Inquirer: We Have NV40 Sample...

mreman4k

Newcomer
First hands-on look at NV40

CeBIT 2004 Shocking power requirements


By Paul Dutton: Friday 19 March 2004, 16:33

THE INQUIRER has managed to get hold of one NV40 sample.
We agreed not to photograph what we saw, but can confirm the following.

When removing the heatsink, we noted that the NV40 GPU itself was, in comparison to what we have seen before, quite large.

However whilst our earlier information indicated that it comprised of 205 -210 million transistors, our sources currently say that what we were holding was a 175 million transistor part.

With the heatsink removed we could see that a small formed surrounding was employed to assist leveling of the heatsink.

Unlike the ill fated GeForce FX 5800 Ultra (NV30) where a heatsink on the back of the PCB could also interface with a small section of the PCB directly behind the GPU as well as the memory modules, additional circuitry in this area on the NV40 prohibits this.

Whilst a heatsink still seems necessary to assist thermal control of the new lower voltage memory modules, the heatsink on the front of the GPU will have to do all the work.

Eight 32MB video memory modules are installed – four on each side of the PCB and as we first reported from Computex in Taiwan – these are of the new GDDR3 type.

However these were not from Micron, which at that time seemed to be both ATi and Nvidia’s GDDR3 development partner. Instead, Samsung seems to be the preferred supplier of the 256MB we saw installed.

The primary shocker was that the board requires two large 4-pin power connectors as opposed to the single power connector on current high-end products.

Perhaps a lot of end-users who have been thinking that they might wait and upgrade to NV40 when it becomes available, should start saving their sheckles to upgrade to a new power supply unit as well.

One final thing, contrary to previous speculation, we think that ‘NV40’ that will come to market may be what was originally planned, as NV45 is, we’re told, the PCI Express part.

Whether that has an AGP to PCI Express bridge chip remains to be seen.

Full article

175 mil transistors? :?:

Two Molex Connectors?? :?:
 
Moving my comment from another thread, because (of course) there is duplication. So confusing to get things in the "right place." Heh...

At any rate, could this make other current scuttlebutt, then, and be the part--the original NV40--that they will launch soon to not lose ground to the R420, and their refresh would be delayed the length many were predicting to become a "bigger refresh" later? It was certainly something I was contemplating myself, and seems to have gotten a bit more confirmation on that front.

Fits the overall picture better, but it certainly doesn't reduce the confusion! :p
 
Difficult to know what's going on here, because the power requirements are more in keeping with the 16 pipe card, yet everything else points at the old NV40. Either:

1) The Inquirer was talking to the wrong people
2) The Inquirer has misreported what they heard
3) That's just an old sample they had to show people at the show
4) They're planning to use the old NV40 in the lineup somewhere
5) There is no new NV40 (doubtful - too many knowledgable people have confirmed it)

I suspect it's either 3 or 4. I doubt that The Inq would have gotten hold of samples shown behind closed doors, so maybe the old NV40 is just floating around on the main floor for non-NDA Journo's to look at.

There's some "behind closed doors" action going on from nVidia at the show, so draw your own conclusions on how much this matters.
 
Two molex connectors?

That doesn't make any sense! One will work just as well as two. The only worry would be pulling so much current through one wire that you burn out the wire, but I don't think most computer power supplies can actually supply that kind of current.
 
PaulS said:
3) That's just an old sample they had to show people at the show
4) They're planning to use the old NV40 in the lineup somewhere

I'm thinking it's a combo of 3 & 4 too. I think that nVidia realized their "new" nV40 would be too far delayed to fit their announced launch so they're going to go ahead and launch the "original" nV40 to keep pace with ATi's launch.

That would explain why there was first going to be a delay, (they switched to the "new" nV40), and now everyone is telling me I'm a moron for thinking it might just possibly be a little late, (they plan to launch the "original" nV40 just to spite me and make me look foolish and bite any profit losses and offset that by the great joy they'll feel by making me look dumb...it's worth at LEAST a few mil to 'em! ;) ), and now FUaD has one of the "original" nV40s which is what they will soon be releasing.

It makes perfect sense. 8)
 
Chalnoth said:
Two molex connectors?

That doesn't make any sense! One will work just as well as two. The only worry would be pulling so much current through one wire that you burn out the wire, but I don't think most computer power supplies can actually supply that kind of current.

Yeah they can. Burning out traces is a common problem if you pull too much current. It happens all the time when people put too-big fans on their motherboard fan-power headers.

Alterntively, it could be a method of keeping power traces short due to interference issues.
 
Chalnoth said:
Two molex connectors?

That doesn't make any sense! One will work just as well as two. The only worry would be pulling so much current through one wire that you burn out the wire, but I don't think most computer power supplies can actually supply that kind of current.

Indeed. I know a case where friend of mine had his Radeon 9700's power wire burned.
 
Ok this is how i am taking it:

1.) Nvidia is going to launch their original NV40 sooner rather than later

2.) the "new" NV40 that was NV45 will launch a bit later (late spring or summer?)

3.) original NV40 that is still coming out is roughly 175M transistors and may or may not be 16 pipelines (8 or 12 pipes)

4.) NV45 (new NV40) is the 210M transistor, 16 pipeline beast, will launch a bit after the original NV40. or they will both launch, both being part of the NV4x lineup.


ok i'm probably wrong but that's how i read it.
 
Both MuFu and Ailuros have said NV40 AND NV45 have already taped out, and that they did around the same time... MuFu even said something about 18/12/03, but i don´t know if he meant it... so they could still launch NV40 NOW and have NV45 be launched in a couple (not TOO many) months although that would really heart NV40 sales if true, unless they have a reaaaaaalllllly nice retail price, as around $350...
 
Megadrive1988 said:
3.) original NV40 that is still coming out is roughly 175M transistors and may or may not be 16 pipelines (8 or 12 pipes)

16x1, pretty much guaranteed even on the original. (Which I don't really get yet, but that's what I'm hearing.)
 
I saw this, too, and think its purpose is just to admit, in its own inimitable fashion, that Inquirer really doesn't know what it's talking about when it makes such colorful noises about nV40...;) Notice that this time there's no reiteration of any "16-pixel-pipeline" claim, and that their new "source" for nV40 info is telling them this "very large" gpu is 175M transistors, instead of the 205-210M transistors their last "source" told them was required by the "16 pixel pipelines" of the other nV40.

Doesn't it seem odd that they stated they believed that "what they saw before" was ~210M transistors, and yet it doesn't seem incongruent to them that the current "175M transistor" nV40 in this case is "in comparison to what we have seen before, quite large"....? Granted that what they "saw before" might not have been an nV40 at all, but then, why say that it was if it wasn't? Heh...;) You'd think that they'd expect this gpu, with fewer transistors, would be physically smaller than what they "saw before," instead of quite a bit larger.

From the way the account is written, it seems apparent the author didn't actually put the card in a system or run it, or even observe it being run inside someone else's system, so it seems at least possible that what the Inquirer was shown might have been a mechanical mockup--totally non functional--or else maybe an entirely different card altogether--maybe just a different mockup of a different 3d card--and somebody is just having fun pushing the Inquirer's buttons.

It also seems odd to me that although they claimed to have "removed the heatsink" and that although the gpu exposed was "quite large," they were unable to decipher any tell-tale silk-screen printing on the gpu which they might have recorded for everyone's enlightenment. That would have been kind of interesting, I think. So maybe if they didn't see it, it was because whatever had been printed on the gpu had been removed...? (Not hard to do.) It seems from their "we could see that a small formed surrounding was employed to assist leveling of the heatsink" remark that they certainly studied the gpu area long enough to see any printing that might have been present. Just guessing, but it would appear they might have thought that the absence of printing on the gpu was "normal" in the case of prototype cards, but I've seen dummy mockups with silk-screen identifiers printed on the chips.

Also, the lack of a heatsink for the ram chips on the flip side of the PCB would tend to indicate a mockup, very possibly. Speaking of the GDDR3 ram modules, it, again, seems odd that no effort was made to record the ram-chip identifiers, which could have been taken from ram on the rear of the pcb even if not from the front after heatsink removal. This remark seemed cryptic: "However these were not from Micron, which at that time seemed to be both ATi and Nvidia’s GDDR3 development partner. Instead, Samsung seems to be the preferred supplier of the 256MB we saw installed." It tends to make me think they weren't sure that it was GDDR3 at all, as they expected GDDR3 to come from Micron. Odd, to me, that they could bother getting the brand name from the ram chips, but not the numbered identifiers, so that their assumptions might have been verified. The fact that there was, they say, 256mbs onboard the card certainly doesn't prove it was GDDR3.

When I saw this, "The primary shocker was that the board requires two large 4-pin power connectors as opposed to the single power connector on current high-end products," I got a very strong mental image of Ricardo Montalban singing, "VO-lar-E', yo-yo-yo-yo!" and making obscene comments about "Reech Cor-een-thian Leathers-s-s-s-s-sssss"....;)

But then, that wouldn't jive with the "in comparison to what we have seen before, quite large" 175M transistor gpu, which was apparently much larger than what they saw earlier and believed to be a ~210M transistor nV40 gpu with 16 pixel pipelines, would it?....;)

And, to go on, I found it odd that they'd think nVidia might take an nV40, mount it on a PCIe pcb, and call it "nV45," assuming they didn't mean that nVidia was going to wait until nV45 before it would produce a native PCIe 3d reference design...;) Assuming that Inquirer meant that at some point nVidia plans to market a native PCIe reference design for a 3d accelerator, I can only assume they meant to say a "PCIe to AGP x8 bridge chip," instead of what they said, which was "Whether that has an AGP to PCI Express bridge chip remains to be seen."

But that's one of the joys of Inquirer "scoops," isn't it? They are always just so crystal clear...;)
 
mreman4k, its bad form to copy the _whole_ article and then paste it somewhere else. As much as I dont trust/like/... the inq, they do have bills to pay and should be compensated for their articles. You should have just pasted a short quote and had a link to the article.

Please make a note of that. It seems others have done similar things, not sure if b3d.com has a policy against this, if not may be it should be added.

later,
epic
 
digitalwanderer said:
Megadrive1988 said:
3.) original NV40 that is still coming out is roughly 175M transistors and may or may not be 16 pipelines (8 or 12 pipes)

16x1, pretty much guaranteed even on the original. (Which I don't really get yet, but that's what I'm hearing.)


That's what I think. The NV4- has been 16 pipe all along.
They knew the 420 had 12 so they upped them a little. :)
 
bkswaney said:
That's what I think. The NV4- has been 16 pipe all along.
They knew the 420 had 12 so they upped them a little. :)

Oh not THIS AGAINNN...

The NVFans seem to have come out of hiding since autumn/winter 02 before NV30 launched. All we heard then was how it would have 8x2 since R300 was 8x1, and we all know NV is always > ATi... (Well, was back then anyway).

There should be some law against speculating without a foundation in truth to base it on. For frick's sake, we don't even know how many pipes R420 has!
 
Back
Top