RV560/570 Gemini roadmap

It also shows S: Now on R480 - sampling? Why sampling? Dodgy graphic. Like the lack of HDMI on the 2006 parts.

Of course the graphic could be a re-labelled ATI graphic, and 80nm refresh was indeed set to be a Q2 release until the wheels fell off. Which, in a nasty kind of way, sorta fits with how things seem to have gone.

As I said the other day, I wonder if anyone at the finance chat will ask ATI why 80nm is taking so long.

Jawed
 
geo said:
Shows P: Now on all those 80nm parts tho. So that doesn't fit.
Click to expand...
Don´t confuse the original dates these roadmaps were done. P: is shown on the gecube roadmap which looks like it is fairly new. HKEPC just provided a roadmap from the beginning of this year (fairly old) to underline that there still are some different estimates floating around, which is nothing surprising, considering those estimates on ATI´s roadmap are extrapolated results (done for the AIBs) rather than taken from actual, working HW.

geo said:
I would guess price points have more to do with what they picked to show against (and maybe some cherry picking, you can't rule that out in PR docs).
Click to expand...
Since GeCube is an AIB they show their portfolio of SKUs, so it makes sense the way they did it. Performance is important, but it´s a lot more important to show at what price points you have that performance available. So this should give us a pretty good idea where they are gonna place those parts price wise.

Now we also know where INQ got that idea that RV560/RV570 may be introduced earlier than their original estimated P: date. It obviously was taken from the GeCube roadmap.
 
Last edited by a moderator:
L'Inq with some "new" Gemini info. Two new codenames have appeared it seems, Hellfire 1 and Hellfire 2 -- "...These are like Gemini, but with a single chip and a different memory speed," according to Fudo, although that doesn't jive with Dave's previous clarification that Gemini=twins, i.e. dual-chip reference board.
 
Umm. "Supports both ATX and BTX formats". Err, what? There are pcie compatible viddy cards that don't?
 
Yeah, I am glad ATI is replacing Radeon X1800 XT and Radeon X1900 GT both of those are too expensive to make in comparison to Nvidia's offering.

Though the RV560 and RV570XL are interesting.

Since we saw that the 7600 G73 core has 16(12)x1/5 with a quad disabled Nvidia can possibly release a 7600 GTX to compete with the RV570 XL.
 
coldpower27 said:
Since we saw that the 7600 G73 core has 16(12)x1/5 with a quad disabled Nvidia can possibly release a 7600 GTX to compete with the RV570 XL.
Click to expand...

Hmm? You're talking about RivaTuner's turned-off quad-bit? That isn't generally accepted yet --I believe NV has said it isn't true. I'm not taking that as gospel either, btw, but it is enough for me to put it in the unproven catagory --unless you can point at someone who has managed to unlock one?
 
geo said:
Hmm? You're talking about RivaTuner's turned-off quad-bit? That isn't generally accepted yet --I believe NV has said it isn't true. I'm not taking that as gospel either, btw, but it is enough for me to put it in the unproven catagory --unless you can point at someone who has managed to unlock one?
Click to expand...

It just something I wanted to insert it there it's may not be provable, since Nvidia utilized means now that prevent any unlocking on G7x GPU's.

Though we will have to see, if RV570 XL is indeed a 256Bit Interface part, then a disabled G71 GPU with 256Bit Interface may be more ideal for that category as an 7800 GT replacement probably just release the OEM 7900 GS to retail and all is well.
 
Hkepc again with a naked gemini board:

15pihsk.jpg


Some bumblings about physics on the card, Dailytech has the english version of the same article.
 
Last edited by a moderator:
The PEX 8532 device offers 32 PCI Express lanes, capable of configuring to 8 flexible ports. The switch conforms to the PCI Express Base Specification, rev 1.0a. The 32-lane switch enables users to add scalable, high bandwidth I/Os to a wide variety of applications including servers, communications, storage, blade servers, and embedded systems. The device is hardware configurable and software programmable, allowing users to tailor their port configurations and QoS operating characteristics to suit their application requirements.
Click to expand...
http://www.plxtech.com/products/pci_express/PEX8532/default.asp

Since the bridge chip would appear to be boasting software programmability that may be where the physics bit is from. Allowing ATi to essentially do what ever they want with one, or of course both, of the GPUs via drivers i would assume.

I also noted there doesnt appear to be the extra pins on the DVI for a CrossFire connection nor the tabs for the new bridges we saw in the pictures now removed. Still depending on the master card problably (they wouldnt do 4GPU over the PCIE Bus like X1300/X1600 would they?) if they allow CrossFire with these cards at all.
 
Last edited by a moderator:
SugarCoat said:
I also noted there doesnt appear to be the extra pins on the DVI for a CrossFire connection nor the tabs for the new bridges we saw in the pictures now removed. Still depending on the master card problably (they wouldnt do 4GPU over the PCIE Bus like X1300/X1600 would they?) if they allow CrossFire with these cards at all.
Click to expand...
Well, I doubt there would be a whole lot of point to that. It becomes harder and harder to manage load balancing the more GPU's you add. With lower-end GPU's like this, I don't see any reason whatsoever. Without even looking to the nVidia camp, one of these puppies might well be outperformed by a Radeon X1900 XTX, and would easily fall far short of two X1900's in Crossfire mode. Crossfire on this dual-GPU board wouldn't improve things much.
 
Chalnoth said:
Well, I doubt there would be a whole lot of point to that. It becomes harder and harder to manage load balancing the more GPU's you add. With lower-end GPU's like this, I don't see any reason whatsoever. Without even looking to the nVidia camp, one of these puppies might well be outperformed by a Radeon X1900 XTX, and would easily fall far short of two X1900's in Crossfire mode. Crossfire on this dual-GPU board wouldn't improve things much.
Click to expand...


but none of these "gemini" cards on the roadmap are ment to take the high end out. I'm not quite sure why thats an issue. AFR should be quite managable as well as far as profiles go as well. I would even venture to wonder if these cards wont be shipping with AFR enabled by default for 3D apps rather then single card mode requiring CrossFire profiles.
 
Dave Baumann said:
Load balancing should be a little more straighforward with supertiling than it is with the SFR.
Click to expand...

But has anyone done a check on what percentage of the time supertiling actually provides performance scaling somewhere even in the general neighborhood of theoretical with modern games? Certainly my impression is much less often than 50% of the time. Much, much less in fact --tho I'd be happy to see a formal analysis of the question with a range of games.
 
Its a little strange to see X1800XT and X1800XL in front of 7800GTX 512MB and 7800GT respectively. Cherry picking benchmark results I guess.
 
What should the "Hellfire" parts be? Fudo's senseless blabber ("like gemini but in one chip", huh?) didn't tell me anything.
 
SugarCoat said:
I also noted there doesnt appear to be the extra pins on the DVI for a CrossFire connection nor the tabs for the new bridges we saw in the pictures now removed. Still depending on the master card problably (they wouldnt do 4GPU over the PCIE Bus like X1300/X1600 would they?) if they allow CrossFire with these cards at all.
Click to expand...

Keep in mind that this is a X1600 Gemini setup which probably don't even support the bridgeconnection like what you saw in the pictures that were removed of the RV560/RV570. So I wouldn't be suprised to see a RV560/RV570 Gemini board which does have the new bridge-thingy to enable 4GPU.
 
You must log in or register to reply here.

Similar threads

O
Microsoft Xbox Roadmap (2013) (VGleaks)
5 6 7
Replies
138
Views
27K
N2O
N
T
Sony Cell BE / RSX shrink & integration roadmap
2 3
Replies
57
Views
26K
babybumb
B
C
An overview of Qualcomm's Snapddragon Roadmap
3 4 5
Replies
91
Views
27K
Laurent06
L
R
Intel Price drop coming and future roadmap
Replies
10
Views
3K
SugarCoat
S
N
Intel G965 to support SM4.0?
4 5 6
Replies
104
Views
19K
Zaphod
Z
Back
Top