nVidia GT200 Series Thread

[Brunnis]

Originally posted by: Kuzi
Isn't the data being "duplicated" in both 512MB memory pools? If that is the case, then we have less memory, 512MB (instead of 1024MB) and "less" bandwidth. Since data is being duplicated, then we are wasting bandwidth.

No, that only applies if the same data is being read in the same clock cycle by both chips. If they read different data, the effective bandwidth is two times that of a single 256-bit bus.

The important thing to realise is that while the effective bandwidth is two times that of a single 256-bit bus, each chip only has a 256-bit bus and each chip works on different frames. That last part is important. An 8800GTX has a 384-bit bus and can use that bandwidth to pull data for single frame, while a 9800GX2 can not.

So, yes, the total amount of data that can be moved is two times that of a single 256-bit bus, but the effect isn't the same as if you'd have the same amount of bandwidth to single chip.

 

[Keysplayr]

Originally posted by: Kuzi
Isn't the data being "duplicated" in both 512MB memory pools? If that is the case, then we have less memory, 512MB (instead of 1024MB) and "less" bandwidth. Since data is being duplicated, then we are wasting bandwidth.

So the GX2 is only 256bits or a little more, but not 512bits. That's how I understand it, I could be wrong.

Each card has 256-bit memory access, but because they do not actually display a single frame at the same time, I wouldn't consider it a 512 bit solution, but a 2x256 solution. However, we will get more fps because we have two cores churning out frames alternately. I guess it all depends on how this is looked at. In the end, performance is what we are looking at, and the GX2 is delivering it pretty nicely.

Keys

 

[bryanW1995]

so is the nda going to be lifted on launch day, or do we get a few days to savor the real benchies first?

 

[nRollo]

Originally posted by: bryanW1995
so is the nda going to be lifted on launch day, or do we get a few days to savor the real benchies first?

NDA is always lifted launch day for NVIDIA.

 

[Kuzi]

GTX280 core speed 602MHz
GTX260 core speed 576Mhz

Kinda low speeds. Check at INQ here

 

[HurleyBird]

The 50% "more powerful shaders" makes perfect sense given this new information.

Current gen: ~1600MHz * 128SPs = 204800
Next gen: 1296MHz * 240SPs = 311040

311040 / 204800 = 1.51875

Or around 50% more powerful in total.

If I'm right, NVIDIA is going to get pwned by R700.

 

[imported_Shaq]

Originally posted by: HurleyBird
The 50% "more powerful shaders" makes perfect sense given this new information.

Current gen: ~1600MHz * 128SPs = 204800
Next gen: 1296MHz * 240SPs = 311040

311040 / 204800 = 1.51875

Or around 50% more powerful in total.

If I'm right, NVIDIA is going to get pwned by R700.

Extremely unlikely the R700 wil pwn the GTX 280. The RV770XT is supposed to be priced at $349 with 1GB of DDR5 memory. If the GTX 280 is $600 it suggests the performance is way higher than the RV770XT. Even adding another GPU to it likely won't be enough to equal it much less pwn it. The 3870x2 is already close to 50% slower than the GX2. The GTX 280 is rumored to be 50% faster than the GX2. If ATI can double the performance with basically the same architecture than more power to them. And that is just to match the GTX 280. pwnage means they need to be 250% faster than the 3870x2. Good Luck!

 

[Tempered81]

the hd 4870 will lose to the gtx280 imo.

a 4870x2 will lose to a gtx280 or be ~ = in many occasions as well. gtx 280 will debut at 600-700 bucks though. and a 4870x2 should less than 500

 

[shabby]

Originally posted by: HurleyBird
The 50% "more powerful shaders" makes perfect sense given this new information.

Current gen: ~1600MHz * 128SPs = 204800
Next gen: 1296MHz * 240SPs = 311040

311040 / 204800 = 1.51875

Or around 50% more powerful in total.

If I'm right, NVIDIA is going to get pwned by R700.

Using your math a pair of 8800gt cards would be faster than the gtx280. Do you actually think nvidia's new high end card is going to be slower than a pair of 8800gt cards that cost less than $300?

 

[nRollo]

Originally posted by: Kuzi
GTX280 core speed 602MHz
GTX260 core speed 576Mhz

Kinda low speeds. Check at INQ here

LOL-
Is it just me, or does that article read like "We're not invited to press day, so we're going to rag on NVIDIA"?

 

[nRollo]

Originally posted by: HurleyBird
The 50% "more powerful shaders" makes perfect sense given this new information.

Current gen: ~1600MHz * 128SPs = 204800
Next gen: 1296MHz * 240SPs = 311040

311040 / 204800 = 1.51875

Or around 50% more powerful in total.

If I'm right, NVIDIA is going to get pwned by R700.

Bet a liter of Jack Daniels on that?

 

[OCGuy]

Originally posted by: nRollo

Originally posted by: Kuzi
GTX280 core speed 602MHz
GTX260 core speed 576Mhz

Kinda low speeds. Check at INQ here

LOL-
Is it just me, or does that article read like "We're not invited to press day, so we're going to rag on NVIDIA"?

That author has an ax to grind..anyone who takes that seriously is a joke.

 

[HOOfan 1]

Originally posted by: Ocguy31

Originally posted by: nRollo

Originally posted by: Kuzi
GTX280 core speed 602MHz
GTX260 core speed 576Mhz

Kinda low speeds. Check at INQ here

LOL-
Is it just me, or does that article read like "We're not invited to press day, so we're going to rag on NVIDIA"?

That author has an ax to grind..anyone who takes that seriously is a joke.

Not that theInquirer has ever been known for its journalistic integrity, but that was about the most unprofessional and blatantly hostile article I have ever read about a Tech part.

"we were not at the press event, which is why we can tell you this stuff early"

ok, if you were not at the press event, then how would you know this stuff? obviously through speculation and hearsay.....what a hack.

 

[Cookie Monster]

With the missing MUL utilized in the GT200, we are going to see this beast produce a theoretical max output of ~930GFLOPs. (1300MHz x 240ALUs x (2+1)FLOPs). I also dont find it surprising that the initial GT200s will be clocked low due to heat/power and yield issues. It was also the problems with G80s. People forget that the 8800GTS640MB had only a shader clock of 1200MHz.

Having its memory interface, ROPs, and framebuffer literally all doubling im going to bet that this thing will scream at 19x12 and above res with AA/AF. (Whereas most G92 based cards take a nosedive because all of this points to the fact that the G92 based cards were limited in bandwidth alot, and maybe even framebuffer for that matter in certain titles and res)

Im betting 3x8800GTX/9800GTX performance at high res AA/AF. SLi setups will not compare since there are far too many drawbacks to consider a dual setup to the GT200. However people who bought the 9800GX2s are left with the shortend of the stick.

 

[HurleyBird]

Originally posted by: Shaq
Extremely unlikely the R700 wil pwn the GTX 280. The RV770XT is supposed to be priced at $349 with 1GB of DDR5 memory. If the GTX 280 is $600 it suggests the performance is way higher than the RV770XT.

By R700 I mean 4870X2.

Originally posted by: ShaqEven adding another GPU to it likely won't be enough to equal it much less pwn it. The 3870x2 is already close to 50% slower than the GX2. The GTX 280 is rumored to be 50% faster than the GX2.

The key term here is "rumored to be". What I'm pointing out is that the rumors could be wrong given this latest info. What's more probable: that "50% faster shaders than current generation" means 50% faster on a per shader basis, or that GTX 280 has 50% more shading power in total given that 240 G80/G92 type SPs @ 1296MHz give pretty much exactly 50% more shader power than the current generation? My common sense (not my fanboy sense -- my last four GPUs have been NVIDIA after all) tells me the latter is more likely and the 280 is absolutely not going to be 50% faster than the GX2. It all comes down to whether the hard numbers are just coincidence or not. Personally, I think that if this is all just a coincidence, it's a pretty big one.

Originally posted by: shabbyUsing your math a pair of 8800gt cards would be faster than the gtx280. Do you actually think nvidia's new high end card is going to be slower than a pair of 8800gt cards that cost less than $300?

What are you smoking and can I have some?

 

[ghost recon88]

From that Insider article

The 260 will be priced at $449 and go up against the ATI 770/4870 costing MUCH less. The 280 will be about 25 per cent faster and quite likely lose badly to the R700, very badly, but cost more, $600+.

:Q

 

[AmberClad]

Originally posted by: ghost recon88
From that Insider article

The 260 will be priced at $449 and go up against the ATI 770/4870 costing MUCH less. The 280 will be about 25 per cent faster and quite likely lose badly to the R700, very badly, but cost more, $600+.

:Q

That, plus the 236W and 182W for the 280 and 260 has me doing a facepalm :roll:. I'm not sure what to say...I was kind of hoping of upgrading this cycle...and I'm kind of rethinking that at the moment. Why exactly is it that we have CPUs coming out that are consistently faster, more power-efficient, and cheaper...and we have this in the GPU front.

 

[HOOfan 1]

why would you even listen to an article written like that...

will most likely lose badly to the R700, very badly

what is that even based on? Trying to declare a winner amongst two chips that haven't even come out yet, is beyond idiotic....and again this is theInquirer.

No wonder that guy wasn't invited to nVidia's press event, if he has any journalistic credentials, they should be yanked.

That article was written more badly than a rabid, delusional fanboy post on a message board.

 

[Rusin]

nothing

 

[bryanW1995]

Originally posted by: Ocguy31

Originally posted by: nRollo

Originally posted by: Kuzi
GTX280 core speed 602MHz
GTX260 core speed 576Mhz

Kinda low speeds. Check at INQ here

LOL-
Is it just me, or does that article read like "We're not invited to press day, so we're going to rag on NVIDIA"?

That author has an ax to grind..anyone who takes that seriously is a joke.

how DARE you impugn charlie's journalistic integrity!!

 

[Rusin]

nothing

 

[nitromullet]

Originally posted by: Rusin
Yeah.. only thing that tops Charlie's journalistic integrity is Fuad.

I have a request for you as the OP of this thread... Since this is the only thread on GT200, the first post should have all the links to the spy pics and info contained in within this thread... I realized this when I tried to pop into the 4870 thread, and was like "forget it"... It's like 30 pages by now. No way I want to wade through that just to find the actual info.

Here's a list of all the pics posted so far and the links to the most relevant info (IMO) if you wanted to drop them into the first post.

http://www.vr-zone.com/article...sign_Exposed/5763.html
http://diy.pconline.com.cn/gra...news/0805/1294786.html
http://img175.imageshack.us/im...805201808234174cz4.png
http://en.expreview.com/img/2008/05/21/GT200-core_1.jpg
http://www.pczilla.net/en/post/17.html
http://www.fudzilla.com/index....view&id=7473&Itemid=65 (first full card PCB pics)
http://www.pcper.com/images/news/folding4.jpg

Info links:
http://www.dailytech.com/article.aspx?newsid=11842
http://www.tgdaily.com/content/view/37554/135/
http://www.pcper.com/comments.php?nid=5679

...and some INQ "data":
http://www.theinquirer.net/gb/...24/gtx260-280-revealed

I haven't been keeping up with the 4870 thread as much, so I have no idea what actual info and pics have been posted, but I'd really appreciate it of someone got the OP of that thread to update the first post with the 'meat' of that thread as well. It's a good idea to lock competing threads, but if we want these to be worthwhile the OP and/or mods need to update the first post with new info and pics.

 

[Acanthus]

Originally posted by: Denithor

Originally posted by: nRollo
That is not the GT200, this is the GT200

Oho, nicely done. Love the nail holding the IDE cable onto the board.

Truly classic!

Yeah the nail is what got me :laugh:

 

[Rusin]

nothing

 

[nRollo]

Originally posted by: Rusin
Funny that everyone who were invited said that it was true terror on performance and will make GF7 -> GF8 jump look smaller and Inquirer which weren't invited said that kind of negative thing
----

Adobe's Photoshop will have GPU and physics support!
http://www.tgdaily.com/content/view/37611/140/

Nvidia did demo CS4 which add's GPU and physics support for their existing multicore support and results were promising: They played with 2GB, 448 megapixel picture like it were 5 megapixel picture with 8-core Skulltrail-system!

Another great CUDA feature . Perhaps Nvidia weren't jokin one year ago when they said that GPGPU will be important part of next gen graphics cards. First PhysX-support and now Photoshop..

I was invited, and planning on attending, but I couldn't make it due to real work conflicts.