NVIDIA GeForce GTX 570 in December; AMD to launch Radeon HD 6900 Series on December

[Alien Grey]

NVIDIA GeForce GTX 570 in December 2010

The NVIDIA GeForce GTX 580 flagship, based on a fully-enabled GF110 chipset was launched on November 9.

A slightly watered-down version of the GF110, or rather chips that do not meet the GeForce GTX 580 spec, will be placed on the upcoming GeForce GTX 570. The GF110 chip on the GeForce GTX 570 will have one streaming multiprocessor disabled, bringing the total CUDA core count to 480 - the same as the outgoing GeForce GTX 480. Performance is said to be around the GTX 480's level, but with a TDP quite similar to the GTX 470. No further specifications are available, but we would expect 1280MB memory on the new card.

Our source tipped us that some of NVIDIA's partners have mentioned that the GTX 570 is confirmed for December, so we'll just have to wait for it to drop.



AMD launches Radeon HD 6970 and HD 6950 on December 13

AMD has officially responded that they will launch the Radeon HD 6970 and HD 6950 late - on December 13 - from November 22, citing market performance and availability of currently available products.

PR speak aside, we are thinking the delay is simply due to the Radeon HD 6970 being not competitive enough against the GeForce GTX 580, or in a worse case scenario, it is just about as fast as the soon-to-be-launched GeForce GTX 570. Therefore, AMD is making last minute tweaks to operating frequencies (and hence why partners did not receive any final BIOS so far).

http://vr-zone.com/articles/nvidia-g...3th/10316.html

[ajaidev]

I think we had a tread on GTX 570 could be wrong tough...

Good news in any case if its 8-10% faster than GTX480 it would be quite popular. But the GTX 560 rumored with 3xx shaders is just pathetic at least they could have given it around the same numbers as GTX470.

[ripken204]

i cant wait for a month from now, plenty of new options to choose, and i need a new card badly.
AMD better get a top of the line card out, if it does only compete with the 570 that would be quite disappointing.

[Dimitriman]

hasnt amd already said officially that adjusting the clock speeds of duzens of thousands of chips could take months not weeks to perform? why do "news" sites keep insisting that it is because it wont be competitive?

[Manicdan]

hasnt amd already said officially that adjusting the clock speeds of duzens of thousands of chips could take months not weeks to perform? why do "news" sites keep insisting that it is because it wont be competitive?

cause drama sells

[informal]

hasnt amd already said officially that adjusting the clock speeds of duzens of thousands of chips could take months not weeks to perform? why do "news" sites keep insisting that it is because it wont be competitive?

Because the news sites don't know better and like the sensational speculation since it attracts (less informed) readers.

[qcmadness]

i cant wait for a month from now, plenty of new options to choose, and i need a new card badly.
AMD better get a top of the line card out, if it does only compete with the 570 that would be quite disappointing.

Can GTX580 beat HD5970 convincingly?

AMD does not need to hurry.

[SKYMTL]

Because the news sites don't know better and like the sensational speculation since it attracts (less informed) readers.

And people saying the HD 6950 would equal the GTX 580 in performance weren't being in the least bit sensationalist?

Honestly, when the facts are lined up, VR Zone's post seems to be much more grounded in reality than many of the others we have seen. One way or another we likely never know what is truly going on behind the scenes.

[zalbard]

Old news...
Will GTX570 have 1.5GB of VRAM? That could make it interesting.

[hurrdurr]

And people saying the HD 6950 would equal the GTX 580 in performance weren't being in the least bit sensationalist?

Honestly, when the facts are lined up, VR Zone's post seems to be much more grounded in reality than many of the others we have seen. One way or another we likely never know what is truly going on behind the scenes.

Indeed. Especially neliz at B3D said he was pretty sure 6950 would "trade blows" with the 580. Now he's talking quite to the contrary. What happened?

[LightSpeed]

Indeed. Especially neliz at B3D said he was pretty sure 6950 would "trade blows" with the 580. Now he's talking quite to the contrary. What happened?

I hear that the 69xx targeted higher clocks, but failed to meet them. Dunno but all I hear is everything didnt go as well as planned. Lets wait till mid December

Make no mistake, im not saying the 69xx will suck, its just not going to be the cow devouring obliterating insanely fast card that people were hoping.

edit: IMO 6970's gonna trade blows with the 580, and the 6950 with the 570.

[Dimitriman]

I hear that the 69xx targeted higher clocks, but failed to meet them. Dunno but all I hear is everything didnt go as well as planned. Lets wait till mid December

Make no mistake, im not saying the 69xx will suck, its just not going to be the cow devouring obliterating insanely fast card that people were hoping.

at the very least it will be faster than the gtx480, that itself ensures it cant be too far off the 580.

[Manicdan]

if the 6950 is a little above the 5870, thats "ok", then the 6970 will need to be above a 480 by about 10% to also be "ok" by me

based on old SP count perf relative to new SP count perf, its starting to sound scary that this new 4D isnt uber compared to older 5D.
unless the SP count weve herd is way off. but sofar it seems that 1600 5D shaders is not much different than 1500-1600 4D that would be in the 6950
then comes the question of size. if the chip used by caymen is like 20% bigger than a cypress, but only 25-30% faster, its hardly an improvement that couldnt be had by simply going with a bigger barts core.

[LightSpeed]

if the 6950 is a little above the 5870, thats "ok", then the 6970 will need to be above a 480 by about 10% to also be "ok" by me

based on old SP count perf relative to new SP count perf, its starting to sound scary that this new 4D isnt uber compared to older 5D.
unless the SP count weve herd is way off. but sofar it seems that 1600 5D shaders is not much different than 1500-1600 4D that would be in the 6950
then comes the question of size. if the chip used by caymen is like 20% bigger than a cypress, but only 25-30% faster, its hardly an improvement that couldnt be had by simply going with a bigger barts core.

Im not saying that its not going to be fast, it certainly will be. You might remember me mentioning a month before Barts's release that its gonna be based on Cypress and be very similar, but some people refused to believe me. Anyway, based on what I hear, with it missing the target clocks (albeit not by much, most likely between 850 and 900 when they targeted 950+) its not going to be faster than the 580, most likely trade blows with it. Similarly the 6950 should be around a 570. Thats around 1.4x and 1.2x of Cypress, and thats what should be expected from a next gen card on the same node IMO.

[Manicdan]

Im not saying that its not going to be fast, it certainly will be. You might remember me mentioning a month before Barts's release that its gonna be based on Cypress and be very similar, but some people refused to believe me. Anyway, based on what I hear, with it missing the target clocks (albeit not by much, most likely between 850 and 900 when they targeted 950+) its not going to be faster than the 580, most likely trade blows with it. Similarly the 6950 should be around a 570. Thats around 1.4x and 1.2x of Cypress, and thats what should be expected from a next gen card on the same node IMO.

yeah that would be much more respectable in my opinion, a gain of 20% vs 40% over cypress is a difference of just 20% on the charts, but 100% extra 'improvement' difference.

[generics_user]

if the 6950 is a little above the 5870, thats "ok", then the 6970 will need to be above a 480 by about 10% to also be "ok" by me

it wouldn't be ok for amd as they could fill that spot with higher clocked barts chips (higher mem clocks)...

6950 has to be at least 20% faster than barts or it wouldn't make any sense to release it...

[Leeghoofd]

Interesting end of year products...

[Your_Boss]

hasnt amd already said officially that adjusting the clock speeds of duzens of thousands of chips could take months not weeks to perform? why do "news" sites keep insisting that it is because it wont be competitive?

So how do we know they were bumping the clocks UP? What if the 6950 was too good and they needed to bump the clocks DOWN so as to get more of them out? Maybe the 6950 with low clocks can trade blows with the 570 maybe even the 580 and the 6970 is the mother load...a monster all in itself. It would be the top enthusist card.

Less than 1 month to go.

[Alien Grey]

Interesting end of year products...

A little competition isn't bad. The closer the performance the lower the price.

[tajoh111]

it wouldn't be ok for amd as they could fill that spot with higher clocked barts chips (higher mem clocks)...

6950 has to be at least 20% faster than barts or it wouldn't make any sense to release it...

Not at all.

The yields on the 6970 can't be the greatest really and they have to do something with salvaged parts.

Additionally, the street price of the 6870 is around 260 right now. Something 20% faster could command a price atleast 20% higher and since performance/price gains are not linear as you go up, they could charge 329 for it.

[Manicdan]

it wouldn't be ok for amd as they could fill that spot with higher clocked barts chips (higher mem clocks)...

6950 has to be at least 20% faster than barts or it wouldn't make any sense to release it...

i threw together some quick numbers based on TPU review for the 580 at 1920 resolution

the 6850 is 61% the perf, with 170% the perf/$
6870 is 71% and 148%
the 580 is 100% and 100% btw

so the 6950 if 325$ would get 81% of the 580, its a little over the 5870 (which is 75%), and its price/$ would be 125% of the 580

then for the 6970 at 400$ if it got 91% of the 580 would get a price/$ of 114%

notice the perf numbers for 6xxx (61, 71, 81, 91). then the price difference was a simple 75$ between each other for caymen.

this is kinda the line i was thinking of as being "ok". the 6950 is slightly ahead of the 5870, and the 6970 is slightly ahead of the 480, or soon to be 570. but both are priced quite competitively, while also not having super awesome perf numbers.

if the 6950 was 85% of the 580
and the 6970 was 96% of it,
then the perf/$ would be much more inline with the 68xx and i would be quite pleased.

please remember these are all fake estimates based on purely what if, which simply speculates on perf relative to price ranges. if AMD released a huge power hungry and expensive monster, then sure its more than expected to hopefully beat a 580, and cost just as much. but i think the gap between a 6870 and 6950 is too big, then we will need a 6930 to fill that gap.

[herderien]

well i still wonder what will be under my xmas tree ^^

[xVeinx]

well i still wonder what will be under my xmas tree ^^

Coal?

[Lanek]

if the 6950 is a little above the 5870, thats "ok", then the 6970 will need to be above a 480 by about 10% to also be "ok" by me

based on old SP count perf relative to new SP count perf, its starting to sound scary that this new 4D isnt uber compared to older 5D.
unless the SP count weve herd is way off. but sofar it seems that 1600 5D shaders is not much different than 1500-1600 4D that would be in the 6950
then comes the question of size. if the chip used by caymen is like 20% bigger than a cypress, but only 25-30% faster, its hardly an improvement that couldnt be had by simply going with a bigger barts core.

how much slower is the 6870 vs 5870, 5 % ( 10% )? do you really think AMD will release a 6950 who will just be a little bit faster of the 5870? it's a non sense, even an overclocked 5870 will do the job so.
Ofc we are speaking about the 6950 but seriously, i see rather the 6950 as fast of the 480 ... i don't think AMD have decided to develop Cayman gpu's in other way... and if the 6950 match 480, it's not really hard for the 6970 to be on the same range of the 580 ( i don't say faster or slower, just in range ). The 580 is not really a revolution in term of performance vs the 480... we are not speaking about a complete new architecture who can draw 50% perf more over the 480 ....

Untill offcourse they have face trouble with the architecture, and something is going wrong... 2-3 weeks more delay, will not really help to this.

But looking at the 6950 vs GTX570 bring an interessant question. 570 will be really close of the GTX480, even equal on paper.. if the 6950 was designed to match it, they are maybe in doubt with it, more of what they can worry about the 6970...

Anyway they can't really change anything 2 weeks before the launch (initially on 22th nov ) , this ask too much of work and money, reflash the gpu's, need send back to validation all gpu's produced, ask to AIB to throw to the trash all the retail boxes ( who get spec on it ), contact their printers, and redo all, without seeing manufacturer workers cost by hours ( nobody work for free ) ... It's not like they will just show a box and 1 gpu flashed to 50mhz more, and all is fine.... this an industry, you can't ask to your partner to do 2x the same work just cause you want it. this will increase the production cost (even for the AIB who just put a stickers on the coolers and print a box ) ... if it was 2-3months ago, and production have not start, ok... but not 2 weeks before the initial launch.

Now AMD could have seen the GTX"580" coming, and prepare the launch differently, keeping a time windows for be able to change 1-2things on last minutes when they will have 100% confirmation of what perf they will face... but this type of practice is not really common on this type of industry where you deal with multiple partners.

[ripken204]

Coal?

worse, a mac