Next Gen Nvidia 2nd half 2011

I would imagine there will be some sort of Fermi refresh sometime in between.
 
This makes me more interested in the 6xxx lol. Unless Nvidia has a refresh, you can almost jump in blind with the 6xxx and not be able to go wrong.
 
You'd have to be an idiot to think there is not going to be a refresh between now and then..
 
You'd have to be an idiot to think there is not going to be a refresh between now and then..
But you'd also have to be an idiot to think that said refresh will offer something significantly more powerful than what will already be on the market when that refresh arrives.
 
But you'd also have to be an idiot to think that said refresh will offer something significantly more powerful than what will already be on the market when that refresh arrives.

Yup, they aren't able to do a shrink like G92/G80 so I dunno what options are out there to make something appealing to counter the 6xxx series. The fact there are no mention of the refresh besides "a refresh is coming" doesn't exactly tweak me to something exciting, you'd think they would have it ready to go already to counter what is coming in October. I honestly don't see anything much different besides a possible dual gf104 card, which has been mostly a fanboy creation/dream imo. In any case it will be interesting what happens in October, if there is no mention of this so called refresh at the same time or shortly after the 6xxx series I know what product I'm pulling the plug on.
 
"3 to 4 times" faster is an awfully vague statement by Huang for a card that isn't even close to being done yet. 12-15 months is probably optimistic, there is always problems going to a smaller dye.
 
But you'd also have to be an idiot to think that said refresh will offer something significantly more powerful than what will already be on the market when that refresh arrives.

Really? You don't think NVIDIA will release a fully operational Fermi... ever? Doesn't sound that far fetched to me.

§kynet;1036206457 said:
Nvidia is getting their ass kicked by an old piece of mold.

Derp. Yea, the 5870 is totally kicking the 480's butt. :rolleyes:
 
Mildly interesting.

Info about a possible refresh would be way more interesting (we know that they have to get out a refresh, but the question will be, based on what? inefficient GF100 like or more efficient GF104?... i think we would rather see them perfect the GF104 at least for the consumer market, leave the GF100 refresh for pro-sumer and tesla markets)

editted to fix typos
 
They didn't list the 2008's 9xxx refresh so not mentioning one on that slide doesn't mean anything. The 465/70/80 cards all suffered from nVidia missing a major engineering problem with the TSMC 40nm process until too late in the game. The more recent 450/460 launches show that they've updated their designs to cope which leaves room for a significantly improved top level chip to be released after the refresh.
 
Actually AMD is still using the moldy old R600 core. You know the one that brought us the legendary HD2900.

Looks like the first HD6000 won't be impressing anyone either
http://hardforum.com/showthread.php?t=1547183

i do not see anything that backs up either of your claims... amd is currently over a year ahead of nvidia its sad to see nvidia just recently responding to cards that have been in the market since this time last year ...
 
i do not see anything that backs up your claim. do not believe everything on the interwebs and please do some real research.... god

The fact that everyone is talking about the HD6000 is aggravating in and of itself.

All anyone says is BETTER and FASTER and JUST KEEP WAITING, justifying their claims without a single reliable source.

Something new is coming. We know that. That is all we know. You either play the waiting game or you don't. We aren't talking about a new CPU socket, here.
 
"3 to 4 times" faster is an awfully vague statement by Huang for a card that isn't even close to being done yet. 12-15 months is probably optimistic, there is always problems going to a smaller dye.

Perhaps he means 3 to 4 times the wooden screws!
 
All this talk of kepler vs southern islands doesn't matter because by the time kepler comes out amd will have their own 28nm products coming out as well. Was nvidia late to the dX 11 game? Yes. Does that necessitate that they will perpetually be behind? Of course not. Time will tell.
 
Three to four times faster than Fermi? Hell keep the thermals the same as Fermi and that's kick ass. AMD will obvisouly have the single fastest card but with 3 480s I'm good for a yeat easy. Not giving up 3D.
 
Really? You don't think NVIDIA will release a fully operational Fermi... ever? Doesn't sound that far fetched to me.
If they could have, they would have released one by now. Or what do you actually think that selling a crippled chip with functional units disabled because of bad yields is good business sense? Yeah that must be it, they are going to surprise everyone and anytime they will release a fully functional 512 shader gtx480...LOL. Next thing people are going to say is they purposely delayed the gtx480 to release it after the Christmas buying season when it was supposed to come out to March right?

Derp. Yea, the 5870 is totally kicking the 480's butt. :rolleyes:
Yeah it is. Just because it wins a few benchmarks and a subset of the already small enthusiasts market are jerking off to that does not mean it wins in the market. Speaking conservatively the ratio of 5870's to 480s should be at least 3:1. If we count GF100 to Cypress chip based products the ratio is 4:1
 
Last edited:
I'm an NVIDIA fan (but not a blind one) and haven't bought any 400 series cards.....same for alot of people. Nothing more needs to be said....
 
I'm an NVIDIA fan (but not a blind one) and haven't bought any 400 series cards.....same for alot of people. Nothing more needs to be said....

I bought a 480 and love it. I'll probably sell it and upgrade for one of these when the time comes.
 
3-4 times faster in a year? lol wtf is he smokin

It's in a specific category: double precision floating point operations per watt. The largest chunk of that "performance per watt" is going to come from moving from 40nm to 28nm. The rest of that performance is just going to come from architecture.

"3-4x performance" is a bit on the sensationalized side. Actual gaming performance is likely going to be what you'd expect from a GPU released in late 2011.
 
The fact that everyone is talking about the HD6000 is aggravating in and of itself.

All anyone says is BETTER and FASTER and JUST KEEP WAITING, justifying their claims without a single reliable source.

Something new is coming. We know that. That is all we know. You either play the waiting game or you don't. We aren't talking about a new CPU socket, here.

Of course there aren't any reliable sources, it's all RUMORS. That is how this always works. We almost never get anything official until the day the card launches.

That said, AMD actually confirmed the 6xxx series is a new architecture and that it will be out this year (so the rumors of some new cards being a month or so away are possible): http://www.techradar.com/news/compu...ics-pre-christmas-717388?src=rss&attr=newsall
 
Yup, they aren't able to do a shrink like G92/G80 so I dunno what options are out there to make something appealing to counter the 6xxx series. The fact there are no mention of the refresh besides "a refresh is coming" doesn't exactly tweak me to something exciting, you'd think they would have it ready to go already to counter what is coming in October. I honestly don't see anything much different besides a possible dual gf104 card, which has been mostly a fanboy creation/dream imo. In any case it will be interesting what happens in October, if there is no mention of this so called refresh at the same time or shortly after the 6xxx series I know what product I'm pulling the plug on.

And AMD is in the same boat, trying to get more performance out of the same process. Unless you're trying to say they have some magical process which doesn't exist yet anywhere...
 
AMD has 28nm gloflo. So for the first time in a great deal of time they are not dependent on TSMC. THAT is going to be what shatters any efficiency gains Nvidia makes.

Magic? No but AMD is claiming some 100 percents on complex test patterns a while ago.

(Before some fanboi mentions their 32nm issues be aware that is a COMPLETELY different process. Made for CPUs and APUs. And is not expected to delay them very long 28nm is going well)
 
I've heard rumors of nVidia's another refined Fermi card GTX 485.

Still rumored, The way these tech guys described it, 480 might be the fastest single GPU solution. But the problem of 480 is high power consumption and enormous heat. Then they released GTX 460 which consumes less power and is cool and calm. So the sources claim that, just 460 there is going to be another refined Fermi GTX 485, which will perform the same as 480 (maybe 4-5% faster than 480) but just like 460, it will consume less power and generate less heat.

Although I don't believe in rumors, but if this is true, then 485 will be a good thing to try out
 
It is going to be far more important to have good yields. The disaster that is 40nm shows that fabs can have a much greater impact on profit and market share than the design company itself!

If Gloflos near 100 percent yields stick around while TSMC thinks it acceptable with much less. AMD will wipe the floor with Nvidia.
 
Since AMD already has the single-card crown with the 5970, they'll inevitably keep it. (They're only another pathetic driver release or two away from losing this though, too. Can't believe people still keep reverting back to drivers from months ago...)

Single-core crown? Had better achieve it.

Multi-core crown? With the sad state of affairs of AMD's driver dept, I am actually anticipating that tri-SLI GTX 480s will keep the multi-core/card crown.

But AMD's already burned a bridge to me with their GSOD fiasco. Not even considering risking that failboat when I know I can just snag a third GTX 480 and have similar performance regardless of what AMD puts out... And PhysX to boot, too...
 
Similar performance perhaps with a nice helping of extra power bill and extra power bill so the AC and push that extra heat out.

The ole "AMD Drivers suck" BS again. I have gone many upgrades with not a single showstopper issue.

As for whoever takes crowns. The only one that matters is Yield and binning. If you cant get good yields and good binning to put superior silicon into top end models it matters not. As he who rules the yield rules the mid to lower end segments and thus the market.
 
Similar performance perhaps with a nice helping of extra power bill and extra power bill so the AC and push that extra heat out.

The ole "AMD Drivers suck" BS again. I have gone many upgrades with not a single showstopper issue.

As for whoever takes crowns. The only one that matters is Yield and binning. If you cant get good yields and good binning to put superior silicon into top end models it matters not. As he who rules the yield rules the mid to lower end segments and thus the market.

AMD are the best for me. Have been using them for a while now with no issues. Guess those who know how to use it are at benefits.
wink.gif
 
And AMD is in the same boat, trying to get more performance out of the same process. Unless you're trying to say they have some magical process which doesn't exist yet anywhere...

No, AMD has one extremely crucial advantage - the 58xx has a much smaller die compared to the GTX 470/480. AMD has boatloads of room to work with 40nm. The 5870 has ~2.15 billion transistors to the 480's 3 billion transistors. Even a 30% increase in transistor count will still put it below the 480's count.

Also, you can do tons to improve performance by improving the architecture (which the 6xxx series is supposedly doing). You don't have to have a process shrink to make a faster chip.

So if AMD combines the two, bigger chip + improved architecture, they could end up with some significant performance gains despite still being on 40nm.
 
They didn't list the 2008's 9xxx refresh so not mentioning one on that slide doesn't mean anything. The 465/70/80 cards all suffered from nVidia missing a major engineering problem with the TSMC 40nm process until too late in the game. The more recent 450/460 launches show that they've updated their designs to cope which leaves room for a significantly improved top level chip to be released after the refresh.

This is the crux. AMD solved leakage problems on the TSMC process with redundant layer interconnects. If nVidia does the same in a refresh, a full 512-shader Fermi running cooler on less power at high clocks becomes possible, and a potentially potent product. As implied by Dan, they probably already did a trial run at this to get the 460.

Remember when the 7900's were exposed by shifts in game software to be only a modest upgrade from the 7800's and not very competitive with the X1900's? Then NV pulled out the 8800's and galloped back to the top? Or 5800>6800? Never count them out.
 
With the sad state of affairs of AMD's driver dept, I am actually anticipating that tri-SLI GTX 480s will keep the multi-core/card crown.

But AMD's already burned a bridge to me with their GSOD fiasco. Not even considering risking that failboat
You post this in every. Damn. Thread.
 
AMD has 28nm gloflo. So for the first time in a great deal of time they are not dependent on TSMC. THAT is going to be what shatters any efficiency gains Nvidia makes.

Magic? No but AMD is claiming some 100 percents on complex test patterns a while ago.

(Before some fanboi mentions their 32nm issues be aware that is a COMPLETELY different process. Made for CPUs and APUs. And is not expected to delay them very long 28nm is going well)

Yeah, but how much capacity will they have, since they are also relying on that for CPUs. They've been using TSMC for GPUs for ages, and I haven't seen anything to indicate that they are going to switch anytime soon.
 
Back
Top