Lots of R600 news [Archive] - XtremeSystems Forums

Lightman

12-10-2006, 11:15 AM

Some good news for waiting fans :D !!

If this chip really consume only 80A @1.2V then we should get less power hungry card than X1950XTX.
Chip: 80A*1.2V=96W
Memory 1GB (16 chips) @ unknown clock: ~2W*16=32W
Other circuity: ~20W
Total: 96+32+20=148W

Remember that this is absolute maximum for chip to consume 80A, everything else is only approximation (hopefully close one ;) ).

nn_step

12-10-2006, 11:19 AM

sexy, more than 800mhz...:slobber:

metro.cl

12-10-2006, 11:20 AM

hmm hard to believe, time will tell

biohead

12-10-2006, 11:24 AM

they better not release same cards with either gddr3/4, one having to be 'lucky' again. like x1900 xt/xtx - u either get 1.1 or 1.2ns.

r600 & gddr3 makes no sense though :/ to me anyways

Rovtar

12-10-2006, 11:24 AM

when we will se2 tests?

i want to see how good will be

NickS

12-10-2006, 11:25 AM

sexy, more than 800mhz...:slobber:

Agreed, ATi FTW! :woot:

Tonucci

12-10-2006, 11:30 AM

they better not release same cards with either gddr3/4, one having to be 'lucky' again. like x1900 xt/xtx - u either get 1.1 or 1.2ns.

r600 & gddr3 makes no sense though :/ to me anyways

Maybe gddr4 is overkill ? lol. Its wise to be marginally faster but with alot of headroom :slobber:

Ubermann

12-10-2006, 11:31 AM

....
R600 DIE SIZE compared to G80 small, the transistor are more than G80
720 million transistors

Sounds more powerful then the ATI RAge+ PCI card i have in this comp right now..might even be twice as fast.

[XC] Malkec

12-10-2006, 11:32 AM

r600 will smoke g80's ass big time :slap: :D

Cuthalu

12-10-2006, 11:33 AM

r600 & gddr3 makes no sense though :/ to me anyways

With 512-bit it does make sense. They obviously don't need THAT much bandwith.

NickS

12-10-2006, 11:39 AM

With 512-bit it does make sense. They obviously don't need THAT much bandwith.

R580+ was still bottlenecked by GDDR4 because of some weird clocking bug. Go figure.

eXceededgoku

12-10-2006, 11:43 AM

But whats the point in not making the card as fast as physically possible ala GDDR4... no point in GDDR3 at all, except for budget models.

safan80

12-10-2006, 11:49 AM

I want to see a picture of the card side by side with a 8800GTX before I'll believe it's shorter.

Cuthalu

12-10-2006, 11:53 AM

But whats the point in not making the card as fast as physically possible ala GDDR4... no point in GDDR3 at all, except for budget models.

It may not increase performance enough? GDDR3 is cheaper, so it might be stupid idea for them to have lower profits without speed increase at all or with marginally little increase.

SparkyJJO

12-10-2006, 11:56 AM

they better not release same cards with either gddr3/4, one having to be 'lucky' again. like x1900 xt/xtx - u either get 1.1 or 1.2ns.

r600 & gddr3 makes no sense though :/ to me anyways
I got 1.1ns memory on my XT yet it can't handle XTX speeds :( I got dud memory I guess.

I like the sound of this R600 :D Too bad my income allows for almost no upgrades.

xs64

12-10-2006, 12:10 PM

i wanna tell a little thing about DDR3 and DDR4

I have 1 X1900XT and 1 X1950XTX
I oced my X1900XT @ 650/720 and keep my X1950XTX at DF

Guess what , in NFS Carbon ( using Fraps ) i got 41 pfs with my X1900XT and 64 fps with my X1950XTX

So i love DDR4 !!!!

DilTech

12-10-2006, 12:16 PM

But whats the point in not making the card as fast as physically possible ala GDDR4... no point in GDDR3 at all, except for budget models.

Simple, they have to leave room to grow.

Both NVidia and ATi are banking on their new architectures to last a few years. Kind of like how the R300 architecture lasted ATi from the 9700pro to the x850xtpe, or how the nv40 architecture lasted nvidia from the 6800ultra to the 7950gx2.

If you have to keep making new designs because you went all out on your new chip, you're essentially shooting yourself in the foot, as that makes your profits on the architecture that much smaller. An obvious example is that the R520 architecture wasn't nearly as profitable as the R300, as it lasted them only about a year, while NVidia's NV40 architecture made them huge amounts of cash because it lasted them well through 2 years.

NVidia left themselves plenty of room to grow with the G80, from die-shrinks to upping the memory bus to faster memory. ATi need to give themselves that same headroom, and if ATi start with a 512bit memory bus and 80nm, the only way to refresh will be faster ram. 65nm won't be ready for chips this complex at TSMC in time to compete with NVidia's refresh.

Think business, they have to give themselves a cheap easy way to refresh the cards, and I can't think of anything cheaper than a pin to pin compatible memory switch. They can keep the same pcb, same cores, and merely update the ram like they did with the x1900xtx-x1950xtx and go right back to charging top-dollar for it again.

Eastcoasthandle

12-10-2006, 12:18 PM

I got 1.1ns memory on my XT yet it can't handle XTX speeds :( I got dud memory I guess.

I like the sound of this R600 :D Too bad my income allows for almost no upgrades.
It may just be lower binned 1.1's

Cuthalu

12-10-2006, 12:18 PM

xs64, its GDR, and those cards are not R600 and they have only 256-bit memory interface.

G80 uses GDR3 and 320/384-bit memory and it isn't very bandwith-dependant, and R600 with it's 512-bit would have even more bandwith than G80.

cantankerous

12-10-2006, 12:28 PM

I am wondering, if this card really is suppose to be SHORTER and require LESS power than the 8800GTX does that mean it will only require ONE PCIE power connector or are we still going to need two? The 8800GTS runs fine on one native and currently available PCIE power connector which makes for a very easy video card upgrade if you don't have to get a new PSU to go with it. There was rumours of the R600 possibly needing a new type of power connector to deal with the new power requirements though it seems this may no longer be necessary. I would love to keep my current power supply in my sig (Powerstream 600w) if the new card truly will use LESS power than the X1950XTX which my system currently runs an equivelant to (X1900XTX overclocked) without issue. I was JUST about to order a GTS tomorrow thinking the R600 would be too long for my case and require a new PSU to power it which made things real expensive and full of hassles. Perhaps I will just hold off now to see how the ATI offering holds up as it just may be better.

Yoxxy

12-10-2006, 12:54 PM

I bought an Enermax Galaxy 1000w and it comes with an 8 pin pci-e connection. In the manual it says "for upcoming generation of cards"

Steensen

12-10-2006, 12:56 PM

Ziwro

12-10-2006, 01:07 PM

I bought an Enermax Galaxy 1000w and it comes with an 8 pin pci-e connection. In the manual it says "for upcoming generation of cards"

Enermax Galaxy 1000W and 8Pin PCI-E connection :confused: Do you sure that ?

Lightman

12-10-2006, 01:12 PM

Sounds interesting.
Would be nice that the R600 can do with the current 6-pin PC-E plug + PCI-E slot, but yet againg, 148W is too close to the standard, for the producers being able to balance the load correct.

Think about overclocking!
I'm certain that we will see 2 PW connectors on R600 boards, but it may work with only one. Besides in link provided by w0mbat they are saying that power circuit on board will be capable of providing 160A for chip only :slobber: with possibility of upgrade up to 240A:eek2: !!

This card will be ready to take as much as 300W:eek: ! Certainly Windows Vista Capable ;) .

biohead

12-10-2006, 01:55 PM

This card will be ready to take as much as 300W:eek: !
umm no. We've been over this several times.

Lightman

12-10-2006, 02:05 PM

umm no. We've been over this several times.

Read my post again please!

I was pointing out that card/PCB design will be ready not will eat :up: .

grimREEFER

12-10-2006, 02:07 PM

has the dx10.1 support been proven true/false?

Helmore

12-10-2006, 02:30 PM

I bought an Enermax Galaxy 1000w and it comes with an 8 pin pci-e connection. In the manual it says "for upcoming generation of cards"
Are you sure it's a PCI-E connector? I like to know this as well, as I'm interested in this PSU, but I can't find any information regarding an included 8-pin PCI-E connector. So please elaborate some more, maybe you can show it off in the PSU section:p: .

heikis

12-10-2006, 02:45 PM

Are you sure it's a PCI-E connector? I like to know this as well, as I'm interested in this PSU, but I can't find any information regarding an included 8-pin PCI-E connector. So please elaborate some more, maybe you can show it off in the PSU section:p: .

im sure it is the 8pin EPS12V connector for the motherboard. look at the review of that galaxy.
it has: http://www.virtual-hideout.net/reviews/Enermax_Galaxy_1000w/galaxy18.jpg
and pci-e: http://www.virtual-hideout.net/reviews/Enermax_Galaxy_1000w/galaxy21.jpg

XSAlliN

12-10-2006, 02:54 PM

has the dx10.1 support been proven true/false?

False - it's incomplete, it would be stupid to ad it in R600 as it is.

Go ATi(AMD), Go! - slaughter those Intel Fanboys with their integrated chips. :frag:

Helmore

12-10-2006, 03:19 PM

im sure it is the 8pin EPS12V connector for the motherboard. look at the review of that galaxy.
it has: http://www.virtual-hideout.net/reviews/Enermax_Galaxy_1000w/galaxy18.jpg
and pci-e: http://www.virtual-hideout.net/reviews/Enermax_Galaxy_1000w/galaxy21.jpg
In that review and in those pics it is confirmed that the newer revision Galaxies have two 8-pin PCI-E connectors. In one of the pics that you are linking you can clearly see an 8-pin connector being labled as 'For Graphics cards only". This would also make the possibility for R600 to need an 8-pin PCI-E connector even greater.

Revv23

12-10-2006, 03:24 PM

Simple, they have to leave room to grow.

Both NVidia and ATi are banking on their new architectures to last a few years. Kind of like how the R300 architecture lasted ATi from the 9700pro to the x850xtpe, or how the nv40 architecture lasted nvidia from the 6800ultra to the 7950gx2.

If you have to keep making new designs because you went all out on your new chip, you're essentially shooting yourself in the foot, as that makes your profits on the architecture that much smaller. An obvious example is that the R520 architecture wasn't nearly as profitable as the R300, as it lasted them only about a year, while NVidia's NV40 architecture made them huge amounts of cash because it lasted them well through 2 years.

NVidia left themselves plenty of room to grow with the G80, from die-shrinks to upping the memory bus to faster memory. ATi need to give themselves that same headroom, and if ATi start with a 512bit memory bus and 80nm, the only way to refresh will be faster ram. 65nm won't be ready for chips this complex at TSMC in time to compete with NVidia's refresh.

Think business, they have to give themselves a cheap easy way to refresh the cards, and I can't think of anything cheaper than a pin to pin compatible memory switch. They can keep the same pcb, same cores, and merely update the ram like they did with the x1900xtx-x1950xtx and go right back to charging top-dollar for it again.

This doesn't really matter, but just FYI, r520/r580 is still based off of r300. Just tacked on more shaders mhz and SM3. r600 is the first complete redesign ati has had in years. it was in the works before even r520 was.

nn_step

12-10-2006, 03:26 PM

False - it's incomplete, it would be stupid to ad it in R600 as it is.

Go ATi(AMD), Go! - slaughter those Intel Fanboys with their integrated chips. :frag:
umm they are already doing that with their current line of integrated graphics chips ;)
Heck even Mark Rein, vice president of Unreal studio Epic, claims Intel graphics are bad for the gaming industry http://www.tgdaily.com/2006/07/12/epic_integrated_graphics/
Not to mention the long list of Games that they can't even support but ATi integrated graphics CAN.

onewingedangel

12-10-2006, 03:33 PM

Flame baiting much?

nn_step

12-10-2006, 03:34 PM

Flame baiting much?
naw I call it Fair and Balanced... Kind of like Fox news :rolleyes:

K404

12-10-2006, 03:58 PM

Im out of date with ATX revisions...have 8-pin PCI-E plugs been officially announced in at ATX revision? Is it part of the PCI-E.2 thing I hear about?

Cybercat

12-10-2006, 05:09 PM

This doesn't really matter, but just FYI, r520/r580 is still based off of r300. Just tacked on more shaders mhz and SM3. r600 is the first complete redesign ati has had in years. it was in the works before even r520 was.the biggest thing retained from the R300 still in the R5x0 is the shader layout (Vec3 + Scalar). Everything else is nearly unrecognizable. In the same vein, the R600 will likely retain the shader layout of the R300-R580, but will change everything else. So I argue that the R600 is just as much a change as the R520 was.

ahmad

12-10-2006, 05:36 PM

the biggest thing retained from the R300 still in the R5x0 is the shader layout (Vec3 + Scalar). Everything else is nearly unrecognizable. In the same vein, the R600 will likely retain the shader layout of the R300-R580, but will change everything else. So I argue that the R600 is just as much a change as the R520 was.

If everybody always started from scratch we would still be driving fords..

I would say r520 was a huge improvement over r300 in terms of innovation and features (avivo). In fact it was so good, that others have decided to copy them...

But I wasn't expecting the r600 to be 720M transistors. Thats just insane.

Cybercat

12-10-2006, 06:11 PM

If everybody always started from scratch we would still be driving fords..quite, ATI just wants to stick with the highly refined shader compiler and such that they've always used, but it does seem like NVIDIA is more inclined to start from scratch more often than ATI is. The G80 has nothing retained from previous architectures that I can see, in fact it's a little surprising they still pack the driver in the same Forceware package with all the others.

Unfortunately for NVIDIA that probably still won't give them the win.

I would say r520 was a huge improvement over r300 in terms of innovation and features (avivo). In fact it was so good, that others have decided to copy them...Other way around, ATI needed a video processing model to compete with Purevideo, so really they were the ones copying, although they took the Google approach and improved on it while offering it free of charge (which NVIDIA eventually did, although the Purevideo codec still costs money).

But let's not look over the other changes, the asymmetric ALU/ROP/TMU configuration, the rebuffed ROPs, and memory controller. Add in the low-level stuff necessary to support SM3.0 and the R520 is nearly incomparable with the R300.

But I wasn't expecting the r600 to be 720M transistors. Thats just insane.I wasn't expecting high clockspeeds with such a high transistor count, and a relatively new manufacturing process.

nn_step

12-10-2006, 06:21 PM

If everybody always started from scratch we would still be driving fords..

I would say r520 was a huge improvement over r300 in terms of innovation and features (avivo). In fact it was so good, that others have decided to copy them...

But I wasn't expecting the r600 to be 720M transistors. Thats just insane.
that is true for all things. It is the nature of Exaptation (http://en.wikipedia.org/wiki/Exaptation)
or the belief that everything is built upon previous work.
True the r580 was based on the r300
by the same Token Linux is based on FMS and Conroe is based on the 4004

Linux used ideas from MINIX
MINIX used ideas from UNIX
UNIX used ideas from MULTICS
MULTICS used ideas from CTSS
CTSS used ideas from FMS

NightCrawler™

12-10-2006, 11:35 PM

This thread is useless without pics.....

Fatal Error

12-11-2006, 12:54 AM

XSAlliN

12-11-2006, 03:20 AM

R600 & K8L is the most expected pc-related products that i can remember for a long time.

AMD & ATI gives more and more information and teasers about their products but ok when are we gonna see them at last?
I hope that they aren't just big talks.

While ATI trying to complete RD600 Nvidia smash her in all categories.

Smash what? - it has no competition at the ending of this year :).

I don't care that much about R600/G80 - easy come, easy go! - nice performance at first half of 2007 but there is still more to fallow, for me a "Multi CPU + Multi GPU platform whit DDR3 Modules" - that is MOST EXPECTED PC-RELATED!:weapon:

Fatal Error

12-11-2006, 03:29 AM

Smash what? - it has no competition at the ending of this year :).

Emm that was i mean :rolleyes:
Nvidia already has the G80 and ATI still rumble on paper work and specs annoucement.

DilTech

12-11-2006, 06:30 AM

Emm that was i mean :rolleyes:
Nvidia already has the G80 and ATI still rumble on paper work and specs annoucement.

Yeah, I still wish ATi would've got their card out the door this year, as I honestly was more-so looking forward to it than the G80, but I just couldn't wait til next year.

NVidia's on a roll right now though, from a business standpoint they've been in an amazingly strong position since 2004, and it looks like they can easily keep it going.

G80 gets 2-3 month headstart on sales, and is the only DX10 core on the market for the christmas sales, so rich parents buying their kids computer parts for christmas will all be getting them those.

The midrange G8x cards should see light around march, so when ATi finally get their top-end out, nvidia will be already in the mid-range/low-end field where the real money is. As of right now we have no idea when ATi will follow suit into the midrange/lowend dx10 game, but so far we have no info on this matter.

To throw salt in the wound, NVidia still have their die-shrunk G80 with GDDR4 and probably 1gb ram, which wouldn't surprise me if they up it to a 512bit memory bus since there's no way to hit 1 gig with 12 memory chips, for sometime around march as well.

ATi have really got to stop their wait and see games they've been playing the last few rounds and instead start getting more aggressive. A lot more aggressive... These delays sooner or later is going to put them in a verrry bad situation.

Sure, they can claim their card is faster when it comes out, and it damn well better be 3 months later, but the problem is by then NVidia is almost down to refresh time and if ATi follow suit they kill their old card's market(x1800xt to x1900xtx anyone?) before they can make their money back on it. They end up ontop for a short period of time, and get slapped back down.

Fred_Pohl

12-11-2006, 08:24 AM

umm they are already doing that with their current line of integrated graphics chips ;)
Heck even Mark Rein, vice president of Unreal studio Epic, claims Intel graphics are bad for the gaming industry http://www.tgdaily.com/2006/07/12/epic_integrated_graphics/
Not to mention the long list of Games that they can't even support but ATi integrated graphics CAN.

Supporting games is one thing. Providing playable frame rates is something else entirely.

XSAlliN

12-11-2006, 08:52 AM

That makes sense, but I don't think ATi's delay for R600 is marketing related, if that would be the case they would have release some by now, same thing happened whit x1xxx generation, plus ATi is part of AMD now, after spending those billions of $ on this company, AMD works with products that have priority and guess they have there hands full by now, you might say that they move "Step by step" - one "big mistake" and it's the End of AMD (ATi).

Besides, There won't be any games whit DirectX10 support until 2007, same goes for Vista, so I guess they count on that, as you can see in the Press Room (http://www.amd.com/gb-uk/Corporate/VirtualPressRoom/0,,51_104_543,00.html) they still work with some changes, that might be suitable for the company, since that doesn't concern the general public - you keep whining like babies that want there gaga (R600).

PS.When you analyze a company try tot examine all the aspects, not just the ones that concern you. :fact:

Yoxxy

12-11-2006, 09:21 AM

xlink

12-11-2006, 09:26 AM

:banana::banana::banana::banana:. I want it now.

I have a GeForce 6200 PCI... 64 bit memory interface...

I have 4000$ in the bank and I'm ready to drain my account...

for the midrange R600 based card though...

Yoxxy

12-11-2006, 09:32 AM

The new gamer's edition galaxy has a 8-pin pci-e connector that says "for graphics card only," 2x 6-pin pci-e connectors labled "for graphics card only," and 2 modular 2x 6-pin pci-e connectors. You are also available to purchase a 8-pin pci-e modular cable when the come out. It is rated to 160 watts according the manual. I don't know what a normal pci-e cable is. It also has 2 4pin motherboard cables that say "for cpu only" as to not confuse you, just so people say I am not confusing the two.

http://www.enermaxusa.com/catalog/product_info.php?cPath=21_49&products_id=115&osCsid=47fdc95799f8b46c42338984a032ded0
Here is the link as well. They call it "PCI-E Version 2.0." Unfortunetly it is not on the cable list but I can assure you it does in fact come with the 8-pin cable.

Ubermann

12-11-2006, 09:43 AM

Whats the problem with the 8-pin cable?

Helmore

12-11-2006, 10:52 AM

There aren't many PSUs, with a 8-pin PCI-E connector, around yet that's what so stupid about it and R600 is rumored to use one.

Ubermann

12-11-2006, 10:55 AM

Yes but doesnt GFX cards usualy come with cables ?

Helmore

12-11-2006, 10:58 AM

yeah but all those extra cables are annoying.....

DilTech

12-11-2006, 11:05 AM

Yes but doesnt GFX cards usualy come with cables ?

The question is, what kind of cables would be required for the splitters.

If a 8-pin pci-e 2 cable supplies up to 160W, even 2 standard pci-e cables wouldn't work, as they offer only 75W a piece.

Vapor

12-11-2006, 11:25 AM

I'd bet the 8-pin will just be a 6-pin plus a molex. Even if it's way past rating.

Yoxxy

12-11-2006, 11:34 AM

From reading the manual (a novel concept), there are 6 12v 17amp rails on the Galaxy Gamer's Edition. The 8-pin PCI-E has a rail of its own, that it shares with a pci-e. Both of these cables cannot be in use at the same time, it stresses that quite a bit. It says the spec is for 160w on PCI-E 2.0, but it is able to supply the full 17amps out of the 8 pin.

Possible new cards will be 1 8-pin or 2 6pin's and a molex. Then 1 8pin and a 6 pin. Then God save us.

ether.real

12-11-2006, 11:52 AM

I want some performance numbers already, DAAMIT.

J-Mag

12-11-2006, 01:07 PM

What I have a hard time understanding, is why we need all these damn connectors anyway? I know it is helpful for split rail PSUs, but if we didn't have that safety standard in the first place there would be no need.

18awg wire can handle up to 16 amps, right? Just one PEG connector has 3 so whats the deal?

I was thinking about making a 6pin to dual 6pin splitter to avoid a PSU swap when I add another 8800GTX.

ether.real

12-11-2006, 01:16 PM

I was looking into doing this very same thing as well, and I did some research. Depending on the construction of the PSU, and what type of power transistors it uses, it may work. However, one type of construction would require perfectly balanced transistors because if one gets a little hotter than the other, it will start pushing more current, until it is handling 100% of the power, which is bad news.

JoeBar

12-11-2006, 01:29 PM

I'd bet the 8-pin will just be a 6-pin plus a molex. Even if it's way past rating.
I believe the same also. If they just don't release it with a pci-e only... :p:

Cybercat

12-11-2006, 03:41 PM

To throw salt in the wound, NVidia still have their die-shrunk G80 with GDDR4 and probably 1gb ram, which wouldn't surprise me if they up it to a 512bit memory bus since there's no way to hit 1 gig with 12 memory chips, for sometime around march as well.or they could try a 448-bit bus (896MB).

turtle

12-11-2006, 05:04 PM

or they could try a 448-bit bus (896MB).

lol.

G80's arch has ROP quads tied to memory channels true-enough and that could work, but I doubt they'd do it odd with 14 memory chips, although they may as few were expecting 12 chips and 384-bit. There are 6x64 (384-bit) mem channels currently, and hence 6 quads of rops. I personally don't think we'll see some weird combo...What we will probably see is a 32 ROP 512-bit (8x64) 1GB part, but as i've said on other forums, that's my guess for G90...Prolly with give or take 160~192 shaders.

I think March will be G80 @ 80nm, just like like last gen w/ G70->G71.