NVIDIA GeForce 4XX Series Discussion

[Bo_Fox]

120hz=120FPS, i thought, because the frames are "shifted", which gives the 3d perspective? The glasses literally oppose another in contrast, so you only really see with one eye at a time?

And thanks for the well wishes...she'll be fine, but the experience is emotionally draining. This is her second of 4 surgeries, and none are that big of a deal, but I still worry, of course, as any parent would.

And yeah, I'm pretty lucky, having a wife that enjoys gaming as much as I do. The competitiveness is a good stress reliever for both of us, and it's easy to get involved in a game, and forget life for a bit. But because she likes alot of things I don't, I get the chance to see alot of different games, and once the kids are thrown in, there's little we haven't played...I'm the one that gets stuck playing the games before the kids do, so we can be sure of the content they consume.

It also allow me to have multiple systems, and no problems with spending the money on new stuff when it comes out...but at the same time, I've definately become highly opinionated when it comes to buying stuff, and it's always good fun to see new stuff come, as it's more toys to play with for my entire family.


The fact that ATI and nVidia currently differ so greatly in thier offerings means alot to me. there more difference there is, the more entertainment I have. When I don't drive, all that extra cash that would have been spent on insurance, or gas, goes to gaming, so dumping $600 on some monitors, or a monitor and glasses, is no big deal to me...as long as it works. If it doesn't, then in the end, I say so. I mean...eyefinity works, I guess, but it's not enough for me.

I don't want Fermi to bring the same disappointment. And that's why it matters to me so much...I may be logging hundreds of hours using it.

That's cool getting to know a little about you! Once again, best wishes for your daughter. My wife also likes to do lots of stuff and she really appreciates the kick-ass rig that I build for her (see my system specs--it's the 2nd rig).. whenever her friends see it, their jaws drop. It's because of the aesthetics and stuff, but I try to not spend too much $$ on it.. I'm also very picky with hardware. After so many headaches with SLI, I'm glad to be a retired SLI veteran, which is another reason why I really look forward to the GT300.

My boys also like playing games (they're just old enough to finally start reading). Spore is probably their favorite game on the PC, by far!

About the frame rate and refresh rate.. those are two different things. If a game is running at 60fps, and your monitor is displaying 120Hz, it means that each frame will be drawn twice (at least on a CRT monitor). With the shutter glasses, you'd still be seeing pretty much every frame (with 120 refreshes split between the eyes). It would still look fluid, although I still notice a considerably more fluid motion with 85fps (at around 90fps, it's almost as fluid as it can get for your eyes but hey, you're right that 120fps is even better anyways).

All the more reason to get a GT300! Maybe it would actually support 32x AF?!?

 

[qubit]

Nvidia to showcase Fermi solutions to mad scientists only

Source

 

[Binge]

thanks for old news? That's footage from the keynote.

 

[Bo_Fox]

Come on, come on, Fermi!!! I'm soooo tired of waiting!! ~2 months left countdown!

 

[SNiiPE_DoGG]

Fermi = (3/16(√2)π)^2/3(h^2/m)(n^2/3)


had a chuckle while going over my E/M cribsheet

 

[qubit]

Fermi f*cked?

IT LOOKS LIKE we were right about Fermi being too big, too hot, and too late, Nvidia just castrated it to 448SPs. Even at that, it is a 225 Watt part, slipping into the future.

.............


Fermi - Now less than promised

The rest however is Nvidia's fault. It designed a chip that was more or less unmanufacturable, and we have been saying as much for more than six months now. It is big, 530mm^2 at the minimum, likely 10+mm^2 more, and way too hot. The 448 SP version is listed at 225W TDP, 190W 'typical', and that is with 14 of the 16 shader clusters clusters fused off. With them on, it would likely be well above 250W, far too hot for a single GPU card.

Then there is the whole problem of 512 SPs promised versus 448 SPs delivered. Fermi is arranged as 16 clusters of 32 shaders, and given that it is turning off 64 shaders, it looks like the minimum granularity it can fuse off is a single cluster of 32. This means it is having problems getting less than two unrecoverable errors per die, not a good sign.

Charlie Dhttp://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/http://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/

This is a shame, a really big shame. :shadedshu

 

[Zubasa]

I wouldn't jump on this until I see the real thing.

 

[yogurt_21]

I wouldn't jump on this until I see the real thing.

I wouldn't jump on this anyway. 448 is still a rather high number for sp's especially compared to the 240 on the current gtx285.

I'd rather see a final product and see how it performs. And if they took some away now it just means well see them back and more refined in the revision cards.

at this point I'd rather them get a working card out than try for something they could always do later.

edit: even if the rest of the specs were true it could be were looking at the gtx360 specs and a made up tdp. hard to tell with so little infor from nv.

 

[Benetanegia]

IT LOOKS LIKE we were right about Fermi being too big, too hot, and too late, Nvidia just castrated it to 448SPs. Even at that, it is a 225 Watt part, slipping into the future.

.............


Fermi - Now less than promised

The rest however is Nvidia's fault. It designed a chip that was more or less unmanufacturable, and we have been saying as much for more than six months now. It is big, 530mm^2 at the minimum, likely 10+mm^2 more, and way too hot. The 448 SP version is listed at 225W TDP, 190W 'typical', and that is with 14 of the 16 shader clusters clusters fused off. With them on, it would likely be well above 250W, far too hot for a single GPU card.

Then there is the whole problem of 512 SPs promised versus 448 SPs delivered. Fermi is arranged as 16 clusters of 32 shaders, and given that it is turning off 64 shaders, it looks like the minimum granularity it can fuse off is a single cluster of 32. This means it is having problems getting less than two unrecoverable errors per die, not a good sign.

Charlie Dhttp://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/http://www.semiaccurate.com/2009/12/21/nvidia-castrates-fermi-448sps/

This is a shame, a really big shame. :shadedshu

That article shows just how ignorant Demerjian really is. Seriously he might have some insider info at TSMC or whatever, but he just knows shit about GPUs...

So the reason the Tesla card is crippled is because the 512 SP chip can't be done and not because they have to play with limited power envelop (namely 225w), which is even more limited by the fact that it comes with 3 and 6GB of GDDR5 (aka 24 memory modules as shown in the chart). At 2.5-5w per memory module 12 extra modules suppose 30-60 watts more, depending if you are talking about average consumption or peak consumption. With 12 modules less the GeForce card will have those 30-60 watts to spend on the GPU. Not to mention they can exceed the 225w envelope if really required.

And regarding the power envelope and 8pin+6pin, and the fact that the card only needs two 6-pin or one 8-pin, remember that I said that long time ago...

Anyway, as always Charlie taking something that is true and making up the rest in order to create something half-believable that in reality only has 5% of truth. Now THAT is a shame...

 

[erocker]

That article shows just how ignorant Demerjian really is. Seriously he might have some insider info at TSMC or whatever, but he just knows shit about GPUs...

So the reason the Tesla card is crippled is because the 512 SP chip can't be done and not because they have to play with limited power envelop (namely 225w), which is even more limited by the fact that it comes with 3 and 6GB of GDDR5 (aka 24 memory modules as shown in the chart). At 2.5-5w per memory module 12 extra modules suppose 30-60 watts more, depending if you are talking about average consumption or peak consumption. With 12 modules less the GeForce card will have those 30-60 watts to spend on the GPU. Not to mention they can exceed the 225w envelope if really required.

And regarding the power envelope and 8pin+6pin, and the fact that the card only needs two 6-pin or one 8-pin, remember that I said that long time ago...

Anyway, as always Charlie taking something that is true and elaborating the rest in order to create something half-believable that in reality only has 5% of truth. Now THAT is a shame...

Not to mention the 448sp part could just be a different model. I do assume that Fermi will be more than one giant ball breaking card! It's a good thing Charlie didn't hear about any new low power mainstream card from Nvidia, he might think that's Fermi as well.

 

[Bo_Fox]

Hmm, that Nvidia PDF is for the Tesla. A 6GB version is still ridiculous.

It could be like a GTX-360, but for the Tesla, with 24 memory chips for either 3GB or 6GB. I'm surprised that there is now a 256MB GDDR5 chip available. Well, I'm still skeptical about the 6GB version, which is 4 times of what the Fermi is supposed to have (1.5GB).

Charlie just took the numbers for the Tesla and applied it to the GT300. Charlie is forgetting that the workstation graphics is usually a tad bit slower than the enthusiast graphics (8800 Ultra, GTX 285, etc..).


Ahh... here, 2Gb GDDR5 memory chips by Hynix: http://www.semiaccurate.com/2009/12/21/hynix-first-40nm-2gb-gddr5-memory/

 

[qubit]

That article shows just how ignorant Demerjian really is. Seriously he might have some insider info at TSMC or whatever, but he just knows shit about GPUs...

So the reason the Tesla card is crippled is because the 512 SP chip can't be done and not because they have to play with limited power envelop (namely 225w), which is even more limited by the fact that it comes with 3 and 6GB of GDDR5 (aka 24 memory modules as shown in the chart). At 2.5-5w per memory module 12 extra modules suppose 30-60 watts more, depending if you are talking about average consumption or peak consumption. With 12 modules less the GeForce card will have those 30-60 watts to spend on the GPU. Not to mention they can exceed the 225w envelope if really required.

And regarding the power envelope and 8pin+6pin, and the fact that the card only needs two 6-pin or one 8-pin, remember that I said that long time ago...

Anyway, as always Charlie taking something that is true and making up the rest in order to create something half-believable that in reality only has 5% of truth. Now THAT is a shame...

Charlie said it couldn't be done and not just power. He's pretty accurate most of the time, too. For some reason people equate his "angry man" style with lack of knowledge, which is just wrong.

 

[Benetanegia]

Hmm, that Nvidia PDF is for the Tesla. A 6GB version is still ridiculous.

It could be like a GTX-360, but for the Tesla, with 24 memory chips for either 3GB or 6GB. I'm surprised that there is now a 256MB GDDR5 chip available. Well, I'm still skeptical about the 6GB version, which is 4 times of what the Fermi is supposed to have (1.5GB).

Charlie just took the numbers for the Tesla and applied it to the GT300. Charlie is forgetting that the workstation graphics is usually a tad bit slower than the enthusiast graphics (8800 Ultra, GTX 285, etc..).

Yup, he has also overlooked the fact that the card is only 9.75 inches long (a bit more than HD5850, a lot less than HD5870) and that critically affects the cooling performance, especially for a suposedly big chip like Fermi. The fact that a 448 SP Fermi can be cooled down on a 9.75 inches long card with 24 memory chips in there already puts to shame all the crap he has put in that article...

Charlie said it couldn't be done and not just power. He's pretty accurate most of the time, too. For some reason people equate his "angry man" style with lack of knowledge, which is just wrong.

Face it, he is an ignorant and knows nothing, except what he is told by his sources (read him again and you will see how he always repeasts the same thing, with exact same words, because he really doesn't know what he is talking about, he repeats what he has been told like my 5 years old cousin does). You just don't say what he has said there, regarding the TDP, power envelop number of SPs and whatnot without taking into account the memory modules, the card length, where the card is going to be used...

If you are waiting to see the fallout of such architectural snafus, look no further. 448 SPs at 225W TDP,

His entire "article" is based on this conclusion, his conclusion, strongly based on those bold letters: "448SPs at 225W". The rest has been made up to reach that conclusion, as if the fact that the Tesla card being crippled to 448 SPs in order to fit the tight restrictions required for HPC mean that every Fermi card will hafve to be crippled. And he is absolutely wrong, because he has overlooked so many things explained above. Not to mention that the white paper doesn't mention 225w being the TDP and it just specifies it will be lower than that in any case...

 

[Goodman]

Sory! if that been post before i didn't want to go through all the 28 pages lol!

But so far nvidia new card is fake & i meen this ;
http://www.semiaccurate.com/2009/10/01/nvidia-fakes-fermi-boards-gtc/

Update 3: When asked initially, a simple response of, "Yes it was a mockup, Jen-Hsun used non-technical language to describe the part" would have made this a non-story. Instead, it is a great story about corporate lies and public deception.

Benchmarks are also fake ;
http://www.dvhardware.net/article39777.html

They don’t match our own information and after talking to NVIDIA we can only conclude that they are very much fake. Someone simply took one of NVIDIA’s older roadmaps and edited it to present what looks like the two cards GeForce GTX 380 and GTX 360, in competition with Radeon HD 5970 and HD 5870.

 

[KainXS]

Sory! if that been post before i didn't want to go through all the 28 pages lol!

But so far nvidia new card is fake & i meen this ;
http://www.semiaccurate.com/2009/10/01/nvidia-fakes-fermi-boards-gtc/



Benchmarks are also fake ;
http://www.dvhardware.net/article39777.html

yea, we know, we know, thats old news man

and Charlie Demerjian, . . . . . .

Charlie . . . . . . is the one of the biggest Nvidia haters ever, so i skip by all his news, its usually crap

even if the telsa is ony 448 sp I expect it to be a very fast card still so charlie can bash all he want lol

 

[cadaveca]

#1. Before we go further, let me say this:

The number of "tech journalists" that actually hold a degree even slightly related to journalism is far less than most think. To expect them to actually have any credibility as a journalist is foolish, at most, and mis-guided at least.

For Canadian ones:

http://www.caj.ca/index.html

American:

http://www.spj.org/ethicscode.asp

How many tech writers are members?



So before you even suggest that these guys have any ethics, check if they are listed members. If they are not, there is nothing there for them to base any ethics on, other than thier personal opinions.

Fact is, most aren't really "professionals" by any means.

#2:

Charlie's nV hate is an act. He is a writer, and the character he portrays is anti-nV(as Fudo was pro-ATI, at least at first), but his only true allegiance is to $$$...which comes by way of page hits. He MUST be a sensationalist to get those hits. his bias is faked, 100%.


#3: I've never hidden the fact I personally am pro-ATI, but, at the same time, as a consumer, I need nV to be a success. Decent pricing demands it. That said, I am confident that nV's next Geforce product will be a success, overpriced, and at the same time, will be in time for DX11 to really hit the marketplace...assuming DX11 is an API that nV will provide hardware for. 3 titles hardly create a demand.

 

[kid41212003]

Beside, who doesn't like Tesla tank?

 

[Bo_Fox]

Charlie said it couldn't be done and not just power. He's pretty accurate most of the time, too. For some reason people equate his "angry man" style with lack of knowledge, which is just wrong.

I'm gonna play the middle ground here and say that he's about 50-50 with being accurate or not! That's why his new site is called "semi-accurate"!

He was rather bold in pointing out the manufacturing defects of the 65nm NV chips (especially the 8600 cards). This was a serious issue upon which Nvidia got sued for. Nowadays, during the bad economy, computer hardware is not as reliable as it used to be. 3 years ago, pretty much every computer component was generally expected to be all-around reliable, for at least a number of years. Now, one would be lucky if every single part of the PC lasts for more than one year. First, it started with Crucial Ballistix, then Nvidia 65nm cards, then now it's the hard drives (especially Seagate), which is the worst of all.

Anyways, since Charlie did not say enough about the Tesla specs, on whether it's also applicable for the GT300 while backing it up with the credentials, I'll just take it with a grain of salt. It desperately needs a huge heap of salt and spices.

 

[qubit]

I'm gonna play the middle ground here and say that he's about 50-50 with being accurate or not! That's why his new site is called "semi-accurate"!

No, it's called SemiAccurate, because it's Charlie's sarcastic humour. And sure, he won't get everything right, but then who does?

He was rather bold in pointing out the manufacturing defects of the 65nm NV chips (especially the 8600 cards). This was a serious issue upon which Nvidia got sued for. Nowadays, during the bad economy, computer hardware is not as reliable as it used to be. 3 years ago, pretty much every computer component was generally expected to be all-around reliable, for at least a number of years. Now, one would be lucky if every single part of the PC lasts for more than one year. First, it started with Crucial Ballistix, then Nvidia 65nm cards, then now it's the hard drives (especially Seagate), which is the worst of all.

Anyways, since Charlie did not say enough about the Tesla specs, on whether it's also applicable for the GT300 while backing it up with the credentials, I'll just take it with a grain of salt. It desperately needs a huge heap of salt and spices.

Indeed, he was very bold. That was a lot of time and money that him and The Inquirer invested in exposing nvidia's lies on this one. This alone should earn him a lot of respect from his readers and silence his critics. He's pretty on the ball most of the time too, from what I can see. I've emailed him several times over the years and he's just the same in his replies, so his reporting style is not an act. He's simply got the balls to call out any company that tries to bullshit its customers and rub their nose in it.

If you can find two or three articles (on any site) where he was significantly off, then I'd be very interested to see them. And I'm not being sarcastic here; I really would like to see them. However, whenever I've challenged anyone to do this, the examples never come...

This challenge is open to anyone, not just Bo.

 

[Bo_Fox]

No, it's called SemiAccurate, because it's Charlie's sarcastic humour. And sure, he won't get everything right, but then who does?



Indeed, he was very bold. That was a lot of time and money that him and The Inquirer invested in exposing nvidia's lies on this one. This alone should earn him a lot of respect from his readers and silence his critics. He's pretty on the ball most of the time too, from what I can see. I've emailed him several times over the years and he's just the same in his replies, so his reporting style is not an act. He's simply got the balls to call out any company that tries to bullshit its customers and rub their nose in it.

If you can find two or three articles (on any site) where he was significantly off, then I'd be very interested to see them. And I'm not being sarcastic here; I really would like to see them. However, whenever I've challenged anyone to do this, the examples never come...

This challenge is open to anyone, not just Bo.

LOL, yes, it's called SemiAccurate!

I've seen like 6-7 examples.

One example that I can recall is when the 8800GTX was profoundly claimed to have 32 ROP's. Another example is when the HD2900XT was going to have 32 TMU's. Charlie sounded pretty sure in his own convictions back then.

There are millions of examples here in this link ( http://charliedemerjianisadouchebag.blogspot.com/ ) and other links.

I do like Charlie myself, and he does have a lot of my respect. He criticized Nvidia's "proprietary" Geforce 3DVision glasses that sold for $200, which made me happy, but he said a few things that showed some weaknesses (lacking in knowledgability about Stereo3D).

Once again, I'm playing the middle ground here, and I do like him myself, and do not think he's a douchebag at all!

 

[erocker]

http://www.siliconmadness.com/2009/12/nvidia-tesla-c2050-c2070-cut-down-to.html

 

[a_ump]

448 is still a good bit less than 512, still wonder how performance will be. I too agree with it having to be yield problems, what other logical reason could there be to lower the specifications.

 

[EarlZ]

Fermi is too advanced for the current 40nm manufacturing.... it should have been made with 32nm

 

[GSquadron]

Nvidia Fermi

I am afraid this time nvidia just wanted to launch its new fermi after hynix developed its new 2 GB with the most high density. Anyway i have predicted a lot of things in other threads, so i hope this time it will happen for sure again. Nvidia just new the plans of hynix and that is why it was not afraid of ati. Hope u reply guys, cuz this move of nvidia i think is not worth waiting so much. It would have better launched its fermis before this, let's say not the high end, but just other mid end gtx 3xx. So for now it is not a surprise that nvidia makes fermi the most powerful cards in the world. Ati must wake up and launch a new series immediately after fermi's launch. Otherwise it will be trouble for amd/ati.

 

[Abir053]

if nvidia wants to ship their top of the line gtx380 at a price tag like us$600 then...........a lot of people gonna back off from nvidia.

and i guess amd is already preparing 5890. so no worries for amd.

nvidia must launch Fermi as soon as possible with a proper price tag.