No Fermi benchmarks / Price & TDP revealed

[Lonyo]

I'm actually curious and not trying to troll, but is anyone dissapointed that the 480 doesn't come with the full 512 shaders?

It seems to me that if the HPC or Quadro market is the only one to see the full chip then this shows where NVs bias lies - and it's not with the GeForce line.....

Given the focus of the chip and the fact that the HPC market is much more profitable, it's not a surprise at all from a business point of view. It makes no sense to release your product as a consumer piece for say $500 when you could sell it for maybe $2000+ in the HPC market.

NV's focus lies in returning shareholder value, and with the diminishing marketshare in chipsets, they need to make up for it somewhere else, the somewhere else is Tegra and HPC, so it makes sense for them to push those markets.

While it might seem disappointing, it is a very logical and good decision.

 

[Soleron]

The HPC market isn't getting all 512 shaders either. The official spec documents say 448SP.

 

[tviceman]

Given the focus of the chip and the fact that the HPC market is much more profitable, it's not a surprise at all from a business point of view. It makes no sense to release your product as a consumer piece for say $500 when you could sell it for maybe $2000+ in the HPC market.

NV's focus lies in returning shareholder value, and with the diminishing marketshare in chipsets, they need to make up for it somewhere else, the somewhere else is Tegra and HPC, so it makes sense for them to push those markets.

While it might seem disappointing, it is a very logical and good decision.

This is likely 92% of the reason why they're releasing the gtx480 with 480 cores instead of 512.

The other 8% is that once yields improve a little and after getting B1 stepping, they'll likely release a gtx485 or whatever (with the full 512 shaders and slightly higher clock) to counter the hd5890.

 

[SlowSpyder]

I'm actually curious and not trying to troll, but is anyone dissapointed that the 480 doesn't come with the full 512 shaders?

It seems to me that if the HPC or Quadro market is the only one to see the full chip then this shows where NVs bias lies - and it's not with the GeForce line.....

I don't think anyone would care if it had only 3 shaders enabled if the performance was there. But coming much later than the competition, and rumored to use much more power/put out a lot more heat for barely being a bit faster then what has been on the market for months just leaves people feeling a bit underwhelmed. I don't think it'll be a slow part (5870 performance is pretty good) but I think most of us were just expecting more out of it.

 

[SolMiester]

It's really funny how I used to hear people bashing Microsoft because their stuff never worked or it never worked properly or it was difficult to use...and now, in the light of the ABSOLUTELY EMBARRASSING FIASCO that NVidia have made the Fermi launch, I can't help but smile. Good God, if any other company in any other sector had made such a complete and utter cock-up of a product launch, their names would have been forgotten by the end of the year. That's why companies avoided this sort of total fuckfest like the plague. Because you're just not supposed to come back from this sort of thing. Being 6+ months delayed in getting product out to customers in an industry where product cycles are only 6 months long, having an NDA that won't lift until 3 days after the official launch for customers who crave hard numbers and comparisons, all this embarrassing PR maneuvering and damage control, deception, even outright lying, FOR FUCK'S SAKE!

<snip>.

I could say the same about ATi...18xx series!

 

[dguy6789]

The X1800 series was late, but when it launched it was as good or better than the 7800GTX in most benchmarks, had double the ram, and could be purchased for less money. The following X1900 that launched was faster than the 7800, X1800, and the 7900 that launched later.

 

[cbn]

In post #230 I am reading about the NDA not lifting till 3 days after the launch date?

I thought reviews were going to be out on the 26th....of this month?

 

[MrK6]

In post #230 I am reading about the NDA not lifting till 3 days after the launch date?

I thought reviews were going to be out on the 26th....of this month?

I believed they were delayed until after PAX ended (or something like that). Very curious timing...

 

[SolMiester]

The X1800 series was late, but when it launched it was as good or better than the 7800GTX in most benchmarks, had double the ram, and could be purchased for less money. The following X1900 that launched was faster than the 7800, X1800, and the 7900 that launched later.

x1800 was crap....the x1900 got it right...

 

[SolMiester]

The X1800 series was late, but when it launched it was as good or better than the 7800GTX in most benchmarks, had double the ram, and could be purchased for less money. The following X1900 that launched was faster than the 7800, X1800, and the 7900 that launched later.

Also, you sound like you are saying that Fermi will not be as good or better than the 58xx series on launch when we already know it is, it also has more RAM and we dont know about price until we can buy!

 

[dguy6789]

x1800 was crap....the x1900 got it right...

What do you mean exactly? It wasn't absolutely faster in all tests or anything nor was it some legendary card by history's standards, but it competed just fine.

http://www.anandtech.com/showdoc.aspx?i=2552&p=10
http://www.anandtech.com/showdoc.aspx?i=2556&p=2

Benchmarks show it very competitive with the 7800GTX. Now keep in mind that at this time, the X1800 had 512MB of memory(7800GTX has 256MB) and was cheaper than the 7800GTX. The only mark against it was that it took a couple months for it to come out after the 7800 did, but nowhere near as long as Fermi is taking. The X1800 was also a full hard launch, availability wasn't an issue when it did come out.

Also, you sound like you are saying that Fermi will not be as good or better than the 58xx series on launch when we already know it is, it also has more RAM and we dont know about price until we can buy!

My post had nothing to do with Fermi. Just discussing history.

 

[Narynan]

5-10%? Where are you getting that? At what resolution? DX10 or DX11? Is this an average across all games that you have seen benchmarks on? Is this the stock version or OC'd version you are talking about? Which drivers were used in these benchmarks?

470 is looking about the same as a 5870, so I would doubt the next card up is only 5-10% faster.

Lets see.
Same as a 5870? """"Where are you getting that? At what resolution? DX10 or DX11? Is this an average across all games that you have seen benchmarks on? Is this the stock version or OC'd version you are talking about? Which drivers were used in these benchmarks?""""

Let's not be so quick to jump on someone here.

 

[betasub]

I could say the same about ATi...18xx series!

Shame that you are over four years too late to point this out

Still, nice distraction tactic. Discussing ancient history (enjoyable enough) instead of focusing on Fermi.

 

[tviceman]

Shame that you are over four years too late to point this out

Still, nice distraction tactic. Discussing ancient history (enjoyable enough) instead of focusing on Fermi.

Agreed about all the talk of past releases. History may be a good indication as to how in the future AMD/Nvidia can respond to adverse conditions, but really it serves no purpose right now.

Bring on the DX11 benchmarks and lets see if Fermi really is a forward looking architecture that Tom Peterson said it is (in gtx480 heaven/3dvision demo on youtube).

 

[Hauk]

Hmm, Voltage Tweak from Asus. Perhaps there will be overhead

 

[dzoner]

Given the focus of the chip and the fact that the HPC market is much more profitable, it's not a surprise at all from a business point of view. It makes no sense to release your product as a consumer piece for say $500 when you could sell it for maybe $2000+ in the HPC market.

NV's focus lies in returning shareholder value, and with the diminishing marketshare in chipsets, they need to make up for it somewhere else, the somewhere else is Tegra and HPC, so it makes sense for them to push those markets.

While it might seem disappointing, it is a very logical and good decision.

Why would Nvidia need a 512 shader chip over a 480 shader chip to 'push' the HPC market? Will that determine Nvidia's reputation in that market? Isn't it in the gaming market where this is important and they most need such a chip? Isn't it vital Nvidia ekes out the last erg of performance lead over the 5870 when they release their halo gaming card when it is uncompetitive in every other metric?

Perhaps the reported extreme scarcity of viable 512 shader chips is the logical reason for releasing a 480 shader halo card and their decision was neither 'good' nor 'bad', but simply that they had no other choice.

 

[SlowSpyder]

Why would Nvidia need a 512 shader chip over a 480 shader chip to 'push' the HPC market? Will that determine Nvidia's reputation in that market? Isn't it in the gaming market where this is important and they most need such a chip? Isn't it vital Nvidia ekes out the last erg of performance lead over the 5870 when they release their halo gaming card when it is uncompetitive in every other metric?

Perhaps the reported extreme scarcity of viable 512 shader chips is the logical reason for releasing a 480 shader halo card and their decision was neither 'good' nor 'bad', but simply that they had no other choice.

I think that seeing as Nvidia released the specs some time ago and they said 512 SP's at the time, I think this gives you a good idea of yields. They have to release a slightly crippled part with the 512 SP part coming in future. If yields are terrible and continue to be bad, and Fermi is just competing with the 5870 and 5850, I wouldn't be shocked if we never see the 512 SP part. 295 watt TDP with a part that yields very poorly and therefore cost a lot to make might not make sense. They may just get these out to show up (like AMD had to do with the 2900XT) with something while they are busy working hard on Fermi 2.

The 2900XT was a poor part compared to the competition, but it gave AMD a great platform to work from. Who would have thought back when the 2900XT was released that a few years later AMD would be kicking Nvidia's ass from the enthusiast standpoint with parts that are based off of the 2900XT? I think Fermi will be similar for Nvidia, right now things look poor, but as they refine the architecture and build next gen parts based off of it, I see them being right back in the thick of things.

 

[evolucion8]

Like Slowspyder said, Fermi will be the next HD 2900XT, but then will be the next HD 3800 series and will be great, ATi is working on a new architecture, so I won't be surprised that it might be the next G80. nVidia and ATi likes to mirror each one swapping roles like in a movie, odd.

 

[ronnn]

Agreed about all the talk of past releases. History may be a good indication as to how in the future AMD/Nvidia can respond to adverse conditions, but really it serves no purpose right now.

Bring on the DX11 benchmarks and lets see if Fermi really is a forward looking architecture that Tom Peterson said it is (in gtx480 heaven/3dvision demo on youtube).

Past history suggests optimized drivers with a nice shimmer. Benchmarks may or may not clarify the situation.

 

[McCartney]

What do you mean exactly? It wasn't absolutely faster in all tests or anything nor was it some legendary card by history's standards, but it competed just fine.

http://www.anandtech.com/showdoc.aspx?i=2552&p=10
http://www.anandtech.com/showdoc.aspx?i=2556&p=2

Benchmarks show it very competitive with the 7800GTX. Now keep in mind that at this time, the X1800 had 512MB of memory(7800GTX has 256MB) and was cheaper than the 7800GTX. The only mark against it was that it took a couple months for it to come out after the 7800 did, but nowhere near as long as Fermi is taking. The X1800 was also a full hard launch, availability wasn't an issue when it did come out.



My post had nothing to do with Fermi. Just discussing history.

You're mistaken about the x1800 getting a full launch, that was the x1900. The X1800 (i had it) was a soft launch that didn't have massive availability on shelves. The x1900 just came out a month or 2 after and was widely available.

 

[hectorsm]

10&#37; faster "internally" is embarrassing for as long as it took, size & power. ATI should be able to counter for parity fairly easy. Well nV will selll a lot because they have a good rep but I'm sticking with my old HW for now.

Agree. Plus castrating the GTX480 to 480sp after they announced it to be 512sp is even more embarrassing if it turns out to be true.

 

[tviceman]

295 watt TDP with a part that yields very poorly and therefore cost a lot to make might not make sense.

You're only guessing/assuming that an unreleased, announced, 512 core GF100 will have a 295w tdp and will, in the future, be yielding "poorly", right?

 

[MrK6]

You're only guessing/assuming that an unreleased, announced, 512 core GF100 will have a 295w tdp and will, in the future, be yielding "poorly", right?

Well 1 + 1 = 2, so yeah.

 

[SlowSpyder]

You're only guessing/assuming that an unreleased, announced, 512 core GF100 will have a 295w tdp and will, in the future, be yielding "poorly", right?

I was saying that with poor yields (seeing as Nvidia has to release a crippled part as the top sku to get Fermi to the market, assuming this information is correct) I wouldn't be shocked if we never see the 512 SP part. If yields improve to the point that it makes sense financially, then of course we'll see it. But, if yields for this monster sized chip on this realatively new, and so far problematic process don't improve, I doubt the 512 SP part will ever show up as Nvidia will rather get Fermi 2 out the door.

 

[SolMiester]

What do you mean exactly? It wasn't absolutely faster in all tests or anything nor was it some legendary card by history's standards, but it competed just fine.

http://www.anandtech.com/showdoc.aspx?i=2552&p=10
http://www.anandtech.com/showdoc.aspx?i=2556&p=2

<snip>

IIRC, the drivers were awful the the benchmarks you now link to are from the updated drivers.....
It was nowhere near the 7xxx series when it came out! unless i'm losing my mind!