Farcry 2 results: GF100 Vs. GTX285

[T2k]

doubtful, that's not how nvidia works, they like to have the faster cards, even if they cost more (to buy). Nvidia knows exactly how fast the 5870 is so it'll be easy for them to make a part that'll be faster. I predict the GTX360 will beat the 5870, not by much but it'll be faster and it will force AMD to lower their prices.

That would be true in normal circumstances, if they weren't

- 6+ months behind schedule
- utterly kicked out of the top and mid-market for several months now (outdated by an entire generation and still uncompetitive prices)
- suffering from horrible yields already
- weren't so badly in need of fresh revenues ASAP (investors, investors, investors!)
- Fermi weren't a power hog with very thermal envelope

Some of these are pretty solid rumors, others are speculations on my part, of course but we're speculating here anyway, aren't we?

 

[tviceman]

AMD didn't raise the price, demand outstripping supplies and etailers did. Unless GTX360 kills 5870 AMD won't lower the MSRPs, only these current exorbitant markups will disappear (they already started winding down.) Remember: officially 5870 is $390 and $5850 is $290 and these are after the November $10 bump AMD did.

Ah, but AMD did raise the price - you say it yourself in the last sentence.

Because I think the gtx360 wil be slightly better than a 5870, if the gtx360 comes in at $399, I fully expect the MSRP of the 5870 to drop at or prior to fermi's hard launch.

 

[tviceman]

Whether or not you're a fan of nvidia or AMD, this is pretty funny.

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

 

[edplayer]

AMD didn't raise the price, demand outstripping supplies and etailers did.

Remember: officially 5870 is $390 and $5850 is $290 and these are after the November $10 bump AMD did.

too funny

the 5850 increase was larger. It came out at $259 msrp

 

[HurleyBird]

IF this isn't a cherry picked benchmark. And IF this is GTX 360 and not GTX 380, then Nvidia will have the faster part but will only only be pretty much evenly matched against ATI in terms of performance per transistor/die area. While that would certainly be an improvement over G200, I don't know if it would make up for the 6+ month lag period.

On the other hand, IF this IS a cherry picked benchmark, and IF this IS GTX 380, Nvidia is pretty much screwed in the consumer space, but I can still see them doing very well in HPC.

If ATI is smart they'll have been working on an HPC only card for awhile now that focuses solely on DP performance, but my guess is they are more focused on fusion.

 

[Tsavo]

Here's a (formerly) secret picture of Fermi's cooling solution:

http://cryptome.org/eyeball/npp2/pict437.jpg

 

[RussianSensation]

Whether or not you're a fan of nvidia or AMD, this is pretty funny.

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

Hehe

$HIT! She pwn3d tri CF'ed 5850!!!

 

[nosfe]

That would be true in normal circumstances, if they weren't

- 6+ months behind schedule
- utterly kicked out of the top and mid-market for several months now (outdated by an entire generation and still uncompetitive prices)
- suffering from horrible yields already
- weren't so badly in need of fresh revenues ASAP (investors, investors, investors!)
- Fermi weren't a power hog with very thermal envelope

Some of these are pretty solid rumors, others are speculations on my part, of course but we're speculating here anyway, aren't we?

1) We don't know why they're 6 months behind AMD (not necessarily behind schedule)
2) This is a different generation so it doesn't apply
3) Yields improve over time, if i'm not mistaken, GT200 had similar yield issues when it came out
4) Pure speculation
5) This hasn't stopped Nvidia in the past, why would it stop them now?

The big question now is whether making all those changes in G100 will pay off. I personally wouldn't have done it, too much can go wrong when making major changes to your architecture and going to a smaller process. This will also make the next generation of cards very interesting because unless they want to be 6months late again, Nvidia probably won't change much of anything with Fermi2

 

[T2k]

1) We don't know why they're 6 months behind AMD (not necessarily behind schedule)

Ehh? They are behind their own schedule.

2) This is a different generation so it doesn't apply

Whadda... what "does not apply"? It's a fact, not subject of a debate, for chrissake...

3) Yields improve over time, if i'm not mistaken, GT200 had similar yield issues when it came out

You're wrong, it didn't. Yes, they improve but they seem to take an awful long time to improve and when they are late by 6-8 months it might be over, it's better to move on.

4) Pure speculation

Except it isn't: they lost their chipset business (over 30% of their entire revenue!), no integrated market, blew their mobile market with faulty chips, do not have nor will have x86 license to fight off the inevitable Fusion-type architectures from noth Intel & AMD and of course, Fermi's R&D incl. foundry costs are astronomical now.
They have to start bringing in money - do you think the Tesla trade-in preorder "Mad Science" promo is an accident?

5) This hasn't stopped Nvidia in the past, why would it stop them now?

Do you understand the context we are talking about here? I doubt it...
FYI NV never tripped over any standard and I highly doubt they ever will, regardless of having the fastest card gives JSH a hard on or not.

The big question now is whether making all those changes in G100 will pay off. I personally wouldn't have done it, too much can go wrong when making major changes to your architecture and going to a smaller process. This will also make the next generation of cards very interesting because unless they want to be 6months late again, Nvidia probably won't change much of anything with Fermi2

They will - they have to unless they can get it on 32nm.

EDIT: if you don't believe me the 6 + months, at least believe for Ryan@AT: "Being 6 months late has hurt NVIDIA, and being 6 months late has hurt consumers through uncompetitive pricing from AMD."

 

[nosfe]

Never seen their schedule (and i doubt that anybody outside of nvidia saw it) but i don't think that's as simple as: send chip to TSMC, get back working chip from TSMC on A0 and release product.

I thought that we were talking about Fermi and not GT200. Just because in the GT200 generation they used G92 chips for mid-range doesn't mean that they'll always be behind a generation on mid-range chips from now on. And i highly doubt that they'll make GT200 based mid-range chips now

There was an article on TGDaily about the GTX 280 yields but i can't find it anymore, too much interference from fermi articles on google

They have their Quadro line to give them all the profit and margins they need. What was it? some 80% market share in the workstation market?
http://www.fudzilla.com/index.php?option=com_content&task=view&id=10375&Itemid=34

Power draw/temperature won't be a big problem on the single chip card, there's still room to grow there, not that i'd like this trend to continue. Personally i like Ati's strategy of making smaller chips better because it brings cheaper cards faster on the market and i never was interested in buying these huge $500+ cards (which end up as $800+ cards where i live after you add tax and all other sorts of fun stuff).

Theoretically speaking, the dual fermi card will also be very interesting. They keep talking about how fermi will wipe the floor with 5870 but they never talk about the 5970 which leaves me to deduce that single chip fermi won't beat the 5970. It's also interesting to note that they're even more tight lipped about the dual fermi than the single chip variant, will they break the 300W barrier? if not, will they have enough performance/watt to beat the 5970? I hope they'll stop at 300W and start working on more efficient chips than just going the easy road and doubling stuff to increase performance. As a side effect this would mean a longer time between generations which also isn't a bad thing