R700 pics

[STEvil]

cadaveca - have you tried adjusting the PCI-E frequency? Maybe the swap file is playing a part somehow?

[metro.cl]

A TDP of 160 is likely the total board power not the die itself. The X2 should consume a decent amount less than 2 4870 512s. My guess would be in the 250-280 range.

The HD 3870 had a TDP of 105, then the HD 3870X2 had a TDP of 190, so you get the 20Watts i'm talking about. If you check NVIDIAs 8800GT to 9800GX2 you see the same range of optimization.

then for HD 4870 you have 300Watts or so, there is no way they can lower 70 Watts to get it to 250Watts that is already a lot of heat to disipate.

[cadaveca]

LoL...Stevil...you got my old dualie out of my BP6...you know I've been doing this a while...So yes, PCI-e Frequency has been changed, with no noticuble difference.


Also note, pls, that as I have mentioned before...that video is with in-game res set to 1280x1024...but 1680x1050 it's barely noticible...so your comment aobut swap file makes a wee bit of sense, however, it's the gmae itself, or the driver interaction, that makes the problem as evident as it is.

I mean really...I don't play @ 1280x1024...the video is only an example of the behavior.


GRID..in corners...when the actual rendering is far less than on a straight(due to draw distance)...it happens on the straights too, but nowhere near as pronouced...


It really makes me think of how the game is being rendered that is the issue...framerate will be 100+...appears like 20FPS...seems like as data flow to the card increases, stutter rears it's ugly head...like the data flowing in the card itself is stalling...maybe it's the data on the pci-e bus...I dunno.



ANyway, I'll end this here for now...maybe I'll start a thread on the issue at a later time, but I'll leave off with this:

The problem IS going away as drivers are updated. WHy, or how, I dunno...but it truly seems OEMs are aware of the issue, and are working to rectify it. We can hope that R700's multi-gpu communication implementation will overcome this problem, or be a fix, but until the card comes out, we won't know.

I also have a feeling that the driver for R700 is what will hold the card back. I do not think AMD wants another relase fiasco like UVD on thier hands, so I remain confident soemthing will happen.

[Crankybugga]

Just because you don't see or feel the microstuttering (or whatever you want to call it), does NOT mean it doesn't exist and that it doesn't bother others. When we are spending hundreds of dollars on these GPUs, we have every right to want them to work how we see fit.

I got no problem with that at all
Its all these experts who dont own and havent seen that fragment this thread.
Surely theres other thread to discuss this problem without having nvidia fanbois pharking up intentionly this one

[Crankybugga]

We can hope that R700's multi-gpu communication implementation will overcome this problem, or be a fix, but until the card comes out, we won't know.

I also have a feeling that the driver for R700 is what will hold the card back. I do not think AMD wants another relase fiasco like UVD on thier hands, so I remain confident soemthing will happen.

I think they gained a lot of experience with the predecessor although they have made huge changes, hopefully with experience gained
Drivers are now ATi's cup of tea, no hope needed.
But i could see a bios flash to improve stuttering problem in current X2

[STEvil]

Yeah, data flow was my thought too

maybe EMI is causing corrupted data flow between the cards?

[Crankybugga]

Not my area, but i wouldnt think so. Seems like bad latency between cores data but i doubt drivers can adress that.

[Helmore]

The HD 3870 had a TDP of 105, then the HD 3870X2 had a TDP of 190, so you get the 20Watts i'm talking about. If you check NVIDIAs 8800GT to 9800GX2 you see the same range of optimization.

then for HD 4870 you have 300Watts or so, there is no way they can lower 70 Watts to get it to 250Watts that is already a lot of heat to disipate.

Rumors as of now suggest that the R700 will use 2 RV770 cores clocked at 4850 speeds but with GDDR5 RAM, so this could mean that power consumption will be closer to 2 times a 4850.

[Dimitriman]

4870X2 = DDR3 = Less power consumed than 2x 4870.

[zerazax]

GDDR5 consumes less power than GDDR3 so why would they step back and use GDDR3? Besides, GDDR5 provides more bandwidth

[naokaji]

4870X2 = DDR3 = Less power consumed than 2x 4870.

that would make it more of a 4850x2 though if they would opt for gddr3, but why would they do that? the 4870x2 will be the top product, so the price of gddr5 cant really be the issue (and it doesnt use more power and is faster).

I think they will just bin the 4870 chips and run them at lower volts to keep tdp somewhat under control.

[Hornet331]

that would make it more of a 4850x2 though if they would opt for gddr3, but why would they do that? the 4870x2 will be the top product, so the price of gddr5 cant really be the issue (and it doesnt use more power and is faster).

I think they will just bin the 4870 chips and run them at lower volts to keep tdp somewhat under control.

i dont think its the price, more a sort a problem of availability.

[Helmore]

that would make it more of a 4850x2 though if they would opt for gddr3, but why would they do that? the 4870x2 will be the top product, so the price of gddr5 cant really be the issue (and it doesnt use more power and is faster).

I think they will just bin the 4870 chips and run them at lower volts to keep tdp somewhat under control.

AMD hasn't decided on a name for their R700 part, we are just using the 4870X2 moniker because that's what we got used to with the HD3xxx series. They might even name an R700 card the ATI Radeon HD49x0 for all we know.

[zerazax]

I've heard even 4890

The 4870X2 SKU is just made up by us since R680 used it but a different name might be telling as to how much it is actually like the X2

[JohnZS]

JohnZS, stay on topic or stay out

Now where are the die shots of the R700

Perkam

Sorry for thread jacking.

Does anyone know if there would be 2GB versions of the R700?
Also (and although this might be slightly off topic again (sorry)) did anyone see the article where 4xHD4850's in CrossfireX mode were slower than 3 and in some cases slower than 2 HD4850's here?

Perhaps the R700 is all ready and tapped out somewhere but is not being released due to the driver issues with 4GPU's?

John

[zerazax]

I've heard late July availability so its likely taped out already. Driver patching up is likely a big part of the issue as well as production of cores + memory

GDDR5 limited quantities won't be too big of a problem if all the indications of the 4870 already being on sale and being launched June 25th are true