GTX 580 rumors

Honestly this makes no difference to me. Because of how my budgeting goes, I don't buy the absolute top of the line, but rather the lower end of the high performance line. For example, with intel's line of processors, I'd be looking at the lowest cost core i7s. With ATI, oh pardon me, AMD, GPUs I'd be mostly looking at the lower x800 or higher x700 cards (based off the current scheme not the new one). On nvidia, I'd probably be looking at (with the newer naming scheme) GTX x60 cards, assuming I didn't buy a generation back.

The only thing that I hope is that they have a "GTX 560" out by the time I start putting together a build, although I'd be perfectly happy with the 460.
 
chinesepiratefood said:
I think nvidia can come up with a great card if they take a 480 and make the same changes they made to the 460 (removing some of the cuda stuff)

Hilarious how anti-nvidia this board has become though.
Click to expand...

That would be great. I do not think you assessment that this forum is anti Nvidia is fair or right.
Sure there will always be fans, but first and foremost people here are enthusiasts and root for maximized competition between red and green, so they can reap the best GPUs possible at the segment they take aim at. I for one, do not have any preference between the two, and have owned many cards from both teams, besides wanting a solid product with the features I want at a price that I consider fair.
 
Well, if Nvidia cuts off a lot of the double precision and GPGPU-specific stuff like they did with the GTX 460, that opens up a ton of extra transistors for them to repurpose.

It's possible, but it would cause the GeForce and Tesla to require two different GPU designs, which I don't see Nvidia doing just yet. I think that they will do that in a couple of years once (if?) the Tesla market has grown, but I don't think they will quite yet.

Dual GTX 460s seems far more likely.
 
Not only that, but it would be a bit of a retrograde step from the direction Nvidia is pursuing on the GPGPU market. I certainly think they'd have a fine product on their hands if they did, but I think they'd need to ask themselves if they want to make this tradeoff.
 
Falling Anvil said:
Not only that, but it would be a bit of a retrograde step from the direction Nvidia is pursuing on the GPGPU market. I certainly think they'd have a fine product on their hands if they did, but I think they'd need to ask themselves if they want to make this tradeoff.
Click to expand...

Any engineer working on DP code should have a Tesla board in their box. DP support on the GTX 400 series is already limited. (They turned it off. The transistors for DP and ECC are still there.)

Bottom line is it's a necessary step if they want to compete with the 6000 series.
 
Lord_Exodia said:
Very nice find. Hmm I hope that doesn't get Kyle in trouble. I doubt it cause he's done this for so long I doubt he was under NDA and slipped up. That's it for me 100% confirmed.
Click to expand...

Yes, Kyle had obviously seen a real roadmap, which by definition must have come from NV, and the only question was if this was a new High-End chip or Low-End chip
- and I think it's obvious now that it must be high-end.

So, if it's still on schedule, we should get some more concrete info soon!
:)
 
eddieck said:
Any engineer working on DP code should have a Tesla board in their box. DP support on the GTX 400 series is already limited. (They turned it off. The transistors for DP and ECC are still there.)

Bottom line is it's a necessary step if they want to compete with the 6000 series.
Click to expand...

I agree. I think they disabled some features but the hardware to support it is still there. So basically a 480 is a Tesla with features turned off and vice versa since people who will be using Teslas won't be using them for gaming.

Does Nvidia need to cull the gpu-gpu stuff from Fermi to have it competitive with the 6000 series? Good question. Honestly, GF104 is a big improvement over GF100 in terms of efficiency. How much of that is attributable to 'cutting out'' gpu-gpu features is unknown. But I do think a retuned and improved GF104 could easily serve as the foundation for GF110. I still think the rumored GF110 is just going to be Nvidia's dual-gpu solution and not an entirely new 'Fermi 2.0'. I'm just not convinced that Nvidia can churn out something completely new by Decemeber. And it would be uncharacteristic of them to skip a dual-gpu solution in any generation of video cards.
 
vjcsmoke said:
I still think the rumored GF110 is just going to be Nvidia's dual-gpu solution and not an entirely new 'Fermi 2.0'. I'm just not convinced that Nvidia can churn out something completely new by Decemeber. And it would be uncharacteristic of them to skip a dual-gpu solution in any generation of video cards.
Click to expand...

The problem is AMD is also going to have a dual-GPU card. Not only will they have the 6970 which is supposed to be faster than the GTX 480, they'll also have the 6990 which will be 2 6970s.

Honestly, the GTX 400 series seems like a stop-gap generation. NVIDIA needed to get a DX11 part in the market whether they were ready or not. At that point, they had already promised their investors and AIBs they'd have something out in Q1. There is no doubt GF100 was not what NVIDIA wished it to be, and they probably would've delayed it even more to make it better had AMD not been giving them any meaningful competition.

In addition to the removal of the GPGPU hardware, what about the large amounts of cache on the GF100? My understanding is that the amount of cache is really only useful for GPGPU applications and not games. I'm not sure how much cache could be removed, but that's certain to free up die space as well. With that in mind, doubling texture units shouldn't be a problem at all. I'd be willing to bet with both of those changes, the resulting die size will still be smaller. Keep in mind, as well, that GF100 was NVIDIA's 4770 - it was a learning experience for them. They know a lot more about the process and TSMC supposedly has fixed some of their problems.
 
i was waiting for the new radeon cards (because they have the displayport input which i prefer). but if nvidia have something as good, i would go the nvidia way. no matter how much the newer most fast card is i am, this time, going to buy the fastest.
 
I am looking forward to a GTX 460-based dual-GPU-on-a-card with NVIDIA Surround support and 2GB of GDDR5. This would fulfill my gaming needs as well as be a Folding@Home powerhouse.

If NVIDIA can coax out a die shrink of the current Fermi cards, I see this as a reality.
 
NobleX13 said:
If NVIDIA can coax out a die shrink of the current Fermi cards, I see this as a reality.
Click to expand...

That's not an option until next year. TSMC is not ready. I believe they may have 28nm LP up and running soon, or possibly already, but last I checked a GPU that would still be over 150W is hardly "low power."
 
A dual GTX 460 on a single PCB board would be pretty awesome in terms of performance vs cost. However the power requirements wouldn't be small.. maybe with a GPU refresh that has less power and heat overhead it'll be possible.
 
vjcsmoke said:
I still think the rumored GF110 is just going to be Nvidia's dual-gpu solution and not an entirely new 'Fermi 2.0'. I'm just not convinced that Nvidia can churn out something completely new by Decemeber. And it would be uncharacteristic of them to skip a dual-gpu solution in any generation of video cards.
Click to expand...

It seems people at Beyond3D also thinks the GF110 is going to be a dual-gpu solution:

768 Cude cores
128 TMUs
2GB DDR5 on a 512bit bus
550mm2 to 650mm2
Compared to something about dual chip from AMD?
50% quicker than GTX480

Now look an uncrippled GF104 (GTX460 1GB)
384 Cuda cores
64 TMUs
1GB DDR5 on a 256bit bus

384 Cuda cores x2 = 768 Cuda cores (check)
64 TMUs x2 = 128 TMUs (check)
1GB DDR5 on a 256bit bus x2 = 2GB DDR5 on a 512bit bus (check)

Sounds like a GTX460x2 to me
Click to expand...
http://forum.beyond3d.com/showpost.php?p=1482766&postcount=6897
 
Tamlin_WSGF said:
It seems people at Beyond3D also thinks the GF110 is going to be a dual-gpu solution:


http://forum.beyond3d.com/showpost.php?p=1482766&postcount=6897
Click to expand...
That would explain why there hasn't been any rumors of a new GPU until now and nothing about a new tapeout. This should be an interesting part. Any more than $400 and it wouldn't sell too well due to people just going for 2x460, but at $400 the GTX 480 looks pointless. If Nvidia also came out with a lower end ~$300 version, I could see them retiring the GTX470/480.
 
siliconnerd said:
That would explain why there hasn't been any rumors of a new GPU until now and nothing about a new tapeout. This should be an interesting part. Any more than $400 and it wouldn't sell too well due to people just going for 2x460, but at $400 the GTX 480 looks pointless. If Nvidia also came out with a lower end ~$300 version, I could see them retiring the GTX470/480.
Click to expand...

And, it would be a single card surround solution. :) The 460's have proven to perform well in SLI surround, so here's an easy alternative.
 
Last edited:
If that GF110 roadmap is indeed true, I wonder how powerful that card would be. And not to mention, the power consumption and heat it produces like the GTX 480, if NVIDIA can reduce both on 40nm I would be greatly impressed.
 
i think nvidia should go with accelero heatsinks, they offer awesome cooling,

one is already out, and it offers 40c less than normal gtx480.
 
768 Cude cores
128 TMUs
2GB DDR5 on a 512bit bus
550mm2 to 650mm2
Compared to something about dual chip from AMD?
50% quicker than GTX480

Now look an uncrippled GF104 (GTX460 1GB)
384 Cuda cores
64 TMUs
1GB DDR5 on a 256bit bus

384 Cuda cores x2 = 768 Cuda cores (check)
64 TMUs x2 = 128 TMUs (check)
1GB DDR5 on a 256bit bus x2 = 2GB DDR5 on a 512bit bus (check)

Sounds like a GTX460x2 to me

a dual gf104 = gf110. there is no way nvidia has been working on a new chip without any leaks. things leak at an alarming rate these days.
 
tricky0502 said:
768 Cude cores
128 TMUs
2GB DDR5 on a 512bit bus
550mm2 to 650mm2
Compared to something about dual chip from AMD?
50% quicker than GTX480

Now look an uncrippled GF104 (GTX460 1GB)
384 Cuda cores
64 TMUs
1GB DDR5 on a 256bit bus

384 Cuda cores x2 = 768 Cuda cores (check)
64 TMUs x2 = 128 TMUs (check)
1GB DDR5 on a 256bit bus x2 = 2GB DDR5 on a 512bit bus (check)

Sounds like a GTX460x2 to me

a dual gf104 = gf110. there is no way nvidia has been working on a new chip without any leaks. things leak at an alarming rate these days.
Click to expand...

I disagree, and all of you guys are forgetting the obvious. Since the beginning of time graphic cards have been named. That name could mean anything, e.g. dual gtx 460=580. However, that is not the case here IMO. What is on it's way is a NEW GPU named GF110. Nvidia doesn't give a internal GPU name to a dual card part. As that would be GF 104 X2.

So speculation can be that it will be a uncrippled GF104. However ALL the rumors are saying that the GF110 (not the GTX 580) will be a 512SP part with 128TMU's a 512bit memory bus and better power draw etc.. This would not make it a uncrippled GF104 as that chip would be 1/2 a gf110. So with plain common sense you can apply to the scenario leaving one conclusion. GF110 is a SINGLE High End GPU

This chip will be high end, due to the rules that have been followed since the original geforce and ati cards
 
There is no way they are going to make a 768 core part with a single GPU. I'm calling bullshit on that rumor right now. The more plausible rumor is the original 512 core rumor, with extra TMUs. It's possible they may add a few more shaders, but 256 more on the same process? No.

AMD is going to have a dual GPU part as well. If NVIDIA needs a dual GPU part to compete with AMD's single GPU solutions, it's not going to turn out well. I'm pretty sure GF110 is a single chip, though it's plausible we may see a downclocked variant of that on a GTX 595.
 
not only 768 sp part but also add a 512mb memory bus? all at 40nm. right.
 
I certainly don't believe the 512 bit rumor either. just doubling the tmus is likely going to make it a bigger more power hungry chip then gf100. really there is no need for 512bit if they can get the issues with the controller taken care of and actually run gddr5 at its available speeds.
 
Lord_Exodia said:
I disagree, and all of you guys are forgetting the obvious. Since the beginning of time graphic cards have been named. That name could mean anything, e.g. dual gtx 460=580. However, that is not the case here IMO. What is on it's way is a NEW GPU named GF110. Nvidia doesn't give a internal GPU name to a dual card part. As that would be GF 104 X2.

This chip will be high end, due to the rules that have been followed since the original geforce and ati cards
Click to expand...

Yes, the GF110 must refer to a new chip, not a Dual Card.

I was fairly skeptical about the 768 SP part at first - but after reading the 4gamer article above, I'm a lot less skeptical about it ... another massive chip, but NV may have learned enough by now to handle it....
 
cannondale06 said:
I certainly don't believe the 512 bit rumor either. just doubling the tmus is likely going to make it a bigger more power hungry chip then gf100. really there is no need for 512bit if they can get the issues with the controller taken care of and actually run gddr5 at its available speeds.
Click to expand...

The removal of the GPGPU equipment and possibly some cache will help. I think that's been discussed in this thread.
 
Lord_Exodia said:
I disagree, and all of you guys are forgetting the obvious. Since the beginning of time graphic cards have been named. That name could mean anything, e.g. dual gtx 460=580. However, that is not the case here IMO. What is on it's way is a NEW GPU named GF110. Nvidia doesn't give a internal GPU name to a dual card part. As that would be GF 104 X2.

So speculation can be that it will be a uncrippled GF104. However ALL the rumors are saying that the GF110 (not the GTX 580) will be a 512SP part with 128TMU's a 512bit memory bus and better power draw etc.. This would not make it a uncrippled GF104 as that chip would be 1/2 a gf110. So with plain common sense you can apply to the scenario leaving one conclusion. GF110 is a SINGLE High End GPU

This chip will be high end, due to the rules that have been followed since the original geforce and ati cards
Click to expand...

512bit bus just isn't happening, I'm calling BS on that right now. Maybe for their next gen, but not for a refresh.

The rumors IMO sound like a fanboys wishlist more than something that has any possibility of actually coming out this year.

I'm not saying there won't be a GF110 or that it won't be out this year, I just don't think it will be what the current rumors (and you) hope it will be.

512sp? Sure, I can see that.
128 TMU? Eh... pushing it, but OK
Better power draw? Let's see.. bigger chip but on the same process? Nope, this one isn't happening (at least not in any meaningful amount)
512bit bus? lol wut?
 
kllrnohj said:
512bit bus just isn't happening, I'm calling BS on that right now. Maybe for their next gen, but not for a refresh.

The rumors IMO sound like a fanboys wishlist more than something that has any possibility of actually coming out this year.

I'm not saying there won't be a GF110 or that it won't be out this year, I just don't think it will be what the current rumors (and you) hope it will be.

512sp? Sure, I can see that.
128 TMU? Eh... pushing it, but OK
Better power draw? Let's see.. bigger chip but on the same process? Nope, this one isn't happening (at least not in any meaningful amount)
512bit bus? lol wut?
Click to expand...

Yes, the 512-bit one does sound extreme, but, then it is NV we are talking about ....

If you accept that there is a GF110 coming, then it has to offer a meaningful upgrade from the current GF100, or at least offer a significant benefit....
- so, the 3 or 4 options presented by www.3dcenter.org (www.3dcenter.org/news/2010-10-13) seem to be reasonable guesses/rumors

Of course, at most, only one of those will be correct...
;)

The 3 GPC option certainly looks do-able, if they can figure out a way round the awkwardness of having 3 GTCs
So, maybe we could have
GF110 = 4 GPC = 768 SP, 512-bit
GF112 = 3 GPC = 576 SP, 384-bit

The 3 GPC part would end up smaller, faster & lower power than the GF100
 
Last edited:
kllrnohj said:
512bit bus just isn't happening, I'm calling BS on that right now. Maybe for their next gen, but not for a refresh.

The rumors IMO sound like a fanboys wishlist more than something that has any possibility of actually coming out this year.

I'm not saying there won't be a GF110 or that it won't be out this year, I just don't think it will be what the current rumors (and you) hope it will be.

512sp? Sure, I can see that.
128 TMU? Eh... pushing it, but OK
Better power draw? Let's see.. bigger chip but on the same process? Nope, this one isn't happening (at least not in any meaningful amount)
512bit bus? lol wut?
Click to expand...

Fermi was supposed to have 512SP/64TMU's and a 512 bit memory bus. Unfortunately at that time the TSMC and nvidia couldn't fit all the gpgpu stuff and cache parts in the gpu with all of that and had to scale the gpu back to 480SPs. The Memory controller also didn't come out correctly and there were issues there that nvidia had to debug for months and eventually cut that back to 384bit bus and pair the gpu's with slower ram and more relaxed timings, so the GPU got delayed until March of this year.

It's my belief that from this time nvidia began working on fixing this. I believe that was the birth of the engineering design's for the GF110 gpu. Kyle knew about it from June, just 3 months after the launch of gf100. I'm thinking that this gpu has been in the works since march or perhaps december giving them plenty of time to make these rumors true.

I've said before, in my own thread that I think the GPU will arrive in January but the rumors say December. I'd concede to say that depending on when they actually decided to work on the GF110 will probably depend on when they will release the newer cards and some of these rumored specs may be inflated. I hope whatever it turns out to be, it turns out to be something very competitive with AMD's hardware, as when both companies fight their hardest it's us gamers who win the most.
 
Last edited:
Lord_Exodia said:
So speculation can be that it will be a uncrippled GF104. However ALL the rumors are saying that the GF110 (not the GTX 580) will be a 512SP part with 128TMU's a 512bit memory bus and better power draw etc.. This would not make it a uncrippled GF104 as that chip would be 1/2 a gf110. So with plain common sense you can apply to the scenario leaving one conclusion. GF110 is a SINGLE High End GPU

This chip will be high end, due to the rules that have been followed since the original geforce and ati cards
Click to expand...

There are several rumors and they originate from Chiphell I think (where some of the wrong and right leaks for 6000 series came from).

That the GF110 being a X2 card doesn't exclude that the GF110 is a GF104 GPU refined.

GTX 495 dual GPU:

Based upon 384SP GF110 which would be a fully enabled improved GF104.
GF110 X 2
384SPX2=768SP
512bit bus
64TMUx2=128TMU

Fits the glove more then having a new single GPU part with 512SP, 512bit bus, 128TMU. Nvidia had issues enough getting out a 512SP Fermi, so I seriously doubt that they would get one out and have increased the bus and doubled TMU's at the same time + lowered the power usage.
 
Tamlin_WSGF said:
There are several rumors and they originate from Chiphell I think (where some of the wrong and right leaks for 6000 series came from).

That the GF110 being a X2 card doesn't exclude that the GF110 is a GF104 GPU refined.

GTX 495 dual GPU:

Based upon 384SP GF110 which would be a fully enabled improved GF104.
GF110 X 2
384SPX2=768SP
512bit bus
64TMUx2=128TMU

Fits the glove more then having a new single GPU part with 512SP, 512bit bus, 128TMU. Nvidia had issues enough getting out a 512SP Fermi, so I seriously doubt that they would get one out and have doubled the bus and TMU's at the same time + lowered the power usage.
Click to expand...

What your missing is that the rumors speak of a gf110 gpu that has 512 Stream Processors/512bit bus/128TMU's. A refined and fully uncrippled GF104 would be less than this. Nvidia doesn't designate a internal name for a dual chip and neither has AMD. The GTX 580 name was a guess.

There is very little chance the GF110 is a dual GPU, not to say that they will make a lower end GF110 or a uncrippled GF104 and use it for a dual chip solution to battle antilles.

The GF110 will be a single high end chip and that's what all the common sense adds up to. A dual GF110 wouldn't be possible in PCIe specs. At least not at 40 nm IMO. I think the 768sp part is a bit far fetched.

So I believe 512SP/512bit bus/128TMU GPU will be GF110, and as mentioned before the birth of the GF104 was attributed with removing some of the gpgpu stuff and cache. Perhaps the same is happening here. A gf100 would have enough space to fit this technology had the gpgpu stuff and cache been removed. Perhaps that is the case here, the last thing to make this feasable would be if nvidia had enough time to make it happen and I believe they did. December seems like a stretch to me, perhaps Janurary
 
The 512sp part of the rumor might only be because people are expecting a 512sp GTX480 to come out.

I think the GF110 will be seen on the market as a dual-GPU solution like GTX 495. :)
 
Tamlin_WSGF said:
The 512sp part of the rumor might only be because people are expecting a 512sp GTX480 to come out.

I think the GF110 will be seen on the market as a dual-GPU solution like GTX 495. :)
Click to expand...

I bet you a 6 pack that it'll be a high end chip ;) Again there would be no GF1xx name for it, if it were a dual chip. ;)

Oh I'll go as far as saying that there will be several derivatives of it but no dual GF110. Dual GF104 possible, dual uncrippled gf 104 is pushing it but possible. Dual GF 110 not possible
 
Lord_Exodia said:
I bet you a 6 pack that it'll be a high end chip ;) Again there would be no GF1xx name for it, if it were a dual chip. ;)
Click to expand...

A 384SP chip would be a high end chip. ;)
I'm not talking about a GF110 dual chip, but a 384SP GF110, which will be part of a dual-gpu product (GF110X2).
 
Tamlin_WSGF said:
A 384SP chip would be a high end chip. ;)
I'm not talking about a GF110 dual chip, but a 384SP GF110, which will be part of a dual-gpu product (GF110X2).
Click to expand...

The GTX 480 already has 480 SP so 384 would be less. Maybe if it had 384SP 128TMU 512bit memory bus/uber fast gddr 5 it would be high end but I doubt it because I dont see that as much competition for 69xx series from the rumored specs. That would be what, like only 20% faster than a GTX 480 It makes more sense for 512SP/512bit bus/128TMU with Uber fast ram as GTX580 and a more cut down and feasable chip to be built as a dual gpu solution to be the rival of antilles.
 
If the "refresh" actually has some architectural organization changes the "lower power usage" could be possible. One of the issues with the gtx480 is the voltage required to get the chip operating at an acceptable clock. If they were able to fix some of the issues they originally had with manufacturing (crosstalk, leakage, etc) I could see them releasing a full fledged chip that ran at a much lower voltage.
 
Lord_Exodia said:
The GTX 480 already has 480 SP so 384 would be less. Maybe if it had 384SP 128TMU 512bit memory bus/uber fast gddr 5 it would be high end but I doubt it because I dont see that as much competition for 69xx series from the rumored specs. It makes more sense for 512SP/512bit bus/128TMU with Uber fast ram as GTX580 and a more cut down and feasable chip to be rival of antilles.
Click to expand...

They need to get the dual card according to specs, which means it must be under 300W. Somehow, I see that as highly unlikely with a GTX 480. ;) With an improved GF104, they can do it.

It doesn't make sense with a 512SP/512bit bus/128TMU single GPU with uber fast ram even as a halo card. Think of how Nvidia has struggled with both yeilds and power/temp on the GTX 480. They had to disable cores to get it out on the market. I doubt they will invest R&D money into such. Most likely, we'll see an improved GTX 460, used in a dual GPU variant and perhaps a single chip variant of the same as a GTX 475 or something. Then it would follow what Nvidia has done in the past.
 
Last edited:
Tamlin_WSGF said:
They need to get the dual card according to specs, which means it must be under 300W. Somehow, I see that as highly unlikely with a GTX 480. ;) With an improved GF104, they can do it.

It doesn't make sense with a 512SP/512bit bus/128TMU single GPU with uber fast ram even as a halo card. Think of how Nvidia has struggled with both yeilds and power/temp on the GTX 480. They had to disable cores to get it out on the market. I doubt they will invest R&D money into such. Most likely, we'll see an improved GTX 460, used in a dual GPU variant and perhaps a single chip variant of the same as a GTX 475 or something. Then it would follow what Nvidia has done in the past.
Click to expand...

I think one of us is spot on. Still willing to bet a 6 pack if your up to it. Remember though, nvidia has had enough time to do this. They even had the GTX 460 as practice as far as stripping gpgpu and removing cache to refine the fermi gpu for more performance/watt. I dont think this is as far fetched as your thinking especially if they've been at it since December 09 or March 2k10 and they are releasing 2 months from now in december. That's enough time for several things to be possible. At the worse case it was announced when kyle saw it on or before mid June leaving at least 6 months for the process. ;)
 
Tamlin_WSGF said:
They need to get the dual card according to specs, which means it must be under 300W. Somehow, I see that as highly unlikely with a GTX 480. ;) With an improved GF104, they can do it.
Click to expand...

None of the rumors are pointing to a an improved GF104 in the short term

All of the rumors are saying 512+ SPs.

There were plenty of rumors of a 2x GF104 card back when the GTX460 was first released
- in fact, I'm fairly certain there was an article saying that the Dual GF104 card was ready to go, but was stopped by NV
Here's the rumor link to that one, could be total BS of course ...
http://translate.google.co.uk/trans...tu-gf110-grafiikkapiiri-suunnitteilla&act=url


If that was true, then that probably means they had a Dual GF104 card as a back-up plan, but have decided they don't need/want it....
 
Last edited:
Lord_Exodia said:
I think one of us is spot on. Still willing to bet a 6 pack if your up to it. Remember though, nvidia has had enough time to do this. They even had the GTX 460 as practice as far as stripping gpgpu and removing cache to refine the fermi gpu for more performance/watt. I dont think this is as far fetched as your thinking especially if they've been at it since December 09 or March 2k10 and they are releasing 2 months from now in december. That's enough time for several things to be possible. At the worse case it was announced when kyle saw it on or before mid June leaving at least 6 months for the process. ;)
Click to expand...

One of us is spot on or both are wrong (or maybe the GF110 gets canceled like the GT212 and nobody's right?). Roadmaps change all the time and we are dealing with rumors. :)
You mention the GTX 460 and what they stripped to get power/size down. Don't forget that they still had to disable 48 cores on it.
Nvidia first "showed" (wooden screws) a supposedly 512 core Fermi 1st (or 3rd) oct 2009. One can reckon they've been at it long before that, trying to make a 512 core. Now its 1 year ago, since JHH showed the first 512 Fermi prototype. There are chances we'll see one, but I doubt it after all this time.
The GTX 480 and 470 are not selling as well as expected, and the total sale is low for Fermi. I think the last numbers were only 10% DX11 market share for Fermi and its been out for 6-7 months already. The GTX 460 is well received and is a better halo card then GTX 480.
Having an improved version of GTX 460 (GF110/GTX475) and a dual version (GTX495) as Nvidia have done before, makes more economical sense. Nvidia is working hard on getting Kepler as fast out as possible. :)

samurai1999 said:
None of the rumors are pointing to a an improved GF104 in the short term

All of the rumors are saying 512+ SPs.
Click to expand...
Here's a collection of the rumors:
http://www.3dcenter.org/news/2010-10-13

GF100b with 512 shaders
GF110 with 512 shaders
GF110 with 576 shaders
GF110 with 768 shaders

I must press that we are talking about rumors in this thread and not confirmed specs. Therefore, I would take any specs with a grain of salt. :)
 
You must log in or register to reply here.
Back
Top