Sign in with
Sign up | Sign in
Your question

AMD taping out Southern Island already?

Last response: in Graphics & Displays
Share
April 22, 2010 2:26:49 AM

Check this out:
http://www.semiaccurate.com/2010/04/21/atis-southern-is...

If this is true, then Nvidia is in big trouble. Apparently SI is just a refresh until NI comes out in 2011. What do you guys think of this?
a b U Graphics card
a b Î Nvidia
April 22, 2010 2:38:07 AM

Yep, I saw that a little while ago. Finally some reasonable info (40nm, etc.) and looks interesting. I would say he is painting a bit overoptimistic picture, but I guess at this point TSMC can't mess anything more up.
m
0
l
April 22, 2010 2:39:10 AM

If this is true I can only imagine what will happen to Nvidia. They might be beaten at every price point and performance point if this refresh comes out this fall. If the demo is in the next month or so as stated, many people may just hold out and see what performance gains the refresh holds.

Also, this will happen again...

http://www.youtube.com/watch?v=If0Bkfnifi4
m
0
l
Related resources
a b U Graphics card
a b Î Nvidia
April 22, 2010 2:40:40 AM

Well, I certainly doubt the NVidia development team for their next chip has been sleeping this whole time.
m
0
l
April 22, 2010 2:41:46 AM

crosko42 said:
If this is true I can only imagine what will happen to Nvidia. They might be beaten at every price point and performance point if this refresh comes out this fall. If the demo is in the next month or so as stated, many people may just hold out and see what performance gains the refresh holds.

Also, this will happen again...

http://www.youtube.com/watch?v=If0Bkfnifi4


Lol i love that video. But if the Northern Islands (the next gen) in on GloFo's 28nm, then there Nvidia is toast, freshly made on a 480.
m
0
l
a b U Graphics card
a b Î Nvidia
April 22, 2010 2:47:59 AM

Hopefully they'll work together better, exchange info, problems, etc. earlier. That could work out amazing for ATI/AMD/GloFo.
m
0
l
a b U Graphics card
April 22, 2010 3:25:40 AM

No matter how bad a company is, we need both of them to battle each other. If nvidia doesn't get into the game with some better cards, the 5970 and 5870 will never come down in price.

Similar to what happened with the 8800 series from nvidia.
m
0
l
a b U Graphics card
a b Î Nvidia
April 22, 2010 3:29:56 AM

I'm still firmly in the belief, despite NVidia's mistakes this round, they'll be back soon unless they decide to drop gaming cards.
m
0
l
a b U Graphics card
April 22, 2010 3:40:07 AM

I think after the sales of the fermi cards, nvidia will realize that they shouldn't have left the gaming world.

In other words, they should have put more money into the R&D department.
m
0
l
a b U Graphics card
a b Î Nvidia
April 22, 2010 3:43:31 AM

Well, fermi would have been an amazing card if not for the TSMC issues. Hopefully they've now learned they need to do their homework (and tests!) on a new process before throwing a humongous card on it.
m
0
l
a b U Graphics card
April 22, 2010 4:04:45 AM

I believe that they did, but because they were so far behind in sales, they made the best card that they can.
m
0
l
a b U Graphics card
a b Î Nvidia
April 22, 2010 4:15:26 AM

crosko42 said:
If the demo is in the next month or so as stated, many people may just hold out and see what performance gains the refresh holds.


Yeah, I don't think there would be a demo in just 2 months.
Tape out to test product is usually D-day + 6/8 weeks on a good process not the problematic 40nm, and that's for internal testing, not the kinda thing you show to outsiders.

So, I agree with EXT64, the timeline seem a little optimistic. I agree with the overall tone, that if ATi is tick-tocking this fast then it's good for product refresh and bad for nV (and unknown for the end user).

Although as mentioned in the article there is some question as to what you can achieve with the same GDDR5 limits and on the current under-performing process. However when they move to GloFo then that's when it gets interesting (although not as interesting IMO as had they still be under AMD control).
m
0
l
April 22, 2010 4:40:42 AM

Man ati needs to support the 5k series before they move on to the next thing. I have to use 10.1 due to amdkmdap errors. And it's still not working great. But it is far better than anything i had with 10.2 or 10.3.
m
0
l
April 22, 2010 7:49:39 AM

Thats bad news. Bye Bye sweet VC prices.

I bougth my 5850 for $260 and now its over $300 = 7 MONTHS LATER

What will happen if Ati release SI this fall? RAPE - thats what

You've seen it so you have to believe it.

Disapointed at nVidia
m
0
l
a b U Graphics card
April 22, 2010 11:45:58 AM

As has been mentioned, nvidia can hardly be sitting, twiddling their thumbs.

Judging from the article though this would appear to be way for ati to test new tech out and see how it fares whilst still incorporating existing stuff that they know works.

As nvidia is a little behind I guess they have that luxury.
m
0
l
a b U Graphics card
April 22, 2010 11:57:15 AM

If you compare this to Intel vs AMD before Core 2 most people were thinking that Intel was going to be out of the picture in a few years if big changes didn't occur. Well, big changes did occur and now it is AMD that needs to step up.

When you go back and look at NVidia vs ATI, NVidia and ATI have been exchanging best position for a long time, NVidia got it wrong with the 5000 series. ATI then lost while NVidia ruled with the 8800. Now it is reversed again but the one thing going for NVidia is that they do have a card, it is just in low quantities. I'm sure that we will be seeing a lot more out of them in the coming years.
m
0
l
a c 130 U Graphics card
a b Î Nvidia
April 22, 2010 12:07:50 PM

The point is that nVidia will get the rest of the GF1XX lineup out, and by the time that's finished, ATI will have gotten their refresh out. Then ATI can get to work on (hopefully!) a new arch, whilst nVidia is working on their refresh.

if nVidia skips a GF1XX refresh, then ATIs SI will kill them, and then they'll be back on track for release timings.
m
0
l
a b U Graphics card
a b Î Nvidia
April 23, 2010 12:45:59 AM

The biggest issue is that nVidia is stuck with their design this isn't their Prescott with a Core2 on the horizon, this is their P4 launch. Fermi is not the refined design it is the blunt instrument design with goals reaching past it's design capabilities, which doesn't do very well under performance per transistor #s, nor performance per watt, which doesn't bode well for their ability to produce other lines with ease.

nVidia needs to bring their design in line because it's their 3-5 year future platform, whereas AMD is just about to launch theirs, and has all the benefit of knowing their existing platform competes, especially on those performance/per #s so they can perfect their design with a test model.

nVidia aren't twiddling their thumbs, but they also aren't doing a complete redesign either, so it's not like they are going to release Faraday to fix the Fermi design, they will have to figure out how to fix the Fermi design without starting over, which would mean another 2 years+ to implement, and by then it's too late.

Short term they need and FX5900/HD3800 fix , the question is what must go and what must stay? The Polymorph engine is big and bulky and hard to scale, but it's what allows all that flexible horsepower in the back be very basic and modular and put all the DX compliance and features at the front. The reliance on Cache is difficult to address in their current design as well, and it won't scale without changes, and even then if AMD/ATi go the route they've mentioned in the past it also won't be enough to simply segment and shift it around.

Anywhoo, once again back to speculation. It can be interesting to see what comes from SI, but it seems too limited in options, where it's the transition to NI on a new process that has the potential to break things wide open if it succeeds on both fronts at the same time.

Still a long way to go, but AMD is focusing on a whole new product and try new things, while nVidia needs to focus on fixing the one they already spent 6 months fixing without breaking it while trying to make derivatives and deliver volume.

I know which position I'd prefer to be in.
m
0
l
April 23, 2010 12:55:05 AM

TheGreatGrapeApe said:
The biggest issue is that nVidia is stuck with their design this isn't their Prescott with a Core2 on the horizon, this is their P4 launch. Fermi is not the refined design it is the blunt instrument design with goals reaching past it's design capabilities, which doesn't do very well under performance per transistor #s, nor performance per watt, which doesn't bode well for their ability to produce other lines with ease.

nVidia needs to bring their design in line because it's their 3-5 year future platform, whereas AMD is just about to launch theirs, and has all the benefit of knowing their existing platform competes, especially on those performance/per #s so they can perfect their design with a test model.

nVidia aren't twiddling their thumbs, but they also aren't doing a complete redesign either, so it's not like they are going to release Faraday to fix the Fermi design, they will have to figure out how to fix the Fermi design without starting over, which would mean another 2 years+ to implement, and by then it's too late.

Short term they need and FX5900/HD3800 fix , the question is what must go and what must stay? The Polymorph engine is big and bulky and hard to scale, but it's what allows all that flexible horsepower in the back be very basic and modular and put all the DX compliance and features at the front. The reliance on is difficult to address in their current design as well, and it won't scale without changes, and even then if AMD/ATi go the route they've mentioned in the past it also won't be enough to simply segment and shift it around.

Anywhoo, once again back to speculation. It can be interesting to see what comes from SI, but it seems too limited in options, where it's the transition to NI on a new process that has the potential to break things wide open if it succeeds on both fronts at the same time.

Still a long way to go, but AMD is focusing on a whole new product and try new things, while nVidia needs to focus on fixing the one they already spent 6 months fixing without breaking it while trying to make derivatives and deliver volume.

I know which position I'd prefer to be in.


+1 on this. The problem IMO nvidia is facing is that they still only have high end next gen cards, and we all know the money is in the mid-low end (partly b/c of laptops). If AMD releases SI soon and saturates all segments of the market as quickly as the 5xxx series did, then this is going to be Nvidia's main problem. Power consumption and heat are huge factors in laptops and other mid-low end parts, and if AMD improves this further with SI, then Nvidia is going to have huge problems competing in these segments.
m
0
l
a b U Graphics card
a b Î Nvidia
April 23, 2010 2:55:34 AM

Had to edit to fix this sentence that I missed a word while at work;
"The reliance on Cache is difficult to address in their current design as well, and it won't scale without changes"

I'll leave the other typos and such there for colour. ;) 
m
0
l
!