ATi 4870 X2 (R700) thread

[Janooo]

Originally posted by: bunnyfubbles

Originally posted by: chizow

Originally posted by: thilan29
the chip they use is different and so won't necessarily be x8/x8. And in all likelihood, if there was a bottleneck in that part of the card from the 3870x2, ATI would have replaced it this time around.

What do you think the bridge chip does? Unless the PLX gen 2 chip is able to make an x16 PCIE 2.0 slot into an x32 PCIE 2.0 slot then it will result in an x8/x8 PCIE 2.0 on a single slot.

Doesn't PCI-e 2.0 double the bandwidth per lane (500MB/s vs. 250MB/s) over PCI-e 1.x (ie an x16 slot on PCI-e 2.0 is 8GB/s vs. 4GB/s of PCI-e 1.x...)?

I'll admit I haven't kept up with multi GPU technology because I've never really considered it as a viable solution for my own use, but have the multi GPU cards already saturated the single x16 slot of PCI-e 2.0 for this to even be a problem on PCI-e 2.0 motherboards?

You are right, it's doubled. I don't believe PCI-e is a bottleneck.

 

[taltamir]

supposedly they do use the PLX chip to split the pcie bus... but the only use it to communicate with the system RAM and CPU...
Inter GPU communcation is now done directly, not through the PLX switch, so that should compensate

The multi GPU configs ARE affected by the pci speed... i have seen a test that shows them dropping 10% in performance going from 16x to 8x, and 30% going to 4x. But that was on older cards that had less total performance.. on the other hand, that was with pcie v1...
So the cards are faster now (requiring more bandwidth), but the baseline is higher, meaning 8x v2 is same speed as 16x v1.

It could still be an issue on pcie v1 motherboards.

But it is also countered by the GPU - GPU connection that bypassess the PLX switch... so the pcie will no longer be used to communicate between gpu, only with the CPU or System Memory.

 

[Janooo]

Radeon HD 4870X2 in the nude, 15% better than 4870 CrossFire?

Radeon HD 4870X2 in the nude, 15% better than 4870 CrossFire?
Written by Andreas G 29 June 2008 20:48

AMD seems to have regained the consumers' support after the launch of the Radeon HD 4850 and 4870. The cards are performing well, especially in scenarios with antialiasing and anisotropic filtering, and even though they cost half ($199, $319) as much as NVIDIA's new high-end beasts ($399, $649) they are not far behind in the reviews. AMD has another thing coming though, the dual-chip Radeon HD 4870X2.

The card, also been known as R700, has been surrounded by many speculations. The most persistent involve how the two GPUs will communicate. Initial rumors claimed a dual-core approach, which is not true. It's not a Radeon HD 3870X2 remake either. The fact is that it's not just the card that has been optimized for multi-GPU work, but the GPU itself as well. The optimization is mainly for improved scaling though, there's still little known about microstuttering. Numbers close to 80% have been mentioned.

Over at Chiphell the following nude pics of Radeon HD 4870X2 have been posted;

The PLX chip between the GPUs now supports PCIe 2.0, but the exact nature of the remaining changes are yet to be unveiled. The card has two 512MB GDDR5 memory buffers, most likely the same 3.6GHz chips found with Radeon HD 4870. We suspect these won't be shared as so many were hoping.

Some performance indications have also been posted and they state that Radeon HD 4870X2 will outperform Radeon HD 4870 CrossFire with around 15% with a decent system in general. We should know more as we approach the launch in late August.

 

[bunnyfubbles]

Right, so as long as you have a motherboard with a RD700, NF700, or X38/48/P45 chipset (or any that I've forgotten that support PCI-e 2.0) there probably won't be any limitations with a 4870X2. Of course noting how popular and extremely common chipsets such as the P35 are...I'm looking forward to seeing some tests even though I have no plans on buying ...I just think its stuff like this that makes me love the move we've made to PCI-e more and more.

 

[lopri]

Originally posted by: chizow
1) Still using a PLX bridge. Basically splits the PCIE x16 lanes for 2 x RV770 GPU, resulting in 2x PCIE x8. As we saw in the Tweaktown review, this can be a problem as PCI x8 2.0 does show bottlenecking with 4850s. This will be an even bigger problem for PCIE 1.1 board owners, as you will only have an effective PCIE x4 2.0 for each GPU.

I don't think this is a correct observation. First, we don't know what kind of PLX bridge that is. It's fairly certain to be PCIe 2.0, but do you know for sure it's got x16 lanes? If I remember correctly that company makes whole bunch of PCIe chips.

And the inter-GPU communication occurs entirely on the board-level in this design (same for 9800 GX2, which had something like x48 PCIe interconnect between the two boards), so what goes to the system are the final frames. So yes, PCIe 1.0 x16 slot can be disadvantaged compared to 2.0, but it's not like the two GPUs are spliting the x16 slot on the motherboard.

 

[BTRY B 529th FA BN]

Lopri, which one of your machines feels the fastest?

 

[lopri]

http://www.plxtech.com/product...presslane/switches.asp

 

[Rebel44]

Originally posted by: BFG10K

When I already have more then 60 FPS I dont care if I have 75 instaed of 79....

So are all cards are equal as long as they?re over 60 FPS?

That was my reply to chizow´s BS about CPU bottleneck. When my bottleneck is my Q6600 @3.4Ghz I am pretty sure that my FPS are more then 60 FPS - and I dont care if I am CPU bottlenecked in 3D Mark or similar stupid benchmark...

 

[lopri]

Originally posted by: BTRY B 529th FA BN
Lopri, which one of your machines feels the fastest?

I'm afraid to talk off-topic, but since I was asked.. My favorite is the E2160/BadAxe2 combo. Can't beat 100% OC and I just love 975X chipset. I think it's the best desktop chipset Intel has made to date.

But I assume you're asking between the rig #1 and #2? Well, I use them totally differently so it's hard to say. Obviously the #1 is the work horse, and heavy multi-tasking and occasional gaming in window mode happens there. #2 is for light work load (i.e. downloading/surfing) and full-screen gaming.

Obviously #2 cannot do what #1 can do (vice-versa), and it's generally less stable than the #1. But #2 does feel snappier for light tasks. I believe it's because Penryn > Conroe. I do, however, plan to compare graphics once my HD 4870 arrives.

/off topic

Anyway, I did notice something interesting from the link I posted above wrt PLX chip.

Gen 2 performancePAK?

1. Dual Cast?s
   Read Pacing?
   Dynamic Buffer Allocation

Hmm..

 

[BFG10K]

That was my reply to chizow´s BS about CPU bottleneck. When my bottleneck is my Q6600 @3.4Ghz I am pretty sure that my FPS are more then 60 FPS - and I dont care if I am CPU bottlenecked in 3D Mark or similar stupid benchmark..

Ah I see, I clearly mistook the context of your original comment. :thumbsup:

 

[Sylvanas]

Lopri:

Gen 2 performancePAK?

Dual Cast?s
Read Pacing?
Dynamic Buffer Allocation


Hmm..

Good find! Could this mean a shared memory pool? It seems like we won't be stuck with the 512mb framebuffer per GPU limitation if the PLX chip 'dynamically' assigns memory per the GPU's needs, which means this card is going to be doing alot to improve the Crossfire implementation. Guess I'll be holding out for this.

 

[bryanW1995]

Originally posted by: magreen
That article and "analysis" by John Peddie was nothing but buying into AMD marketing. The 256-bit memory bus w/ faster ram? A stroke of genius? Gimme a break. Every techie knows you can do that. Making smaller chips that work together? They haven't done it yet. They've yet to release a card to compete with this generation's high end. The 4850/70 are very nice value cards. But don't count your X2 chickens before they hatch. Their proprietary bus instead of XFire is the one thing mentioned in that article that could put them ahead. But it's a complete unknown. "AMD bests nvidia with graphics chip strategy"? FUD.

...damn Peddiephiles.

amd is kicking the shit out of nvidia this round. 85% of the performance w/ 4xAA and 90-95% w/ 8xAA for 46% of the cost is an old fashioned ass-whuppin' where I come from. The nvidia pr machine has been cranking out more cuda info and other bs than even amd did last gen with their "dx 10.1" howls. Face it, when you're on top you focus on performance (or in this case the incredible price/performance), when you're struggling in comarison to the other guy you talk about ancillary benefits or supposed long-term advantages that won't show up until after the end of the card's useful life.

 

[bryanW1995]

Originally posted by: chizow

Originally posted by: SlowSpyder
You are right, something is bottlenecking the CF'd 4870's a bit, but I don't see how it is the CPU when all over the graph that same CPU can go well past the 4870CF FPS. Driver problems? PCIE lane bandwidth? Something? I don't know what, but I don't see how it is the CPU.

Well I wouldn't focus so much on 5-10FPS differences, especially when comparing to different vendors. Sure there's going to be scaling/driver/vendor/chipset/bus issues that are going to prevent exact results but its much more clear when you compare products from the same vendor. That example and a few others in that review are classic examples of CPU bottlenecking and show you may want to upgrade that CPU before you add that 2nd GPU.

I mowed through most of the bfg/chizow slugfest. In that bfg did agree that cpu throttling did occur on some of these ultra high-end multi gpu configurations. That makes sense, especially with something like sli gtx 280 or 4870x2; ie, don't buy a 4870x2 for your x2 3600+ and expect to pwn your buddy in crysis

 

[AmberClad]

The power connectors - they seem like they're facing downward? I wonder if they're keeping that on the final card.

 

[chizow]

Originally posted by: lopri

Originally posted by: chizow
1) Still using a PLX bridge. Basically splits the PCIE x16 lanes for 2 x RV770 GPU, resulting in 2x PCIE x8. As we saw in the Tweaktown review, this can be a problem as PCI x8 2.0 does show bottlenecking with 4850s. This will be an even bigger problem for PCIE 1.1 board owners, as you will only have an effective PCIE x4 2.0 for each GPU.

I don't think this is a correct observation. First, we don't know what kind of PLX bridge that is. It's fairly certain to be PCIe 2.0, but do you know for sure it's got x16 lanes? If I remember correctly that company makes whole bunch of PCIe chips.

And the inter-GPU communication occurs entirely on the board-level in this design (same for 9800 GX2, which had something like x48 PCIe interconnect between the two boards), so what goes to the system are the final frames. So yes, PCIe 1.0 x16 slot can be disadvantaged compared to 2.0, but it's not like the two GPUs are spliting the x16 slot on the motherboard.

Again, it doesn't matter what kind of interconnect is between the GPUs, if I had to guess that's what the CF sideport is for. You still run into the problem of the physical PCIE 2.0 slot the card is placed into. AFAIK you cannot make this larger than x16 if the physical traces and the slot itself are limited to 16 lanes. Even if you were passing everything through a single GPU and not splitting the PCIE bus you'd have ~2x data (approx. 4870CF) passing over an x16 bus rather than 1x data over an x8 bus. Normally you would think this sufficient but based on the Tweaktown review of 4850CF on P45 and X48 that's not the case.

Originally posted by: bunnyfubbles
Right, so as long as you have a motherboard with a RD700, NF700, or X38/48/P45 chipset (or any that I've forgotten that support PCI-e 2.0) there probably won't be any limitations with a 4870X2. Of course noting how popular and extremely common chipsets such as the P35 are...I'm looking forward to seeing some tests even though I have no plans on buying ...I just think its stuff like this that makes me love the move we've made to PCI-e more and more.

Yes PCIE 2.0 doubles bandwidth but based on the Tweaktown review, running CF with two x8 PCIE 2.0 lanes starts showing degraded performance on 4850CF. Considering the 4870s would require more bandwidth that might be more pronounced. The linked Tom's comparison does show PCIE 1.0a x8 (PCIE 2.0 x4) decreased performance but little/no change with PCIE 1.0a x16 (PCIE 2.0 x8) and PCIE 2.0 x16.

 

[Aberforth]

Radeon 4870 X2(R700) will priced @ $499 and will be released in 4th week of July

Source: http://www.techpowerup.com/index.php?64378

 

[AmberClad]

Is that right? I would have expected MSRP $599, at least to be in line with the 4870 price.

 

[Aberforth]

Originally posted by: AmberClad
Is that right? I would have expected MSRP $599, at least to be in line with the 4870 price.

It is definitely right, Radeon will never be priced at GTX 280 range.

 

[nitromullet]

Originally posted by: AmberClad
The power connectors - they seem like they're facing downward? I wonder if they're keeping that on the final card.

The 3870 X2 was exactly like that. It would be nice if they would at least move the middle connector to the end.

Originally posted by: Aberforth
Radeon 4870 X2(R700) will priced @ $499 and will be released in 4th week of July

Source: http://www.techpowerup.com/index.php?64378

Provided it offers the performance we think it will at that price, I'd deal with the power connectors not being in the perfect spot.

 

[Aberforth]

Originally posted by: nitromullet

Originally posted by: AmberClad
The power connectors - they seem like they're facing downward? I wonder if they're keeping that on the final card.

The 3870 X2 was exactly like that. It would be nice if they would at least move the middle connector to the end.

Originally posted by: Aberforth
Radeon 4870 X2(R700) will priced @ $499 and will be released in 4th week of July

Source: http://www.techpowerup.com/index.php?64378

Provided it offers the performance we think it will at that price, I'd deal with the power connectors not being in the perfect spot.

With a single 4870 performing very close to GTX 280 it would be logical to assume R700 is much faster than GTX 280 and will probably be close to 280 SLI.

 

[uclaLabrat]

Originally posted by: Aberforth

Originally posted by: nitromullet

Originally posted by: AmberClad
The power connectors - they seem like they're facing downward? I wonder if they're keeping that on the final card.

The 3870 X2 was exactly like that. It would be nice if they would at least move the middle connector to the end.

Originally posted by: Aberforth
Radeon 4870 X2(R700) will priced @ $499 and will be released in 4th week of July

Source: http://www.techpowerup.com/index.php?64378

Provided it offers the performance we think it will at that price, I'd deal with the power connectors not being in the perfect spot.

With a single 4870 performing very close to GTX 280 it would be logical to assume R700 is much faster than GTX 280 and will probably be close to 280 SLI.

IF that's true, and the rumors of the X2 being 15% faster than CF'd 4870's, it may be a bigger coup than the 4850. $499 and beats $1300 of SLI'd GTX's? can't wait!

 

[Tempered81]

Pretty sure that...

2x GTX 280 SLI > R700 by 20% average
R700 > GTX 280 by 15% avg in X-fire scaling games
2x R700 > 2x GTX 280, 3x GTX 280, & 2x 9800GX2 by who knows how much after driver improvements in 4-gpu scaling...

 

[Tempered81]

Pretty sure that...

2x GTX 280 SLI > R700 by 20% average
R700 > GTX 280 by 15% avg in X-fire scaling games
2x R700 > 2x GTX 280, 3x GTX 280, & 2x 9800GX2 by who knows how much after driver improvements in 4-gpu scaling...

 

[bryanW1995]

Originally posted by: jaredpace
Pretty sure that...

2x GTX 280 SLI > R700 by 20% average
R700 > GTX 280 by 15% avg in X-fire scaling games
2x R700 > 2x GTX 280, 3x GTX 280, & 2x 9800GX2 by who knows how much after driver improvements in 4-gpu scaling...

so you really think that gtx 280 sli will be only ~ 40%? also, you're figuring even less than that for % increase over 4870 for 4870x2. that doesn't make any sense

 

[deepinya]

so ATI > Nvidia.....finally?

Good for them