ATI Northern Islands (6xxx) also at 40nm?

[Kuzi]

As I've already mentioned a few times on the forum, both TSMC and GF would not have the ability to manufacture a high end chip at a process size under 40nm this year. Now Fudzilla is claiming that ATI's next gen will be a 40nm chip too. You can check the link below:

http://www.fudzilla.com/content/view/17782/1/

 

[soccerballtux]

It's not like they'll need to produce a new chip this year the way Fermi's looking. The smart thing to do for the investors would be to keep prices high and wait for games to catch up. Currently not many games (read: none besides crysis) stress the 5870.

 

[Keysplayr]

Fermi isn't looking like "anything" yet. Bad or good.

 

[Bateluer]

Fermi isn't looking like "anything" yet. Bad or good.

Thought its been looking huge, hot, power hungry, and expensive for months now?

 

[Keysplayr]

Thought its been looking huge, hot, power hungry, and expensive for months now?

To who?

 

[Kuzi]

Fermi isn't looking like "anything" yet. Bad or good.

True, but the 5-6 month delay can't be good, especially if the release next month is a paper launch with very limited quantities and excessive prices.

 

[Keysplayr]

True, but the 5-6 month delay can't be good, especially if the release next month is a paper launch with very limited quantities and excessive prices.

As far as I can tell, the only thing that can't be good about a 5-6 month delay, is a 5-6 month delay. Annoying for sure. But what can we do?

 

[bunnyfubbles]

It's not like they'll need to produce a new chip this year the way Fermi's looking. The smart thing to do for the investors would be to keep prices high and wait for games to catch up. Currently not many games (read: none besides crysis) stress the 5870.

no need to take the foot off the accelerator if they don't have to

besides, even if Fermi is a debacle much like the GeForce 5800, the 5900 wasn't too far around the corner, as well as the 6000 series that really gave nVidia their edge back.

and from the looks of things, something would have to be really broken in the hardware for Fermi to be as bad as the 5800 was

 

[LtGoonRush]

Rumors are that TSMC canceled their 32nm process to focus on the transition to 28nm, so I don't think it's realistic for AMD to release any further die-shrunk GPU parts this year. What is possible is for them to achieve higher transistor density on a more mature 40nm process, and possibly at lower voltages, which would drive down leakage currents and improve power efficiency. This probably precludes a 100% performance jump like from R4870->R5870, but a smaller, more power-efficient GPU that delivers somewhat better performance at a lower cost to manufacture is nothing to complain about. Unless you're nVidia

 

[Ben90]

It's not like they'll need to produce a new chip this year the way Fermi's looking. The smart thing to do for the investors would be to keep prices high and wait for games to catch up. Currently not many games (read: none besides crysis) stress the 5870.

I would love to work for you.

"Alright guys, we did a good job designing this BGM-109 Tomahawk, lets go out to the bar while we wait for North Korea to catch up"

 

[nonameo]

I would love to work for you.

"Alright guys, we did a good job designing this BGM-109 Tomahawk, lets go out to the bar while we wait for North Korea to catch up"

You can still be developing something new in the wings while waiting for NV to catch up. For instance, they could work on refresh that uses less die space but generally has the same performance. This would allow them to stay on top and reduce their manufacturing costs. This could be used as the next generation's "budget/mainstream" line when NV comes out with something competitive in price and performance. It probably wouldn't be a bad idea to work on a real next generation in the process too. I don't think the OP was intending to say that they should sit on their laurels, but rather that it doesn't make any sense to release a bigger better faster product if you have no competition. Just focus on reducing your costs instead.

 

[yh125d]

To who?

Everyone but you?

Regarding NI, considering the rough numbers we've heard for their planned release (holiday season '10-ish), It's likely they already know what node it will be on. Designing GPUs is a long process, and process node is probably one of the earliest factors nailed down. It would seem unlikely to me that they would change nodes only ~10 months before intended release.

If that is true, and they are still aiming for decemberish, 40nm is probable. But IMO, they'd be smarter to release a 5890 refresh here in a few months, and wait to release 6xxx at 32nm in a year or so

 

[Falloutboy]

I thought the next gen chip amd was moving to GF anyway? why not just wait till there ready for 32nm there? nvidia sure ain't going to build a much bigger chip on 40nm from the sound of it

 

[MrK6]

Fermi isn't looking like "anything" yet. Bad or good.

Already on damage control, I see.

I was kind of hoping AMD would jump in with 32/28nm (or whatever is next), but I suppose it's correct to assume that they shouldn't partake in such a risky venture unless they had to. Considering the time and effort they put into 40nm, and the troubles it gave them along the way, it would be worth it for them to milk their investment for one more generation. I just hope Northern Islands really delivers and improves on the 5xxx family. AMD is on a roll right now, I hope we keep getting better and better products.

 

[dguy6789]

Wouldn't surprise me if Northern Islands was built using 40nm. Of course everyone would prefer a smaller process but it's possible it won't happen.

 

[Phil1977]

I thought the next gen chip amd was moving to GF anyway? why not just wait till there ready for 32nm there? nvidia sure ain't going to build a much bigger chip on 40nm from the sound of it

That was my understanding also...

My next card won't be a 40nm. I wan't 32 or 28nm and preferably coming from Dresden, Germany

 

[soccerballtux]

I would love to work for you.

"Alright guys, we did a good job designing this BGM-109 Tomahawk, lets go out to the bar while we wait for North Korea to catch up"

this was how business operated till everyone got onto the stock market and managers had to continue producing. Until then, if you were top dog, you didn't have to report to anyone-- just stay on top.

 

[waffleironhead]

Fermi isn't looking like "anything" yet. Bad or good.

well, it is looking late. They missed their 60 day launch window.
its looking castrated. They cut the sp count down.
Whether those are good or bad in the whole scheme of things we'll have to wait and see.

 

[Rezist]

I was hoping for a 28 nm chip this year but given how badly 40 nm was I can see why that won't be happening. Hopefully early next year we have something, I can wait for quite a while.

 

[Munky]

Fermi being 6 months late is not a "good" thing by any measure.

 

[bryanW1995]

there have been rumors that amd is considering gf for the next process node, but nothing concrete has been announced yet. I would be shocked if gf wasn't going hard after both amd AND nvidia business right now, however, especially in light of all of tsmc's recent issues. maybe 1/2 nodes isn't the best way to go, maybe they'll shoot for a 1/2 step down to 32nm this year at gf or tsmc and just go back to even nodes going forward. nvidia in particular has been playing the 1/2 node catch-up game for a while now and with fermi it appears to have caught up to them.

 

[Grinja]

I was hoping for a 28 nm chip this year but given how badly 40 nm was I can see why that won't be happening. Hopefully early next year we have something, I can wait for quite a while.

I would guess that they would try one of their less complicated chips on the new process.
If the foundries are ready early next year we may see a budget version of Northern Islands or perhaps some mobile chip?

 

[v8envy]

Since a test run of the 4770 worked so well for ATI (and lack of any test run so poorly for NV) I think it's safe to assume we'll see a budget part on a new 32 or 28nm process at least 6 months before anything interesting.

From that viewpoint 40nm for the rest of the year is likely for both companies.

 

[soccerballtux]

Since a test run of the 4770 worked so well for ATI (and lack of any test run so poorly for NV) I think it's safe to assume we'll see a budget part on a new 32 or 28nm process at least 6 months before anything interesting.

From that viewpoint 40nm for the rest of the year is likely for both companies.

maybe the 5770.

 

[BTRY B 529th FA BN]

5890