolutely demolish the 780/Titan with their new release . They had the ability to with the regular gcn 1.0 . All they had to do was increase the sp/rop/tmu count and decrease the voltage to keep power consumption in check , because the increase in wattage from the 7970 to 7970ghz is almost purely because of the increase from 1.175 to 1.250 that was ENTIRELY unneeded . But AMD didn't do that; instead, they decided to wait and continue developing their next architecture to let that combat the 780/titan and I don't see any reason why they'd wait that long unless their new cards were going to demolish the competition .

my numbers came from this tpu review : http://www.techpowerup.com/reviews/Powercolor/HD_7790_Turbo_Duo/1.html

 

[boxleitnerb]

Demolish Titan? No way, AMD likely uses the same process and they are bound by the laws of physics, too. At best they can be maybe 10% faster at the same power consumption. But I think we'll be looking at a parity when it comes to perf/W.

 

[The Alias]

Demolish Titan? No way, AMD likely uses the same process and they are bound by the laws of physics, too. At best they can be maybe 10% faster at the same power consumption. But I think we'll be looking at a parity when it comes to perf/W.

wattage is really not the concern as they have ~55 watts to play with coming from the 7970 . Performance per clock, however, is the thing that the next card will either live or die from because the Titan is ~40% more powerful per clock (kind of hard to measure because of the boost thing) and the 780 is ~35% more powerful per clock . So the major thing AMD should be shooting for with Hawaii is being at least 10% than the Titan at the same clocks, so when it comes down to the nitty gritty when both cards are oc'ed, Hawaii will come out on top .

 

[boxleitnerb]

Perf/clock is irrelevant. You cannot compare different architectures on this. And what does it matter how performance is generated? Perf/W is all that matters in the end since it is power and thermals that are the ultimate limiting factor.

 

[rgallant]

Perf/$$$ I think is the over riding metric in video cards, with all other things = like IQ ,driver
-if amd wanted it I think nv offerings after 2 yrs down the road could be walked over if amd wanted to do that [build it] , in the past they wouldn't ,but what about 2013 ?
GPU'S are a easy target for mind share [6 bil trans] vs intel their 2 nodes behind in cpu's.

 

[The Alias]

Perf/clock is irrelevant. You cannot compare different architectures on this. And what does it matter how performance is generated? Perf/W is all that matters in the end since it is power and thermals that are the ultimate limiting factor.

Actually you can . The only reason we haven't is because of the different clock limits of the respective architectures, but now is a different case because qboth gcn and kepler seem to have the same overclocking limits so now perf/clock is now the reigning comparison point for chip comparison

 

[boxleitnerb]

Actually you can . The only reason we haven't is because of the different clock limits of the respective architectures, but now is a different case because qboth gcn and kepler seem to have the same overclocking limits so now perf/clock is now the reigning comparison point for chip comparison

It still does not matter. Clocks determine performance, but power consumption determines clocks. What people do in regard to overclocking and how far they are willing to go (efficiency, mod bios, alternative cooling) is another matter altogether.
The only valid argument would be that Hawaii will have custom models allowed while Titan does not. That is a seizable advantage.

 

[The Alias]

It still does not matter. Clocks determine performance, but power consumption determines clocks. What people do in regard to overclocking and how far they are willing to go (efficiency, mod bios, alternative cooling) is another matter altogether.
The only valid argument would be that Hawaii will have custom models allowed while Titan does not. That is a seizable advantage.

the overclocking is what I'm trying to account for . If Hawaii beats the titan but is clocked so high that the 780 still surpasses it when both are overclocked, that's all that we'll hear because it's Nvidia (remember the 460 ? compared to the 7970 vs 680 after 12.7 ?) . But if Hawaii beats the Titan at stock with clock closer to that 925 and still be able to clock in the 1300's on air, it would be a clear cut win for AMD then .

 

[The Alias]

Perf/$$$ I think is the over riding metric in video cards, with all other things = like IQ ,driver
.

I wish, but history has proven that whoever has the performance lead ends up leading in sales

 

[rgallant]

I wish, but history has proven that whoever has the performance lead ends up leading in sales

I agree
they need hardware not free games.

 

[Keysplayr]

the 7790 had 40% more shaders and 33% more bandwidth which equated to ~28% more performance with a ~13% increase in wattage . What makes you think AMD can't do that to the 7970ghz going from not only gcn 10 to gcn 1.1, but gcn 2.0 when they have had almost 2 whole years to design this chip ?

Because they're still on 28nm.

 

[The Alias]

Because they're still on 28nm.

They were still on 28nm going from the 7770 to the 7790 . They did that from architectural improvements alone

 

[Keysplayr]

They were still on 28nm going from the 7770 to the 7790 . They did that from architectural improvements alone

Different. The 7790 is probably relatively easy on 28nm. The 7970 is already quite large on 28nm and there is not a lot of room left. It's really not a good comparison.

 

[The Alias]

Different. The 7790 is probably relatively easy on 28nm. The 7970 is already quite large on 28nm and there is not a lot of room left. It's really not a good comparison.

Room left for what ? Last time I checked there wasnt really a limit on how big chips can get

 

[Keysplayr]

Room left for what ? Last time I checked there wasnt really a limit on how big chips can get

Ohhhhh... I see.

 

[Elfear]

Different. The 7790 is probably relatively easy on 28nm. The 7970 is already quite large on 28nm and there is not a lot of room left. It's really not a good comparison.

Isn't Titan at 551mm^2? With Tahiti at 352mm^2 it looks like they have a lot of room to grow.

 

[Zodiark1593]

I agree
they need hardware not free games.

AMD had both (the 7970 traded blows with the GTX 680), and Nvidia still had the lead, even when AMD dropped the price several times.

Though I do have to agree to an extent, people who buy a GPU just want the hardware first and foremost.

 

[Zanovar]

judging by your interest you might be thinking about one of these cards Keys?*coughs*

 

[raghu78]

Different. The 7790 is probably relatively easy on 28nm. The 7970 is already quite large on 28nm and there is not a lot of room left. It's really not a good comparison.

there is not a lot of room left in your opinion. the fact is AMD has atleast another 120 - 130 sq mm and they are still below 500 sq mm. thats quite a bit of space to add 30 - 40% more sp remember sp are very dense. the question is are they on 384 bit or 512 bit memory bus. i still think 512 is overkill and would impact perf/watt. but maybe AMD has other ideas. hawaii is big and you will get confirmation soon.

 

[3DVagabond]

Perf/clock is irrelevant. You cannot compare different architectures on this. And what does it matter how performance is generated? Perf/W is all that matters in the end since it is power and thermals that are the ultimate limiting factor.

+1
Agreed perf/W is ultimately the limiting factor.