AMD Cayman info (or rumor)

Quote:

---

Originally Posted by Jowy Atreides

From 1120 5d in barts to 1920 4d vliw tweaked in cayman alone is a 71% increase.
Factor in the ipc improvement, the much larger framebuffer, the new offload tesselation engine and you have a very feasible 80% bump in speed.

Edit: Don't forget the clockspeed being 860Mhz compared to the 725Mhz of Barts xt.

Think about the logical evidence of the architectural changes shown above and you'll most likely conclude like myself, that the 580 is NOT going to be serious competition.

---

A lot of that is based on unknowns, how will games scale 4D - 5D, how will the launch drivers be, etc. The GTX 580 seems only okay, there's definitely an opportunity for AMD to strike a heavy blow here if they can pull it off/if they are again looking to compete in the high end single gpu segment. I'm hoping they deliver though. :yepp:

20% more shaders + 10% more efficiency, that 30% from 5870 I don’t expect more if clocks are 860MHz

Quote:

---

Originally Posted by Aten-Ra

20% more shaders + 10% more efficiency, that 30% from 5870 I don’t expect more if clocks are 860MHz

---

go ahead its much better to be positively surprised.:up:

one thing though, nobody knows what gains there will be with 4d as this is a completely new approach.

Quote:

---

Originally Posted by Dimitriman

go ahead its much better to be positively surprised.:up:

---

Definitely :up:

Quote:

---

Originally Posted by Dimitriman

one thing though, nobody knows what gains there will be with 4d as this is a completely new approach.

---

i think its funny that they jumped to 5D so early then.

Quote:

---

Originally Posted by Aten-Ra

20% more shaders + 10% more efficiency, that 30% from 5870 I don’t expect more if clocks are 860MHz

---

Quote:

---

Originally Posted by Mechanical Man

There you go wrong, 5D is rarely utilised fully, its easier to utilize 4D fully. So, that will bring lot of improvement.

1600/5 = 320 and 1920/4 = 480, thats 50% increase alone. Now say that Cayman SP is little underpowered compared to 5D SP of evergreen. So say there alon 40% improvement. Ofc it will eat more power if there is no improvements on that side too. Put in there tesselation improvement that will be >150% compared to evergreen. Maybe not 80% more power, but 50-60% more is very realistic approximation.

---

Do some research on AMD arch. Really.

Quote:

---

Originally Posted by Rai

A lot of that is based on unknowns, how will games scale 4D - 5D, how will the launch drivers be, etc. The GTX 580 seems only okay, there's definitely an opportunity for AMD to strike a heavy blow here if they can pull it off/if they are again looking to compete in the high end single gpu segment. I'm hoping they deliver though. :yepp:

---

1 Question.

1. Would AMD switch to vliw4 if there was only negative changes?
No.

So, at worst case, vliw4 is equivalent to Barts in performance.

Would 71% shaders extra to a barts architecture plus double the frame buffer and a new tess engine provide a big performance increase?

There's your cream.

Afaik Evergreen was never starved for shaders. 5870 and 5850 have basically the same perf clock-to-clock. 5830 sucked because it had only 16 ROPs, so it barely beat a 5770 with 40% more shaders. It was easy for Barts to increase the efficiency by increasing the ROP count compared to the amount of shaders. It's basically an OC'd 5830 with double the ROPs. That's the secret to Barts increased perf per mm^2. Now with Cayman the same doesn't apply, so it's up to the shader architecture changes to give similar increases in performance.

Quote:

---

Originally Posted by Pantsu

Afaik Evergreen was never starved for shaders. 5870 and 5850 have basically the same perf clock-to-clock. 5830 sucked because it had only 16 ROPs, so it barely beat a 5770 with 40% more shaders. It was easy for Barts to increase the efficiency by increasing the ROP count compared to the amount of shaders. It's basically an OC'd 5830 with double the ROPs. That's the secret to Barts increased perf per mm^2. Now with Cayman the same doesn't apply, so it's up to the shader architecture changes to give similar increases in performance.

---

i love this explanation of Barts

except that fact you related AMDs best perf per watt gpu, with its worst, lol.

Quote:

---

Originally Posted by Pantsu

Afaik Evergreen was never starved for shaders. 5870 and 5850 have basically the same perf clock-to-clock. 5830 sucked because it had only 16 ROPs, so it barely beat a 5770 with 40% more shaders. It was easy for Barts to increase the efficiency by increasing the ROP count compared to the amount of shaders. It's basically an OC'd 5830 with double the ROPs. That's the secret to Barts increased perf per mm^2. Now with Cayman the same doesn't apply, so it's up to the shader architecture changes to give similar increases in performance.

---

Funny thing is that you can't put Barts efficiency gains on increase in ROP. here is what AT said in its review:

Quote:

---

it’s worth noting that internally AMD was throwing around 2 designs for Barts: a 16 SIMD (1280 SP) 16 ROP design, and a 14 SIMD (1120 SP) 32 ROP design that they ultimately went with. The 14/32 design was faster, but only by 2%. This along with the ease of porting the design from Cypress made it the right choice for AMD, but it also means that Cypress/Barts is not exclusively bound on the shader/texture side or the ROP/raster side.

---

http://www.anandtech.com/show/3987/a...range-market/2

Quote:

---

Originally Posted by Jowy Atreides

From 1120 5d in barts to 1920 4d vliw tweaked in cayman alone is a 71% increase.
Factor in the ipc improvement, the much larger framebuffer, the new offload tesselation engine and you have a very feasible 80% bump in speed.

Edit: Don't forget the clockspeed being 860Mhz compared to the 725Mhz of Barts xt.

Think about the logical evidence of the architectural changes shown above and you'll most likely conclude like myself, that the 580 is NOT going to be serious competition.

---

how come people dont ask you what you smoke when saying such things?

If it beats the 580, it be a good thing due to 40nm, and we cant expect miracles even with the 4d set of shaders.
Overclocking it to 1ghz+ however might turn it into a great card.
If r300 rumours are true, then it has to beat the 480gtx by around 80%
as that was what the 9700 did back in the day.(not going there not even whatever I might smoke)

That depends how AMD did reason when they saw what fermi did, and thought, Gee, the upgrade Nvidia are going to do at 28nm gonna be something, so they did their homewerk and produced a slightly bigger chip to crown the single card performance planned at 32nm but adjusted to 40nm which they might cut things out that wasnt needed this time until 28nm.

The size of the card seen as far suggest a powerhungry one.
See, Nvidia knows it, more power, more speed and more performance and forget to save the whales, we let the oceans boil.

6970 should be 20.3% faster then a 5870, putting it ahead of the 480 a few frames in some games & since i'm on a roll i'll guess 3-10% slower then a 580.

http://vr-zone.com/articles/amd-rade...res/10240.html

Quote:

---

Originally Posted by cowie

6970 should be 20.3% faster then a 5870, putting it ahead of the 480 a few frames in some games & since i'm on a roll i'll guess 3-10% slower then a 580.

---

Which would make the 580 a disaster, if that was the case.

we've already seen part of cayman in the 6870, we're now going to see what 4d shaders does to it.

I would expect a 20% performance increase over the 480. Which strangely, is something I expect the 580 to be like.

The only question to us, the buyer, at that point. Is what is the benefit of each card. Nvidia will have physx and a decent set of 3D, AMD will have efficiency on their side and 100 monitors support.

Take your pick, depending on what you want (I also suspect AMD can produce it cheaper).

Quote:

---

Originally Posted by cowie

6970 should be 20.3% faster then a 5870, putting it ahead of the 480 a few frames in some games & since i'm on a roll i'll guess 3-10% slower then a 580.

---

I would have say 20.4% , from where come this 0.1% difference, does i have miss something ?

lol
speculation can go + or - .03 - 70.6% depending on the moon phase ofcoarse.
Now i'm sorta with motiv
leaning 580 1.5 - 3.9% over the 6970.
I would thinnk in all fairness 20 - 25% over the 5870 asking much more would be a strech

here is another thing to talk about:
Antilles cooling system
;)

Quote:

---

Originally Posted by flopper

how come people dont ask you what you smoke when saying such things?

If it beats the 580, it be a good thing due to 40nm, and we cant expect miracles even with the 4d set of shaders.
Overclocking it to 1ghz+ however might turn it into a great card.
If r300 rumours are true, then it has to beat the 480gtx by around 80%
as that was what the 9700 did back in the day.(not going there not even whatever I might smoke)

That depends how AMD did reason when they saw what fermi did, and thought, Gee, the upgrade Nvidia are going to do at 28nm gonna be something, so they did their homewerk and produced a slightly bigger chip to crown the single card performance planned at 32nm but adjusted to 40nm which they might cut things out that wasnt needed this time until 28nm.

The size of the card seen as far suggest a powerhungry one.
See, Nvidia knows it, more power, more speed and more performance and forget to save the whales, we let the oceans boil.

---

That assumes that global warming is man made and is caused by Co2.
That damn Co2 from our GPU's seems to be heating up the Sun and the rest planets in our solar system as well.

haha nice. No way I could fit Antilles in my case if it is much longer than the 5870 :shakes:

Quote:

---

Originally Posted by SpuTnicK

here is another thing to talk about:
Antilles cooling system
;)

---

Hum lets see performance speculation from an nvidia boy point of view:

Well gtx 480= 30% faster than 5870 on ALL cases at the very LEAST
gtx 580= 20% faster than 480 AT LEAST its fact and I know it cause I saw a vantage mark somewhere in a chinese forum.
gtx 580 = faster than 6970 - this is just given and automatic fact.

since gtx 480 is 30% faster than 5870 and 580 is 56% faster than 5870 (at least and everywhere) then, 6970 will be right in the middle (cause AMD is mediocre & budget) and is 28% faster than 5870.

Quote:

---

Originally Posted by SpuTnicK

here is another thing to talk about:
Antilles cooling system
;)

---

two 8pin power

Anything less than 50% over a 5870 and i'll be quite disappointed with cayman XT.

Quote:

---

Originally Posted by SpuTnicK

here is another thing to talk about:
Antilles cooling system
;)

---

According to one of the comments that cooling system was tested, and failed on the X1900XT. I would rather have fans pushing than pulling anyway, especially on a card like antilles. Cool vid none the less.

Quote:

---

Originally Posted by Dimitriman

I saw a vantage mark somewhere in a chinese forum.

---

maybe you saw a vantage score of HD5970?

Quote:

---

Originally Posted by SpuTnicK

here is another thing to talk about:
Antilles cooling system
;)

---

Don't do this again. This might as well be fan fiction about Goku and Vegeta finding their love for each other.

Quote:

---

Originally Posted by flopper

how come people dont ask you what you smoke when saying such things?

If it beats the 580, it be a good thing due to 40nm, and we cant expect miracles even with the 4d set of shaders.

---

When it does, I'll sig you.

Fancy a monetary bet on which is faster?
100 usd via moneybookers?