http://farm4.static.flickr.com/3179/...0ab5f6de_b.jpg
http://farm4.static.flickr.com/3121/...dd77d49b_b.jpg
Chiphell
Printable View
This might be the first ATI reference card with a black PCB in a long time, if ever...
<3 my black pcb 8800 ultra
Beautiful!
Is there any word on the launch date?
Nvidia is screwed !
Well... it's a beast for sure...
Oh wow, that thing looks sexy...
I'm wondering if the performance will be the same with two 4870's in CF.
It really looks pretty.
Oh man that's nice. Should pack a punch too!
Me wants a black PCB on my 4870XT :(
It looks fantastic,but I hope that will work even better
nice design
they could make that pcb even shorter...
Benches we need, lots of them ;)
PLX chip in the middle = no performance gains over regular 4870 in CF just like the 3870x2
I mean, I could be totally wrong, but I for some reason am under the impression that crossfire-on-a-chip performs better than normal crossfire.
Correct me if I'm wrong, anyone.
Does that PLX chip act as a PCI-E bus or something of the sort? Kind of like emulating crossfire?
Jesus that PCB is a mess.
Well this makes it it sound like it won't be CF on a card:
CNET Article
Quote:
For ATI, the execution of this chip-ganging strategy is the key. And this is where ATI appears to have been successful. "The inter-processor communications. Getting that to work has been the trick. This is what ATI has done. They've come up with this stellar way of doing inter-processor communications so they can in fact get the scaling," according to Peddie.
So basically the card will not be CF on a board but will be a new interconnect between the GPU's. Also, it seems to suggest that the CrossFire connection was redone in the RV770 or at least for R700 since Peddie suggests that the CF connection is no longer at the very very end of the pipeline. Maybe that's why recent CF results of the 4800's show incredible scaling when it works? Or if as anandtech suggested and it was disabled on the RV770 cards, then it might be turned on for this and is different from the current 4800's.Quote:
AMD-ATI's upcoming R700 (rumored to be called the 4870 X2) two-chip graphics board will be the ultimate test of this strategy.
"It's a new proprietary inter-processor communication technology. If they put these two chips on one board and it does scale properly, then they have pulled off a coup," he said.
"When you gang up graphics chips (using the traditional Scalable Link Interface or CrossFire technologies) they roll off pretty fast. ("Roll off" implies that performance doesn't scale up well.) "So when you put two boards in, you don't get twice the performance but you (only) get one and a half. You put four boards in and you (only) get about 1.7, 1.8. What ATI is saying is that with two chips using (their) proprietary inter-bus, they will get 1.8 (the performance) with two chips. If that's true, you can expect to see four of them giving you something around 2.5."
Getting 2.5 times the performance from four boards would be a masterstroke for ATI.
The previous ATI dual-chip solution was very different, Peddie said. "The HD 3870 X2 was not a proprietary bus but a CrossFire connection. The CrossFire connection and the SLI connection are at the very, very end of the pipeline. Not the most efficient place to do an inter-processor communication. That's one of the reasons ATI has abandoned it."
Anyways, with the plethora of information of R700 that was just released it seems like ATI is pushing hard to get these cards out fast.
And quit thread crapping Sr7. This isn't the place for agendas.
Has the cooler been spotted anywhere?
I fraking KNEW it....
Anyone else notice the size of the PLX chip..? I think AMD leveraged some type of their hyper-transport technology with this newer PLX chip. It would confirm the hints AMD was mentioned a few months back about a new internal bridge.
Did I mention is was fraking black...?
nice.. black PCB;)
DAAMIT look at the transistors on the back of the PCB :eek:
Black PCB FTW !!
WTF! Why is it black? Does that mean something? Maybe it means Nvidia is about to get slaughtered :shocked::eek:
So I guess this is the new smaller and better PLX?
Its not HT - the CNET article says its a new proprietary interconnect so I doubt its based on anything before. Likely what the port does is allow for information to be shared earlier in the pipeline instead of at the end like SLI and CF currently are.
so the 4870X2 will feature a CF Version 2 ??
frankly I dont care what they decide to call it.... I just want to see how 2 of these 4870X2 cards stack up against 3 x GTX 280 and then against 2 x 9800GX2s
Which would give better performance hopefully?
I'm just a little disappointed in some of the crossfire scaling I've seen (from what I've seen it of course helps a lot at high resolutions, but at lower resolutions it's not too impressive). If this could give better performance at lower resolutions, this would be an awesome high-end card for anyone.
Multi-GPU at lower resolution will always tend to scale lower because its more likely to be CPU limited at those resolutions early on. That and games have to be designed to support multi-GPU's. For instance, FEAR seems to scale exceptionally well while other games took a bit longer to get appreciable advantages with multi-GPU (like World in Conflict, it took a few driver releases from both sides before multi-GPU scaling improved).
Dude, are you so incredibly dense, that all you do is spout off and spew banter without any recollection or reference...?
AMD had mentioned (and it was rumored) months ago that ATI was going to leverage alot of AMD's chip-to-chip communication on their newer cards. Looking at this newer PLX chip and taking the tine to study the actual picture.... one can see there is indeed something going on.
Then add in all the technical reviews of the actual R7XX architecture, you would notice there is a new bus interconnect. Then there is the press release about this new "more powerful" interconnect.
I learned to connect-the-dots when I was 5...
Your pessimism is alarming, it's one thing to question or be weary, it's another to troll and thread-crap.
OK guys - he's gone for now. Please continue topic discussion
Guys, I'm not a mod and that's not my job, but please, stop being that aggressive with Sr7 just because you disagree with him, even if he's really wrong. Don't call him "stupid" or "thread-crapper" or whatever just because he's got a different point of view, that's not interesting, he didn't insult anyone ! We're talking about graphic cards, that's all :)
I agree we don't need it, but it's been carried on from thread to thread and you can look at his history yourself if you want. Good riddance for now. Now, back on topic:
Xoulz:
Chip to chip communication doesn't necessarily mean HT but the interconnect can certainly share information for processing to sale.
Eric Demers posted this on B3d a few days ago on why scaling can be super linear:
One possible use of the interconnect is to access the cache of the other chip in order to improve scaling.Quote:
As for super-linear performance, it's possible -- The actual amount of cache is increased as well as the number of functional units. That means that it's very possible for data to stay in cache longer in xfire, than in regular rendering on a single card. That will lead to performance higher than 2x.
Either way though, it seems like ATI is being very hush hush about R700 still, especially the interconnect since at the press briefing they didn't elaborate on the interconnect or CrossFireX Sideport seen on the architecture diagrams.
The only reason I don't think AMD had a ton to do with this was because R700 was supposedly laid out two years ago, before AMD even acquired ATI. Unless they were waiting at that point already for AMD's technology, which seems unlikely, they probably had other designs already in the works.
:rolleyes: please
It's a single card.... it doesn't use Crossfire or SLI. *dee dee dee*
AMD isn't pushing Crossfire, because they had mentioned about 5~6 months ago that their "enthusiast" cards would most likely be dual-core type cards.
Knowing the early problems Intel had with their first gen "dual-core" CPU's and knowing what AMD did... can also be done with GPUs as they advance and mature into sophisticated multi-GPU solutions. That is the direction the GPU market is heading btw... or do you think motherboards with two video cards, lots of noise and sucking watts is the future..?
Just looking at the PCB, indicated that this might be AMD (ATI) second gen multi-GPU. You know nothing and indicate you've read even less. Indicating that everything you've just tried to spew may be moot.
But, since you know so much, I will yield the floor to you, while you explain to the community exactly how this newer PLX chip works and the technology behind it, since your somewhat an authority on this.
Looks very nice indeed. Think ill be picking up one of these babies when its released.
my Hammer..
Please don't call people "stupid" in the threads even if they are.
Make your points without attacking the person please.
If we don't have courtesy between us we have caos and that can't be allowed. I know, at times it is tough when you see someone say something that you know is dead wrond but calling them names doesn't help.
Thanks for reading and now this old fart needs sleep.
Night guys.:up:
Looks ok. ATI really need to work on the CF drivers if they want to continue using dual gpu cards as their very top end cards in the future.
please please please lack latency and micro stutter issues, if that is killed, my Nvidia days might be over. Hard to pull away from CUDA though, being that i actually play some of the games on the list.
I have another wish, ATI talk with Nvidia and release CUDA/physx drivers and im sold. One can wish...
Just thought I would point out the one CF bridge...
Bye bye +5TFlops of GPU power.
It's supposedly 2 x 2TFlops anyways for the two in CFX. I doubt they'd add the double CF connectors and allow 6 or 8 GPU configs anyways... the driver coding for that would lead to suicide
ah, 5 tflop (4x 1.2) unOCed is good enough for the time being :D
guess rv870 is gonna come out in 6-8 months, doubling (i guess even more..) flops again :clap:
Looks like this thread is officially swarming with mods...
Either way, the card looks promising. If there's no microstutter, and it scales like it should, I may end up getting one myself. We'll have to see how this all pans out!
why would u want CUDA when u have openCL, we need a free SDK more than CUDA, and NV will have to share agea or it will fail since havoc is openCL and owned by intel so its coming
dose any1 else like the digital PWMs with controller chips instead of choke, if it works like my dfi board this will be a great improvement and drop the TDP by about 10%, its just to bad that it has qibmonda and not samsung, qib isnt the only gddr5 maker now so it should be retired since its unstable and the samples arnt at the 4.5-4.6ghz that we were promised
I must admit this looks like a huge step forward (again) from the red or are they now black? team. Who would have thought that when Jen Husung or whatever his name is from nVidia was talking about opening a can of Whoopass he was talking about a Can given to him from those clever Canadians at AT Technologies.
Does anyone know when this R700 card is going to be released?
I was waiting on the HD R4870 1GB and planning to CF them, but if this card is out in July then I would be wise to wait as it maybe cheaper or around the same price but better performance than 2 x HD4870 1GB cards.
Would you be able to access the full 1GB of RAM on the R700 or is it going to be the same old 512MB of VRAM due to the way multiGPU's work?
John
I'm in the same boat. I will be in a position to buy 2 x 4870 on the 8-10th of July. I was wondering when this will be launched since i won't have to change my motherboard if i get it instead.
Surprised to see it's black and that there is a bridge chip...
But anyways, it's multi GPU that means I'm not interested.
Looks like no one has one of these up yet, but here's a comparison on PCB length compared to the HD3870X2 and 8800GTX/Ultra and the mounting holes for aftermarket compared to the HD3870X2 (Looks very close. edit: Correcting for lens distortion (the HD4870X2 was taken at a slight angle), the mounting holes likely match it seems :) ):
http://dl.getdropbox.com/u/2028/Xtre...B_Overview.jpg
The mounting holes look close to the 3870X2... that would be sweet if we didn't need a new water block for it!
But there isn't any aftermarket cooler except for water,and you cant use it because of different digital position (those between rv770 chips and above plx)
It appears there's a dedicated phase for the bridge chip with inductors and all. Gotta be one hot bridge...
Below the vDDQ inductor which is below the slave GPU's choke pack. Big square inductor with tantal cap on it's left and low&high mosfets on the right.
yeah looks like its not shared memory (or is it?? ) ....
hopefully it will be less than $599USD MSRP with all those GDDR5 chips
Believe me, that thing is gonna crush gtx280 so hard, even chuck n. could do that :D
It's clearly just like HD3870X2; 256bit per GPU.
Shared memory is nothing but a myth. It's practically impossible to implement as no bridge chip can offer the required bandwidth.
i think it's really cool to see ATi catching up.
Now, all i need is a new fps that is so terribly demanding on graphics power and i'll get one :D
good looking board.like it alot better than red:up::up:
@largon
ahh thanx for clearing that up
This 4870 X2 will come with gddr5? Or with gddr3?
DDR5 normally. They would be stupid to put ddr3 on their topend card.
Chips' part #s and traces looks like GDDR5.
good positioning of the power connectors facing down like that. works well routing cables down
Too bad in most cases the PSU is above the cards...
Meh, so the single memory pool rumors are wrong, and it's 512MB GDDR5 for each card? Meaning this uber-top end card will have only 512MB memory?
I guess 1GB DDR5 x 2 would be too expensive, though.
i hope that's a super-clever-chippy inbetween the gpu's :slobber:
...any chance of squeezing more ram onto that card? or is it just impossible at this point?:shrug:
either way; yes, i'm interested. :D
thats the chip your looking at:
http://www.plxtech.com/products/expresslane/pex8648.asp
Hornet331
But certainly, it does not look like a PEX 8648?
damn i see a damn hot card
hope the cooler is better than the other 40**
So this proprietary new interprocessor communication they talked about.. Either it was a load of hype BS about a PCI-E 2.0 PLX switch, or they're saving it for R800, or.. There's more to that chip?
I don't think that most of the communication will go through the PLX chip but rather through a direct connection between the 2 chips. That connection will probably (I'm pretty sure though) not provide enough bandwidth to make some kind of memory pool sharing possible but it will probably provide increased scaling.
@Hornet331: I think it will beat 2 4870 in CFX in most situations, it will only lose on some rare occasions.
Dave Baumann (ATI employee) has hinted over at Beyond3d that it may not be that PLX chip that you linked to Hornet....
Linky:
http://forum.beyond3d.com/showthread.php?t=48691&page=8
further down on that page.
BTW guys, I'm just speculating...I hope you guys have no objections with that :p:.
As for the chip, it probably just has a metal heatspreader over it.
Not to mention all PLX chips seem to look alike. :D
Do you have any info about whether this card is going to have a possible mechanism to reduce or eliminate microstuttering or multi-GPU driver issues? If not I don't think it will be beating the GTX280 in games with poor or nonexistent multi-GPU support since a single 4870 can almost never meet/beat a GTX280. OTOH, unless something goes horribly wrong the R700 certainly should utterly destroy the GTX280 in frame rates in games with multi-GPU support. In theory 2 4870s should give about the same performance as GTX260 SLI. I am really hoping that AMD/ATI is going to do something about the microstuttering issue though. Having parts of the game perform at the same level as a single card quite significantly reduces the value of multi-GPU for me.Quote:
Originally Posted by w0mbat
damn, i do not see how a 6-pin plus 8-pin would be nuff 4 this...
tis like powering 2 8800gtx chips on 1 card..
i thought it would require 2 8-pin power connectors!
1300W PSU minimum for xfire???
there's only one problem with your theory. A single 4870 beats the gtx 260. Considering xfire is supposed to scale better than sli and the possible there may be an internal ht bridge or something like that for a fast low-latency inter-connect, then it's safe to say the 4870x2 will be closing in on gtx 280 sli
2 x 8pins? Even the 1.1 PCIe + 6pin + 8pin is 300W, if it was over 300W imagine the cooler needed.
The HD4870 only beats the GTX260 if you choose your reviews and the games in the reviews and your resolution with great care. OTOH, I could go cherry pick reviews that clearly shows the HD4870 losing badly to the GTX260 in multiple benchmarks. The fact is that the GTX260 and the HD4870 are trading blows. Although in reading reviews from sites that I know and trust the HD4870 seems to be losing by greater margins when it loses than the GTX260 in the games where it loses. Also, it looks to me like the GTX260 has more OC headroom. With an apples to apples water cooled OC comparison I believe the GTX260 OC is going to win against the HD4870 OC in nearly every benchmarks except those few where the HD4870 does unusually well.Quote:
Originally Posted by aliG
Also what is with this CF is "supposed to" scale better than SLI? In the benchmarks I have seen if anything it seems to scale worse and CF is often stated to have worse driver support in games than Nvidia. The game industry has been very Nvidia biased since the 8800GTX took the crown. Hell there was a lot of speculation that AMD/ATI may not even be around anymore due to their huge financial problems which are not even over yet.
The internal HT type bridge idea is extremely speculative. In fact most of the current rumors seem to be that R700 is going to be some sort of crossfire on a card with a faster bridge chip. Remember all the crazy rumors before R600 was released? Some of the rumors were even talking about improved GPU communication IIRC. I suggest we adopt more of a wait and see attitude rather than these type of ultra-optimistic assumptions.
When is premier of this beast ? :)
Happy to see a black pcb on ATI.