Nvidia Fermi GF100 working desktop card spotted on Facebook

[NapalmV5]

Rys throws down with some serious speculation (supposedly seasoned with a little insider info).

http://techreport.com/articles.x/17815/1

It's a bit long but a good read and sums up a lot of the speculation so far plus a few new tidbits.

4200mhz = 201.6gb/s = 256bit
4200mhz = 302.4gb/s = 384bit

[Piotrsama]

Isnt ati having the next generation out soon?

No, optimistic rumors put it around sept-nov 2010. (Note I said optimistic, IMO it will be later than that)

In the same way, people compared the 5870 with GTX 295, so Fermi vs 5970 is a valid comparison as well.

Fair enough.
Although, I would always prefer the single gpu card over the double gpu. (If performance is somewhat similar).

[BenchZowner]

4200mhz = 201.6gb/s = 256bit
4200mhz = 302.4gb/s = 384bit

Hmm... no?!

384BIT / 8 * 4200 = 201600
256BIT / 8 * 4200 = 134400

Too much Vodka ?

[Grandpaq]

we've known it would be March at the absolute earliest for a few weeks now, it's only news that journalists have posted article's on it.

Not according to Fudzilla

Fermi expected in January
Written by Fuad Abazovic
Tuesday, 24 November 2009 10:04


In retail & e-tail

Multiple sources now claim that Nvidia has promised to ship Fermi in January 2010. We are talking about GF100 Fermi, the GPU that is expected to defeat ATIs Radeon HD 5000-series offerings. To make matters more clear, January is the time when you can hope to buy a high-end Geforce desktop graphics card based on Fermi.

Nvidia Fermi-based cards in the enthusiast market have been pushed from a late November launch to a January launch mainly due the fact that the A2 silicon chip revision was not good enough for volume production, so A3 is about to ship to partners soon. With this in mind and if all goes well, Geforce graphics cards based on Fermi architecture are expected to ship in January.

We have also heard that the launch won't take place at CES 2010, which starts on January 7th and ends of January 10th. Instead, Nvidia might launch it later at some point, but nevertheless still sometime during the month.

[Helloworld_98]

that table loses its credibility in the first box of data, nvidia has already confirmed 3.1 billion transistors, that table is 100 million off.

[NapalmV5]

Hmm... no?!

384BIT / 8 * 4200 = 201600
256BIT / 8 * 4200 = 134400

Too much Vodka ?

hey!! why u messs wid my fun mann?!? just trying to scare ati dudes why u messsin wig my calkulasions man?!? was enjoin dat 300gb/s @ 5 inda mornin.. that thing better oc to 5000mhz 240gb/s huh?!? yeh.. alright good nite everyone!

[BenchZowner]

Hehe, I'm a downer, sorry

[trinibwoy]

dual Fermi was not officially announced, not even rumoured, while the 5970 is already on the market and it will probably be the direct competitor to Fermi in it's price area. So a fair comparison should include the 5970 as well, even if it is a dual gpu.

Well not only should it be included but it should be the primary competitor to Fermi. If after all this hype GF100 doesn't get 70%+ of Hemlock's performance it'll be a wash.

[FischOderAal]

Isnt ati having the next generation out soon?

Who knows for sure? Especially as ATI/AMD said RV870 will be the last GPU based on R600. Maybe they will be just as "late" as NVIDIA's Fermi

[Warboy]

Who knows for sure? Especially as ATI/AMD said RV870 will be the last GPU based on R600. Maybe they will be just as "late" as NVIDIA's Fermi

woo woo, Lets start up the rumor mill. Because we're making some theorybread tonight!

[tajoh111]

Rys throws down with some serious speculation (supposedly seasoned with a little insider info).

http://techreport.com/articles.x/17815/1

It's a bit long but a good read and sums up a lot of the speculation so far plus a few new tidbits.

Those clocks seem a little high, particular the shaders.

If they are true those, I have no doubt it will be competitive with the 5970.

[Helloworld_98]

Those clocks seem a little high, particular the shaders.

If they are true those, I have no doubt it will be competitive with the 5970.

I doubt it will be competitive with the 5970, those specs would put it at about GTX 295 performance on release unless they're still having performance issues.

[orangekiwii]

I doubt it will be competitive with the 5970, those specs would put it at about GTX 295 performance on release unless they're still having performance issues.

Thats assuming its the same core. Its a totally new core. This leads me to believe its a fair bit faster than gtx295. Unless the core itself is a total flop... just raw performance this thing should come close to a gtx295 and couple that with the fact that its a new core... Unless the core is worse which is highly unlikely... its unlikely for it to have similar performance to the gtx295 imo

[BenchZowner]

Thats assuming its the same core. Its a totally new core. This leads me to believe its a fair bit faster than gtx295. Unless the core itself is a total flop... just raw performance this thing should come close to a gtx295 and couple that with the fact that its a new core... Unless the core is worse which is highly unlikely... its unlikely for it to have similar performance to the gtx295 imo

Worst case scenario is an Asus Mars performance ( only worst case is being worse clock per clock than the G200, which is kinda impossible ).

[eric66]

Worst case scenario is an Asus Mars performance ( only worst case is being worse clock per clock than the G200, which is kinda impossible ).

why its worst case lol nvidia probaly will jump with hapiness if they can achive that. ati couldn't do that with 5870 and fermi specs aren't that different frm double gtx 285

[orangekiwii]

new architecture?

g80 specs weren't SO far different from g70... and yet it performed way better... why? because of architecture

5870 from 4870 is just a hardware increase with minor core tweaks

gf100 from gtx280 is a LARGER hardware increase AND a new architecture... its foolish to think gf100 is going to fail... it might not be godlike but to think Nvidia has no idea what they're doing is just plain ignorant

[madcho]

new architecture?

g80 specs weren't SO far different from g70... and yet it performed way better... why? because of architecture

5870 from 4870 is just a hardware increase with minor core tweaks

gf100 from gtx280 is a LARGER hardware increase AND a new architecture... its foolish to think gf100 is going to fail... it might not be godlike but to think Nvidia has no idea what they're doing is just plain ignorant

No were you can read official change of architecture.

[highoctane]

At any rate Nvidia squeezes more work per core than ATI and they more than doubled their core count so the potential is there even without new core tweaks along with the 384bit memory controller for bandwidth.

[orangekiwii]

and its stupid to think its not a new architecture

how long has it been since g80? 3 years? at least? 3 years without any major architectural changes... do you REALLY think they sat on their asses doing NOTHING this entire time... just because they seem stupid doesn't mean they are... and they clearly aren't no business is that stupid and second if they are they would go bankrupt in a matter of years and nvidia is FAR from going bankrupt

[BenchZowner]

why its worst case lol nvidia probaly will jump with hapiness if they can achive that. ati couldn't do that with 5870 and fermi specs aren't that different frm double gtx 285

Single GPU with 512 "SPs" is different ( efficiency-wise ) than a dual 256 "SPs" GPU solution
More TMUs/TFUs/ROPs
Way higher tris/clock
Architectural changes
Enhanced memory controller
refined "SPs"
...

[algorhythm]

Guys I know this has been a long wait, but we are really starting to digress here. There is no doubt that Fermi is new architecture. That is not something your imaginary Nvidia fanboy demons are lying to you about. We have the freaking white paper, the architecture is different.

[gumballguy]

Thats assuming its the same core. Its a totally new core. This leads me to believe its a fair bit faster than gtx295. Unless the core itself is a total flop... just raw performance this thing should come close to a gtx295 and couple that with the fact that its a new core... Unless the core is worse which is highly unlikely... its unlikely for it to have similar performance to the gtx295 imo

Comments on new architecture aside, 5870 doubled the transistor count, but we got less than that back

Its unlikely to be slower, but it could be a flop, and we could not see the returns we expect when judging it by its components

I personally have a bad feeling about fermi, but until the 2gb 5870's come out, I'm just waiting and watching. We'll see how close fermi is to release when those 2gb'ers get here, and then I'll decide if I wait for nvidia or not.

[eric66]

Single GPU with 512 "SPs" is different ( efficiency-wise ) than a dual 256 "SPs" GPU solution
More TMUs/TFUs/ROPs
Way higher tris/clock
Architectural changes
Enhanced memory controller
refined "SPs"
...

well 4870x2 prooved that wrong dual 800 shaders performed better plus Rop are lower than dual gtx 285 according to that report and we also don't know anything about tmus so there is big possibility for it to perform on par with dual gtx 285s or worse. and just because its new arch doesn't mean it will be something phenomenal nv 30 and r600 were new archs too lol there is big chance for failure cause nvidia tries to acomplish too much thing with same gpu .....

[spursindonesia]

It is a REVISED, new mArch, yes, but will it be THAT GOOD efficiencywise for gaming purpose compared to its older brother mArch? who knows ? All that we have is speculation, and both perspectives, optimist and pessimist one, have their own merits.

[InCredible]

now correct me if im wrong..but its only double the SP's but nothing else is double correct? as far as TMU's/TFU's/ROP's?..