Fermi 'slips into January'

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Also, would love to see Fermi take a huge bight in HPC as well.
As we know,m the cpu markets been maligned, so itd be great to see gpus pick up the slack, and no, REAL gpus, extend their life/usefullness
 



In a couple of years, TSMC will find itself in a hole of TSMC's making.


Nvidia and ATi are now champing at the bit to get manufacturing done at globalfoundaries... if only to mitigate against the risk of this farce happening again.
 


or it was an AMD fab. i thought its suppose to operate independently after AMD asexually reproduced giving birth to globalfoundries.

and considering a big chunk of globalF's stock is owned by adu dhabi sheiks, its still plausible as far as neutrality is concerned.



 
I don't understand the relationship between AMD and global foundaries. If someone could explain it to me. I would appreciate it.

From what I know, AMD sold off their chip making business to some rich arab and that business became Global Foundaries. GF is kind of like TSMC as far as AMD is concerned. They're paid to make chips for AMD.

At the same time, I hear AMD is investing in Global Foundaries, so this would lead me to believe they have an ownership stake.
 
They own 34% of GF
They have 50% voting rights, for decisions etc
The "investments" are for R&D and upgrading, which AMD is 34% responsible for, meaning, if AMD puts as much monies into it as they always have, GF actually has 3x as much monies as before for upgrades and R&D
 
It seems that Nvidia has been to busy with refreshes. I would've though that they were refreshing so much to buy them time for a really good GPU but it seems they were just greedy.

Too busy with Intel bashing as well.
 
Everyone seems to be blaming TSMC here a lot. I'm a huge fan of nVidia, I'll be the first to say it, but seriously, nVidia screwed up huge here, a lot of it is their own fault. TSMC was given some brutal expectations by nVidia....

As for when they are coming out. I'm betting March, there's no way it'll be June if nVidia wants to actually save face here. If it IS June, well, they might as well drop the GT300, get out of the market for a year and go balls to the wall on their next card and beat AMD to the punch 😛

Poor, poor nVidia stock holders...
 
Well, even at 3 x a 280, it wont be much faster than a 5870 refresh.
Now, itll come down to costs as well.
I dont like this much.
I was on the R600 wait train, and ended up with a 8800, and I know how it feels.
If the perf isnt there, the difference is, alot of nBidia buyers buy into that halo killer effect, which has been diffused on the ATI side for ATI buyers, and Im not so sure itll play well with nVidia customers if its late and doesnt come out a killer card
 
actually jaydee, don't oversell the 5000s on that 1, you can't talk about a refresh like. I mean look at the 4890 refresh. what lik 10-15 % faster? And the 5870 is anywhere from 15%-50% faster than the 285 GTX depending on the game?

even with another 50 faster refresh on the 5870, and it matches 2 285s. So technically ur 3 280s would still be faster.

Although I know your speaking in estimations and theories, come on your a respected user of the community don't over sell something that can sell on its own.

don't be 1 of those users that compare the 4870 X2 and the 295 GTX and say OMG the 295 is WAY STRONGER or 4870 X2 is WAY WEAKER.

Use sensible words😛


EDIT:

Off my phone hope it all made sense, and don't take it as a flame or any insult or anything just an opinion.
 
Way

Anyways i don't see a refresh from ATI or Nvidia coming out so soon as that depending on the performance gain either pushes the price of the refresh up or pushes the prices of the other cards down, and considered nvidia has sold no cards and ATI has sold only a small margin, due to shortages, i don't see either company pushing for a quick refresh.

ATI might do it for the release of the w.e we are calling the consumer end chip of fermi now if fermi performance is a bit higher then the 5870, but short of adding more stuff or finding a way to make a 25% performance gain while keeping costs down i don't see ATI doing a refresh soon.
 
@ L1qu1d

I think JDJ is factoring in all aspects here as well, 3 X 280 wold not scale nearly as well as a single 5870. Just a thought...

My personal opinion in regards to Fermi: Simply that Nvidia couldn't produce the beast they thought they were going to. It's like they're still trying to work out the kinks. They planned to get the whole GPGPU thing going, but apparently didn't anticipate the consequences that it cause on the GPU preformance. Just my 2 cents...
 


Thanks for the explanation.

But, on the downside they would only get 1/3rd of the profits and they could theoretically invest in R&D that competitors would benefit from.
 


He's not literally talking about 3 280s, he's talkinga bout the power of 3 280s so scaling isn't an issue.😛
 
Why, yes I can.Take the changes for instance.
The usage of the new shaders stands to make driver perf better, as ATI tunes them more.
Who knows what the clocks will end up being?
So far, it shows good scaling with clocks like the 4870 and 4890 did.
Add the 2 together.
Say a conservative 5% increase by then for drivers, add a conservative 10% for clocks with scaling, thats 15%.
That right there puts it around 60% faster than a 280.
Now, we all know that Fermi wont be 3x as fast, and itll most likely be 2x at best, not x2 but 2x, no sli figures here.
Shaders and clocks will limit going much beyond 2x, best case,
So, thats at best 30%, and Im cutting ATI low, and going best for nVidia.
We know thats not very likely, so well most likely see what we had before with the 4890 vs the 285, and being as late as it is, by that time, it wont be enough, not impressive enough, taking so long.
If we turn all these numbers, itll lose a few games to a 5870 refresh, and that doesnt bode well at all for nvidia, again, because it took so long.
The leap we were waiting for has become a small jump if its this late, and thats my point.
If you look at the original 4870 numbers and the original 4890 numbers, youll see a greater than 10% to 15%, just due to drivers.
And again, theres larger changes with this arch than the 4890 change which wasnt much at all, or even coming from the 3 to the 4 series, which ATI said was a bigger change with the 5 series, but not as large as the 1xxx series to the 2xxx series.
This allows for better ending perf thru driver opts.
So, look for 20% here, with refresh, if clocks are 15% or so. And thats still being conservative,
Also, I did say the 280, not the 285. which of course was nvidias refresh, and had those highwer clocks, which we wont see on the first Fermi card.
Im being very realistic here.
Not a good scenario, but doubling 4xxx to 5xxx didnt show 100%, going 2.13 wont show 106.5% either. And most importantly, the clocks will be nowheres near higher , and will most likely be lower in gains than what we saw with 4 to 5 series
 

On the upside tho, is that all the profits just wont go to TSMC, 1/3 will go back to ATI/AND for their cards from here on out
 
pushed back 'til January '10, oh no, i don't get to see new GPU benchmarks and flame wars with actual hardware uintil then

though really, i do want to see what the G300 arch can do, i won't get one (as i have a 4870 and thats fine for now), though i don't mind having competition in the GPU sector to lower prices