The Official RV630/RV610 Rumours & Speculation Thread

[gunblade]

FUD or not, I think it provides a glimpse into the graphic card cost breakdown.

From here, it is very obvious why mid-low end chip still remains 128bit Memory IF.
The ASP of the mid range chips are average 32~38 and the double of memory cost will definitely be very significant. Though, PCB cost will also be up to about 3 ~5 dollar more but that is defnitely more manageable than doubling the memory chips.

 

[Rufus]

Worst FUD (weak) campaign I've ever seen from Nvidia so far.

Meh, it's not particularly new or unique. ATI and Nvidia have both been putting out presentations like this for customers / reviewers / channel / marketing / watever, and they've often been much more biased. The only difference is they're starting to leak online more.

 

[BrynS]

Yeah, you know you're on to a winner when the competition cranks out a PowerFUD. As Rufus mentioned, it is noticeable that these have been increasingly more prevalently leaked over the past few years -- remember the fuss over Nvidia's Kyro 2 'concerns', which seemed fairly unprecedented (i.e. both the leak into the public domain and the nature of the document) at the time.

Perhaps UVD wasn't enabled in the drivers prior to these slides being compiled?

 

[Anarchist4000]

It's not that UVD wasn't necessarily enabled through the drivers. They didn't even test UVD on a card that actually had it. For the video acceleration benchmarks they tested against a 1950Pro. They aren't even showing UVD performance at all.

 

[BrynS]

Well, that's was I was suggesting: either UVD wasn't enabled in the drivers or it produces better results than VP2, so NV used a X1950 Pro for comparison instead.

If RV610 produces similar CPU usage when decoding (with UVD not yet enabled in the drivers) as the R5-series, then presenting that as representative of UVD performance would be a bit too disingenuous even for one of these documents, especially when the eventual reviews will presumably show drastic reductions if UVD performs well.

 

[Anarchist4000]

UVD should be showing almost no CPU utilization so more than likely that's the reason for the omission. That along with HDMI should be the major selling point of the low/mid range cards.

 

[neliz]

But ATi isn't better:

hey.. but that stuff is actually true!

 

[Kaotik]

Well, that's was I was suggesting: either UVD wasn't enabled in the drivers or it produces better results than VP2, so NV used a X1950 Pro for comparison instead.

If RV610 produces similar CPU usage when decoding (with UVD not yet enabled in the drivers) as the R5-series, then presenting that as representative of UVD performance would be a bit too disingenuous even for one of these documents, especially when the eventual reviews will presumably show drastic reductions if UVD performs well.

It does work, and wasn't shown due the fact it works (based on preliminary benches) better than nVidias solution (not to mention nV does only 3 of 4 VC-1 decoding phases on hardware)
In the benchmarks somewhere while back, the performance was.. hum.. IIRC 10-15% CPU utilization depending on movie & scene on 2800+ Sempron.

 

[neliz]

IIRC 10-15% CPU utilization depending on movie & scene on 2800+ Sempron.

yeah.. and close to 5% on anything over a 4000+

 

[AnarchX]

First shot of the GeCUBE twin GPU HD2600
http://www.hexus.net/content/item.php?item=8895

Nice card, R600-killer?

 

[WaltC]

http://www.forum-3dcenter.org/vbulletin/showthread.php?p=1845413#post1845413


But ATi isn't better:


http://images.thgweb.de/2003/07/02/...mit_hase_und_igel/press_tour_presentation.zip

As I recall, though, the 5200 was pretty nearly completely unusable as a "DX9 accelerator" for DX9 games, even though nVidia advertised and sold it as such. That actuality would remove those ATi comments from the realm of FUD, wouldn't it?

I don't see anything objectionable about a company telling the truth about a competitor's products--the FUD part is what I find objectionable. FUD is pretty easy to spot as it always involves either skewed testing or skewed product descriptions, or both. In this case it is rather obvious as the skewed commentary revolves around skewed testing in which the product being criticized was not even the product tested.

 

[INKster]

As I recall, though, the 5200 was pretty nearly completely unusable as a "DX9 accelerator" for DX9 games, even though nVidia advertised and sold it as such. That actuality would remove those ATi comments from the realm of FUD, wouldn't it?

I don't see anything objectionable about a company telling the truth about a competitor's products--the FUD part is what I find objectionable. FUD is pretty easy to spot as it always involves either skewed testing or skewed product descriptions, or both. In this case it is rather obvious as the skewed commentary revolves around skewed testing in which the product being criticized was not even the product tested.

Do you find the following as telling the truth ?

Radeon 9200 will play DX9 class games faster and better through DX8.1 code path

This is pure nonsense. Everyone knew -even back then- that the 92xx family was a 8500 low-end derivative, so why advertise DX9 class games compatibility when they couldn't even run them that way ? What is "DX9 class" doing in that sentence anyway ?

The FX 5200 might have been a truly horrible game card, but everyone knew that it wasn't meant to be a high-end card in the first place. I know of no one that actually thinks that a current Geforce 7300 or Radeon X1300 can play contemporary games at reasonable settings and speed...
Can the Radeon 9200 claim to have Vista AERO Glass desktop compatibility (DX9 hardware support required), as the FX 5200 does now ?

 

[Zaphod]

This is pure nonsense.

IMO, only slightly nonsensical. What they're obviously saying is that DX9 in the 5200 was purely a check box feature and that the 9200 would play DX9 games - say HL2 - better at the settings where both cards could reasonably run said games. This is true.

 

[Davros]

what i want to know is why are ati giving nvidia their rv610's to test they havnt been released yet so nvidia must have sample boards ?

 

[Zaphod]

nvidia must have sample boards ?

How do you gather that? Seems to me that they're citing 'Pcinlife' as the source. It's not like it would be the first time that numbers grabbed from the net somewhere and taken out of context have been used in PR- or sales presentations.

Although, I wouldn't be surprised at all to learn that nVidia had gotten their hands on boards, I'd be very surprised if they were used in a manner such as to get numbers for the PR department...

 

[nutball]

How do you gather that? Seems to me that they're citing 'Pcinlife' as the source.

How do you gather that? Seems to me that Pcinlife have got hold of some NVIDIA slides and stamped their logo all over it before releasing it onto the net.

 

[Geeforcer]

Marketing slides, heh.. Quick, how many flops/s does R580 put out? Take your time...

Back on topic: 96/24 approach (R600 * 1.5) would also make sense. I don't know if 4:1 ratio is the architectural keystone or more of a function of transistor budget limitations. After all, RV630 has the 3:1 ratio, and RV610 2:1 ratio

 

[Zaphod]

How do you gather that? Seems to me that Pcinlife have got hold of some NVIDIA slides and stamped their logo all over it before releasing it onto the net.

Probably, yeah. Didn't look to closely at those slides. I just thought it so unlikely that any decent marketeer would use such direct comparisons without without citing sources or obscuring the numbers. It's just stupid. But then again, it's often policy to discard policy lower in the ranks, and most PR-leaks seems to stem from internal/unofficial material cooked together for sales teams or others with their eyes more focused on the next bonus check than the big picture.

 

[AnarchX]

GeCube 2600Pro listed in Europe:

http://geizhals.at/eu/a260174.html
http://geizhals.at/eu/a260173.html

...one shop says they will be deliverable on 22.06.2007


But they are crippled by DDR2 memory.

edit:

+
GeCube Radeon HD 2400 Pro
http://geizhals.at/eu/a260175.html

 

[Julidz]

do u know why isn't the 512bits from HD2900XT making the difference against 320bits from 8800GTS ?