NVIDIA GeForce 4XX Series Discussion

[wahdangun]

http://img12.imageshack.us/img12/3892/34925239510d4107fc15.jpg

Source: XtremeSystemsForums

WOW, it's that really fermi it's look really wicked, i never seen prototype before but how they connect it to PCIe? i know now why nvdia use fake board

 

[qubit]

More GT300 news

All right, putting aside obviously fake prototypes, here's two more news items about GT300 from Fudzilla:

Nvidia has high supercomputer hopes for GT300, Fermi

GT300 Fermi supports GDDR3

 

[KainXS]

GDDR3 on Fermi is a big Big thing, GDDR3 on a 384bit bus = less rops than on the same bus and were talking half the rops now, 384bit GDDR3 should max at 24 rops.

Hence I think that news is fake.

 

[qubit]

GDDR3 on Fermi is a big Big thing, GDDR3 on a 384bit bus = less rops than on the same bus and were talking half the rops now, 384bit GDDR3 should max at 24 rops.

Hence I think that news is fake.

I don't think it would be fake: the performance would be fine for an 8400GS level card, wouldn't it? In fact, it would be rather better than that.

 

[largon]

Did you read the rest of the thread? Fermi's working form (durring the keynote) was a bundle of sockets, boards, and wires. Read between the lines.

I asked saaya (the original poster of the image) about the proverbial pic, just to amuse you, and as I expected, he said he pulled it from Google image search.
As The Comedian would say: "It's a joke. I's all a joke."

,|||. Thumb, index, middle, ring, pinkey ya hack.

No idea what that should mean.
But you misstyped "pinky".

 

[qubit]

I asked saaya (the original poster of the image) about the proverbial pic, just to amuse you, and as I expected, he said he pulled it from Google image search.
As The Comedian would say: "It's a joke. I's all a joke."

No idea what that should mean.
But you misstyped "pinky".

I can't believe anyone fell for it.

 

[Benetanegia]

GDDR3 on Fermi is a big Big thing, GDDR3 on a 384bit bus = less rops than on the same bus and were talking half the rops now, 384bit GDDR3 should max at 24 rops.

Hence I think that news is fake.

ROPs have nothing to do with the memory width and much less with memory type. Since G80 to GT200 Nvidia put 4 ROPs per memory channel to form each cluster, just as Ati did before RV870. But they could put 5, 6, 7, 8 whatever they want to. Now RV870 has 8 ROPs per channel and probably Nvidia is doing the same. RV870 suports GDDR3 too AFAIK. ROPs have only been linked to memory width because they reused the same design since G80.

384 bit + fast GDDR3 still has a lot of punch for a midrange card anyway, remember that HD4850 uses average GDDR3 too and does 64 GB/s. 384 bit/8 x 2200 mhz GDDR3 = 105 GB/s, 10 GB/s less than GTX260, more than enough to make something that is to the GTX2xx what the 8800 GT (57 GB/s) or 9800GTX+ (70GB/s) was to the 8800 GTX (86 GB/s). And this time we are talking about much more raster power in general, even though G92 struggled at 1920x1200 4xAA in comparison to G80, this won't happen now. A midrange card doing 1920x1200 4xAA is more than enough.

 

[KainXS]

I don't think it would be fake: the performance would be fine for an 8400GS level card, wouldn't it? In fact, it would be rather better than that.

after thinking about it I am thinking that since they say devices I think they are talking about integrating it into other things similar to multichip server clusters and whatnot, kinda like a evolution of telsa, or maybe products like tv's(which would be awesome) or next gen consoles maybe.

 

[Binge]

I asked saaya (the original poster of the image) about the proverbial pic, just to amuse you, and as I expected, he said he pulled it from Google image search.
As The Comedian would say: "It's a joke. I's all a joke."

No idea what that should mean.
But you misstyped "pinky".

I can't believe anyone fell for it.

Hold on guys I found the real image of Fermi!

Now I can't believe you two, especially qubit (he's the smarter chuckle-head), can't read between the lines. Nobody has a picture of a working Fermi, and everyone knows it. Some people can't grasp that a GPU could be an unfinished rough product, and so I cited that image as an example. There are words in my original post that show my example, the picture of "Fermi" by saaya, as a simile to Fermi.

-Blinge

 

[Binge]

GDDR3 on Fermi is a big Big thing, GDDR3 on a 384bit bus = less rops than on the same bus and were talking half the rops now, 384bit GDDR3 should max at 24 rops.

Hence I think that news is fake.

The memory controller is 384-bit wide in Fermi “GTX” single chip 512 shader version, but in some less expensive chips they will use less memory bandwidth and even GDDR3.


READ!!! I can't believe this. First two chuckle heads actually think I'm a dolt, and now you can't read that some less expensive chips will use DDR3. NOT the Fermi GTX, but a less expensive chip.

 

[erocker]

Love is blind. Especially when you love "X" brand computer product.

 

[KainXS]

The memory controller is 384-bit wide in Fermi “GTX” single chip 512 shader version, but in some less expensive chips they will use less memory bandwidth and even GDDR3.


READ!!! I can't believe this. First two chuckle heads actually think I'm a dolt, and now you can't read that some less expensive chips will use DDR3. NOT the Fermi GTX, but a less expensive chip.

with all the talk of fermi and how scalable it is I gots confused

still, I guess this means that the entire GT300 series could be based off the fermi which is an awesome thing, I am looking forward to it now, this will be the first time since the 8 series came out that nvidia did that.'

now what I wonder is is the GTX380 going to launch alone or will it launch with a whole array of cards, from low to high end.

 

[qubit]

Hold on guys I found the real image of Fermi!

http://exophase.com/wp-content/uploads/2009/09/fermi.jpg

Now I can't believe you two, especially qubit (he's the smarter chuckle-head), can't read between the lines. Nobody has a picture of a working Fermi, and everyone knows it. Some people can't grasp that a GPU could be an unfinished rough product, and so I cited that image as an example. There are words in my original post that show my example, the picture of "Fermi" by saaya, as a simile to Fermi.

-Blinge

I don't see what I got wrong here? I saw straight through that wire contraption masquerading as a prototype and I posted Charlie's rant about that "prototype" card that nvidia held up was a fake too.

And finally, I'm now officially a chuckle-head!

 

[largon]

READ!!! I can't believe this. First two chuckle heads actually think I'm a dolt, and now you can't read that some less expensive chips will use DDR3. NOT the Fermi GTX, but a less expensive chip.

Post reported and welcome to my ignore list. I have no interest in reading insults after correcting people's mistakes.

PS. "Fermi" is not a specific GPU or a card. Fermi is the architecture.

 

[Binge]

I don't see what I got wrong here? I saw straight through that wire contraption masquerading as a prototype and I posted Charlie's rant about that "prototype" card that nvidia held up was a fake too.

And finally, I'm now officially a chuckle-head!

None of us are wrong. Chuckling is good for people. I love playing the pissed off kid who needs to be right more than the last, but my shot at humor ended up in a lame argument over nothing with you and some guy who's ignoring me. I won't name names

To be honest I'm supremely frustrated with both nV and the most posts regarding ATI/nVidia discussion. It's one thing to crack jokes, and it's another to get an idea of what's really going on. For pages and pages I've seen people get the wrong idea and end up left in the dark. The ignorance turns into hate and confusion. There are a few facts that just get pushed to the side because of the media, and the point I was trying to make all along is that the working model at the Keynote has no photograph leaked on the web. We don't even know it existed at all. The presentations could have been pre-rendered, but there were segments of the Keynote which claimed the card was rendering a scene or working on a physics application. Think about it. If it's all false claims then nV will be in serious sh*t because their white paper makes a LOT of claims which are pretty impressive if they can back it up with proof at some point. To me that means more than the wait.

 

[qubit]

None of us are wrong. Chuckling is good for people. I love playing the pissed off kid who needs to be right more than the last, but my shot at humor ended up in a lame argument over nothing with you and some guy who's ignoring me. I won't name names

To be honest I'm supremely frustrated with both nV and the most posts regarding ATI/nVidia discussion. It's one thing to crack jokes, and it's another to get an idea of what's really going on. For pages and pages I've seen people get the wrong idea and end up left in the dark. The ignorance turns into hate and confusion. There are a few facts that just get pushed to the side because of the media, and the point I was trying to make all along is that the working model at the Keynote has no photograph leaked on the web. We don't even know it existed at all. The presentations could have been pre-rendered, but there were segments of the Keynote which claimed the card was rendering a scene or working on a physics application. Think about it. If it's all false claims then nV will be in serious sh*t because their white paper makes a LOT of claims which are pretty impressive if they can back it up with proof at some point. To me that means more than the wait.

Firstly, did you see that I thanked you for your post? And I did a little toastie, too. I gots da humour.

I still have no idea why you thought I got anything wrong, but hey, it's cool.

I agree with you about the keynote being iffy, with evidence of working parts rather elusive. It's the sort of thing that Charlie picks up on all the time and seems to cause flames, for some reason. From what I can tell, I'm the only one that thinks that what Charlie says has any merit. Dunno, maybe I'm missing something.

 

[Animalpak]

Guys do you think that Nvidia is planning to introduce a new Technology like ATI eyefinity for the next GT 300 cards ?

Because i saw some videos on youtube and is frkkin awesome to play games with 3 monitors.

 

[phanbuey]

Guys do you think that Nvidia is planning to introduce a new Technology like ATI eyefinity for the next GT 300 cards ?

Because i saw some videos on youtube and is frkkin awesome to play games with 3 monitors.

**SPECULATION**
I think nvidia's new tech will be less in the X montiors department, and more in the "your entire computer will run faster bc of CUDA department."

kind of like physx accelaration, but something that will kick in automatically for many-core threaded applications.

For instance - you run a benchmark like wPrime... it can use as many threads as you let it (i think)... the driver might then detect that you ran an application that can use more than X threads and the GFX card will kick in and do the crunching.

so instead of using a 50-100Gflops of your processor... you will now use 1-2Tflops. so that 32M calculation will be done in just a few seconds. At least that is the impression I am getting from where nvidia is trying to go.

***Again, SPECULATION***

 

[Benetanegia]

**SPECULATION**
I think nvidia's new tech will be less in the X montiors department, and more in the "your entire computer will run faster bc of CUDA department."

kind of like physx accelaration, but something that will kick in automatically for many-core threaded applications.

For instance - you run a benchmark like wPrime... it can use as many threads as you let it (i think)... the driver might then detect that you ran an application that can use more than X threads and the GFX card will kick in and do the crunching.

so instead of using a 50-100Gflops of your processor... you will now use 1-2Tflops. so that 32M calculation will be done in just a few seconds. At least that is the impression I am getting from where nvidia is trying to go.

***Again, SPECULATION***

I agree. They are working on speeding up virus scans too, and that is something that I would definately like to have.

http://www.fudzilla.com/content/view/15832/1/

Actually they are working on speeding up almost everything that can work in parallel, which IMO is amazing.

For gamers they still have Stereo3D and PhysX and both need a lot of work in order to be properly implemented, so I don't think they will add anything else to their list, although better multi-monitor support souldn't be very difficult to implement.

 

[qubit]

GT300 better come soon - GTX 295 price down

Overclockers has them at £289.98 inc VAT. Looks like Charlie's article Nvidia will crater GTX260 and GTX275 prices soon was right (he includes the 295 in the price crash in the article). Frankly, it didn't take much to see that this had to happen, given the performance of the 5850 & 5870.

And please, no flames!

 

[Binge]

I've linked a CNet article below about Fermi's Architecture and a hint that nVidia will miss the holiday season. This is not great news to folks who can't wait, but it is not so bad once you take a look at this article written by Peter Glaskowsky. He is one of the folks who wrote a white paper outlining the Fermi Architecture.

http://news.cnet.com/8301-13512_3-10369441-23.html

Nvidia hasn't released any numbers yet, but Brookwood estimates the GF100 cores will run at about 1.5 GHz; ATI says the RV870 will be clocked at 850 MHz.

If Brookwood is right, the theoretical maximum throughput of the GF100 on single-precision floating-point operations works out to 1.536 GFLOPS, whereas ATI announced the figure of 2.72 GFLOPS for the RV870.

On double-precision data, the ranking is reversed. Nvidia would have 768 GFLOPS, and ATI, 544 GFLOPS. Nvidia may also have advantages in memory bandwidth, since its design has more memory close to the cores and six channels of GDDR5 memory rather than ATI's four, and ATI's programming model may lead to slightly lower overall efficiency.

 

[Animalpak]

Well sounds good.

I'm sure that whoever comes to see this thread is a loyal nvidia GPU buyer. Me too

I want to know making you a question :

What are the main reasons for which we select nvidia instead of ATI ?

( is not a fanboy question but for our personal information )

Me first.

-Maximum performance in all games
-Stable performance in games
-Drivers reliable and frequent updates with continuous improvements in games
-Many options and settings ( nvidia control panel, EVGA precision )
-graphics cards aesthetically beautiful ( i love the green and the black )


I always chose nvidia since my first GeForce 7800 GS AGP8X

 

[qubit]

Well sounds good.

I'm sure that whoever comes to see this thread is a loyal nvidia GPU buyer. Me too

I want to know making you a question :

What are the main reasons for which we select nvidia instead of ATI ?

( is not a fanboy question but for our personal information )

Me first.

-Maximum performance in all games
-Stable performance in games
-Drivers reliable and frequent updates with continuous improvements in games
-Many options and settings ( nvidia control panel, EVGA precision )
-graphics cards aesthetically beautiful ( i love the green and the black )


I always chose nvidia since my first GeForce 7800 GS AGP8X

Your question is a good conversation starter, but is off topic for this thread. Why not start your own thread with it?

 

[kid41212003]

The GF100 won't be out in time for the holiday buying season, but ATI's already shipping the RV870. As long as ATI can deliver its new chip in volume, it should do very well over the next few months. Nvidia doesn't seem concerned, pointing to its recent history of market-share leadership, but gamers can be fickle. They tend to buy what's best when they're ready to buy, rather than waiting for something potentially better to come along.

This is going to be a really long wait, AMD have the whole Q4 to sell their cards. I'm getting impatience .

 

[GC_PaNzerFIN]

I now have experience with HD 5870s, and I can't wait to see what NVIDIA has to offer! Maybe I am crazy... or just addicted to gfx cards.

Anonymous graphics card enthusiast.... is there such?