NVIDIA Fermi: Architecture discussion

[jimmyjames123]

Anyone speak Turkish?

Here is an interview with Luciano Alibrandi @ NVIDIA:

http://www.donanimhaber.com/Nvidiadan_Luciano_Alibrandi_ile_konustuk-17219.htm

http://www.overclock.net/hardware-news/634148-donanimhaber-nvidia-confirms-dual-gpu-fermi.html

Luciano was formerly a product manager at 3dfx corp!

 

[Simon F]

Anyone speak Turkish?

Only enough so say hello and order a couple of coffees ... and I didn't hear him ask for any

 

[neliz]

What about back to topic?! Launch in February is most likely to Fermi now?!

"Yes" for Launch.
"Uhm..." for availability

Ask any board partner and they will tell you they got the GT200b EOL message from nVidia back in Q3.

 

[Sontin]

Only enough so say hello and order a couple of coffees ... and I didn't hear him ask for any

This will help :

Donanimhaber.com published a new video interview with Luciano Alibrandi form Nvidia.

Some highlights from video;

Alibrandi said;

- Fermi based GPUs will be fastest solutions in the market.
- We are very happy to see internal test results of Fermi.
- Fermi architecture developing for every single segment, dual GPU is also possible for Nvidia.
- We are very happy to work with TSMC but also we are always looking for another oppurtinies.
- ION 2 and Tegra 2 are very near to launch.

http://www.nvnews.net/vbulletin/showthread.php?p=2151031#post2151031
http://vr-zone.com/forums/526529/nv...u-is-possible-and-we-are-happy-with-tsmc.html

 

[Rys]

Yep, but if A3 yields are poor at 1400Mhz, what are the odds that we see Rys' 1700Mhz prediction on Geforce parts?

Have we got any solid indication that they are poor at 1400 MHz for any product yet? I'm still confident about my numbers despite the Tesla product configuration, although I'm waiting for the driver to gain some maturity before worry about the GeForce side of things.

 

[Ailuros]

Would sparse supersampling reduce the blur effect of their current ordered grid approach? It looks pretty but some detail is lost.

4xOGSS wherever present in the unofficial modes (hybrid or pure SSAA) offsets LOD by -1.0 as you'd expect it to. Textures become by a healthy portion sharper without adding texture aliasing. However if you don't mind texture aliasing you can always turn off the clamp and offset the LOD even further through 3rd party applications (uhhhm *yuck*).

In any case the LOD offset usually comes with supersampling irrelevant of sample pattern. Afaik AMD initially didn't have the LOD offset for their RGSS (probably due to the way they enabled it) but there's an option for it in the drivers nowadays.

Only enough so say hello and order a couple of coffees ... and I didn't hear him ask for any

At least some turkish delight should have been on the table LOL

 

[trinibwoy]

Poor for 225W? Then how much breathing room do GeForces have? Do you expect the GeForce GTX 380 —or whatever it's called— to have an even higher TDP?

Why not? GTX 280 was 236W.

No, but I am a little down about it. I do understand why it isn't a big deal, I asked at the Evergreen launch (two days before actually), and the memory controller designers told me why it wasn't a problem. I agree with what they said.

I don't know what they told you that assuaged your concerns but the numbers speak for themselves. Doubled units resulted in only a 50% performance boost. The only feasible explanation would be that the bottleneck lies elsewhere and more bandwidth therefore would not improve performance. I'm still not clear why it's not a big deal for Cypress but it would be for Fermi.

Have we got any solid indication that they are poor at 1400 MHz for any product yet? I'm still confident about my numbers despite the Tesla product configuration, although I'm waiting for the driver to gain some maturity before worry about the GeForce side of things.

I think it's just rjc's link to that chinese forum post so far. But I'm not seeing any reason to be optimistic about Geforce clocks with all the talking Nvidia is doing and the recent revelation that Tesla SKUs don't have a full complement of cores enabled.

 

[aaronspink]

The HPC market is a clear target, now more than ever, because of the high profit margins it provides.

HPC and profit margins in the same sentence. Someone doesn't know much about HPC.

 

[trinibwoy]

HPC and profit margins in the same sentence. Someone doesn't know much about HPC.

Or maybe your "knowledge" is rooted in the old school way of doing things where there is strong competition between the CPU guys. I find it hard to believe that a company could offer considerably higher efficiency hardware and not benefit from that advantage.

Of course Nvidia will offer attractive prices to get a foot in the door but after a while the hardware will speak for itself. Under which scenario do you think GPU ASPs will be constrained in the HPC market if they deliver on the promised efficiency gains over commodity CPUs?

 

[Rys]

I think it's just rjc's link to that chinese forum post so far. But I'm not seeing any reason to be optimistic about Geforce clocks with all the talking Nvidia is doing and the recent revelation that Tesla SKUs don't have a full complement of cores enabled.

Well, we're at the stage now where even if I printed what I knew, it'd come across the wrong way to an excitable few, because Charlie's ruined being able to be (even cautiously) optimistic about the company.

Plus, they're just opening up with both barrels and blowing both feet off on a regular basis with what they put on Twitter and Facebook (please make it stop) and the like, Intel's Insides (that's actually vomit-worthy), and various careless whispers to careless people.

CES will be a laugh!

 

[MfA]

Of course Nvidia will offer attractive prices to get a foot in the door but after a while the hardware will speak for itself. Under which scenario do you think GPU ASPs will be constrained in the HPC market if they deliver on the promised efficiency gains over commodity CPUs?

I think their efficiency is a bit overblown, sure they are great compared to desktop CPUs ... but I doubt it would have been able to beat the PowerXCell 32i, a processor which IBM didn't think worth finishing based on the HPC market alone.

The HPC market no doubt will give NVIDIA some money, but the death of Cell should put the potential in perspective IMO.

 

[jimmyjames123]

Well, we're at the stage now where even if I printed what I knew, it'd come across the wrong way to an excitable few, because Charlie's ruined being able to be (even cautiously) optimistic about the company.

Plus, they're just opening up with both barrels and blowing both feet off on a regular basis with what they put on Twitter and Facebook (please make it stop) and the like, Intel's Insides (that's actually vomit-worthy), and various careless whispers to careless people.

CES will be a laugh!

Well said, pretty much spot on.

I wish NVIDIA would put an end to Intel Insides cartoons. NV would be better off just letting their own products do most of the talking. Also, no point in antagonizing Intel fans (many whom may want to use NVIDIA products in the future). Saying too many negative things about a competitor will not be well-perceived by many consumers, I hope they understand this. At this point, we all know of Intel's monopolistic tendencies, no need to keep going on and on about it.

Yes, CES will be a blast I'm sure

 

[trinibwoy]

I think their efficiency is a bit overblown, sure they are great compared to desktop CPUs ... but I doubt it would have been able to beat the PowerXCell 32i, a processor which IBM didn't think worth finishing based on the HPC market alone.

Well you said it yourself - that was a product targeted at HPC. Tesla on the other hand is heavily subsidized by Geforce so it's a completely different equation.

 

[seahawk]

You mean CES will be a blast, because Fermi will miss it as well.

 

[Rys]

You mean CES will be a blast, because Fermi will miss it as well.

They'll demo Fermi there to press. I think both Quadro and GeForce will get a run out, too (or at least there'll be Quadro-specific demonstrations running on the same hardware, but locked to that product at launch).

 

[Silus]

Plus, they're just opening up with both barrels and blowing both feet off on a regular basis with what they put on Twitter and Facebook (please make it stop) and the like, Intel's Insides (that's actually vomit-worthy), and various careless whispers to careless people.

Not sure if I understood you. Why do you say "blowing both feet off" with what they put on twitter and facebook ? From what you know of Fermi based GeForce's performance, what they are revealing isn't going to save it from disappointment ? Or the other way around ?

CES will be a laugh!

In a good or bad way ?
Come on Rys, you can't throw a bone and then expect us not to go after it

 

[MDolenc]

Not sure if I understood you. Why do you say "blowing both feet off" with what they put on twitter and facebook ? From what you know of Fermi based GeForce's performance, what they are revealing isn't going to save it from disappointment ? Or the other way around ?

I'd say that given all the crap that's flying around it would probably be better to keep quiet till whole thing is ready to launch then feeding small bones to the wolves.

 

[2senile]

I'd say that given all the crap that's flying around it would probably be better to keep quiet till whole thing is ready to launch then feeding small bones to the wolves.

I'd say that just one "industry standard" benchmark, verified by "press", would alleviate all the crap flying around. It would keep the usual suspects occupied for at least three months.

CES will be a laugh!

If the press kit contains a woodworking manual & set of wood chisels, can I have them? [/Charlie mode]

 

[aaronspink]

Or maybe your "knowledge" is rooted in the old school way of doing things where there is strong competition between the CPU guys. I find it hard to believe that a company could offer considerably higher efficiency hardware and not benefit from that advantage.

Of course Nvidia will offer attractive prices to get a foot in the door but after a while the hardware will speak for itself. Under which scenario do you think GPU ASPs will be constrained in the HPC market if they deliver on the promised efficiency gains over commodity CPUs?

If that is true then they aren't competing against commodity CPUs, they are competing against commodity GPUs, which I might remind you have even LOWER margins than cpus. In the HPC space you generally aren't competing against the other guys expensive parts, but against your cheapest parts.

AKA why would I pay nvidia 2K for something they are selling for $300 when its them that want to push it?

 

[Sontin]

AKA why would I pay nvidia 2K for something they are selling for $300 when its them that want to push it?

Because you want the support? They tell eveybody that Quadro and Tesla business is a solution market and that's the reason why they don't see amd as a competitor.
And for the margin: They spoke about 60% for Tesla.