AMR Up and Running at Cebit

[perkam]

This thread corresponds to the recent news that AMR (or whatever its gonna be called - the inq has stopped calling it that though ;) ) as up and running at Cebit and you can use two different cards like an x800 and an x850, and higher clock card will automatically downclock to lower card speeds and they'll run in multi-gpu mode. Apparently Inq was quite impressed at how far ATI has gotten with this which can only be good news as SLI Revolution is well under way.

Hopefully we can find the mobos in quantities unlike their cards :rolleyes: No word yet on whether R520 will be compatabile with multi-vpu, but if x800/x850 will be and there shouldnt be a reason why not.

Also it seems that there might be a refresh of the x700 series (THANK GOD !!! RIP X700XT :) ) in Q3 2005 using R520 architecture and an R515 for lower end as well.

Perkam

[ZeroHero]

Sounds good.

[Geforce4ti4200]

wow a refresh of x700 with r520 core? mmmmmm ps3 goodness, 12 pipe, fast ram goodness all for $199 :slobber: as for amr, I thought the x800s didnt support it? doesnt the chip have to physically support amr? I dont see a slot on top to use a passthru cable to connect the two cards

[Mathias-K]

Some time ago, it was rumoured that ATi cards would run their multi vpu at driver level, and thus would require no connector of any kind. Could be bull, though :)

[reject]

nah thats true, at least a true rumour
its aweosme that you can use any cards, u can get an x800xl for gaming, then when u wanna bench chuck in a x700. would the faster card only use the same number of pipes as the slower card? so a x800pro and x800xl would run 12x1 and 12x1?

[DudeMiester]

When they say they run it at driver level, I'm 95% sure they mean they just use the PCIe bus to transfer the data between the cards that the bridge transfers in Nvidia's SLI. It just means less bandwidth free over the PCIe bus, but given that there is so much bandwidth there I doubt it will be an issue with current games, but in the next-gen there may be a small difference.

[googles]

...you can use two different cards like an x800 and an x850, and higher clock card will automatically downclock to lower card speeds and they'll run in multi-gpu mode...

have a question, what happens if i have a x800 that ocs to x850 speeds or more, is it possible to run amr at higher speeds if both cards can operate at these speeds?

[sabrewulf165]

wow a refresh of x700 with r520 core? mmmmmm ps3 goodness, 12 pipe, fast ram goodness all for $199 :slobber: as for amr, I thought the x800s didnt support it? doesnt the chip have to physically support amr? I dont see a slot on top to use a passthru cable to connect the two cards
ATI VPUs have supported multi-VPU configurations for several generations now if I'm not mistaken

[gokickrocks]

This thread corresponds to the recent news that AMR (or whatever its gonna be called - the inq has stopped calling it that though ;) ) as up and running at Cebit and you can use two different cards like an x800 and an x850, and higher clock card will automatically downclock to lower card speeds and they'll run in multi-gpu mode.

most likely the higher end card wont be downclocked, it will just downgrade to the other card's capabilities, ie: a ps3.0 card will use ps2.0

[Kanavit]

where can i find more information regarding AMR and Cebit? Can someone post me a link pls?

[b|gf|sh]

From X-Bit - Day2 CeBIT Coverage (http://www.xbitlabs.com/articles/editorial/display/cebit2005-2.html): -

http://www.xbitlabs.com/images/editorial/cebit2005-2/7.jpg

:cool:

[RAndomaN]

interesting none the less... would like to see some benchmarks :)

[althes]

Nvidia looks in trouble

[perkam]

Apparently its not good to create multiple threads on the same topic so I'll keep this update brief.

A name has been decided upon for AMR multi gpu....so officially its going to be:

SLI vs MVP !!!

http://img.hexus.net/v2/internationalevents/cebit_hannover_2005/store/mvp_board_big.jpg

AND NOW WE KNOW HOW IT WORKS !!!


MVP supports split frame rendering using supertiling, where the screen is split up into tiled areas with each tile processed on a GPU, using any GPUs that support supertiling. That's anything from R300 up, but it's likely to be limited to R4xx GPUs. You can use X700 and X800, X800 XL and X850 XT PE, or any other mix that you can think of. There's the potential to increase anti-aliasing IQ using supertiling (multipassing the tiles through a GPU) and MVP.

More info is available here: http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD0xMDQx

Perkam

[Bloody_Sorcerer]

sounds fairly interesting... wonder what happens when you throw 2 NV cards in, or an ATi and a NV.

[perkam]

sounds fairly interesting... wonder what happens when you throw 2 NV cards in, or an ATi and a NV.

Patent Infringement LOL

Perkam

[IvanAndreevich]

Bloody_Sorcerer
Tomshardware have done this. Apparently there was a lot of driver issues.

[stealthbomber]

Sounds impressive, looks like ATI has managed to pull white bunnies out of their empty hats! :idea:

[Geforce4ti4200]

Nvidia better do something quick, if ati can pull off amr with their r520 then they will just crush nvidia. One r520 is speculated to beat 2x 6800u, imagine two especially if it has no cpu overhead like SLI which eats 10% cpu power

[EMC2]

Someone (GPU or CPU) has to supervise and coordinate between the cards, so there will be some degree of overhead. The Q is how much. As far as NVidia doing something... how does 4 GPU's sound? :lol:

*edit... darn trailing semicolons*

[charlie]

Nvidia better do something quick, if ati can pull off amr with their r520 then they will just crush nvidia. One r520 is speculated to beat 2x 6800u, imagine two especially if it has no cpu overhead like SLI which eats 10% cpu power

Yeah, those ineficient 6800U's :rolleyes: with all that CPU draw they barely manage to pull of 24K in '03 and 11K in '05 :stick:

C

[isp]

I thought the CPU draw was closer to 1-2%, is it really 10?

[Geforce4ti4200]

Yeah, those ineficient 6800U's :rolleyes: with all that CPU draw they barely manage to pull of 24K in '03 and 11K in '05 :stick:

C



They are effeciant in gpu based applications such as 3dmark 2003 and 2005, but lets mention 2001, the top sli score is 40k and thats with 490 nature. All the other tests are considerabily lower than what a single x800xt gets. In fact a single 6800u will score higher in the cpu test in 2001 than two 6800s, your car and lobbies will drop when you use sli. The hit in 2003 and 2005 are far less cause those benchies are very gpu intensive but you still dont get doubled score

[EMC2]

Regarding the overhead issue and the ATI "solution"...

This in from X-Bit Labs (http://www.xbitlabs.com/articles/editorial/display/cebit2005-2.html)



ATI does not reveal the Multi-VPU concepts to the general public, but for its add-in card partners it has posted a brief overview of its technology in its roadmap update. At least for now there is one important update from what was previously reported – ATI wants two identical graphics cards to be installed, not two different boards as said before based on information from unofficial sources. Here is what ATI says to its partners:

* Multiple ATI RADEON X800 XT boards cooperatively rendering a single frame;
* Requires two physical x16 connectors on the mainboard;
* Load balancing and synchronization implemented entirely in software;
* No physical connector requires between devices;
* Currently assumes two identical graphics devices installed in both connectors;
* Offers several user selectable modes of multi-processing;
* Works with any PCIe north bridge.


Notice item 3 in the list... also think about the possibly not-so-great implications of item 4.

[EMC2]

--- Geforce ---

It's called "not it's intended use". The entire plus of SLI is to speed up graphics, not "cpu test in 2001" :p Look at how it performs where it's intended to be used, very intense graphics.

And of course you don't get double... there is always some amount of "overhead" involved, whether it be in SW, HW, or both. Maybe a lot more in the case of ATI's "solution" based on the tidbit from X-Bit Labs I posted above.


--- Charlie ---

Hey man, if you know anyone wanting to throw away their "inefficient" 6800u's, send em my way ;) I'll bury them for 'em :lol:

[Marquzz]

Ohh, I'm so glad I didn't jump the nVidia train! One ATI mb and another X800XT for me please :)

[Der_KHAN]

* Load balancing and synchronization implemented entirely in software;
* No physical connector required between devices;sounds like a buttload of cpu-load...

[mcnbns]

Can someone explain to me why not having a physical connector is bad? I'm not making the connection... :slap:

[HaLDoL]

Ohh, I'm so glad I didn't jump the nVidia train! One ATI mb and another X800XT for me please :)
Without nvidia you wouldn't have that 2k1 score. Seems you're on the nvidia train after all. I know you mean the graphic cards but still, you should be thankfull to nvidia for bringing such good motherboard chipsets. Still have to see that from ATi.

[Geforce4ti4200]

sounds like a buttload of cpu-load...



If this is the case, then blah on amr too! I hope it has equal or less cpu load than nvidia's sli!

[perkam]

then blah on amr too!

Its called MVP now ;) Though my interest is in the fact that you can use mvp with different ATI PCI-e cards :O

Perkam

[SAE]

Hey guys... :D

saaya and me, we've met ATI and the hexus guys at the Cebit.
Hexus seems to have gotten way more info than saaya got (tbh ATI was not telling anything NEW to him iirc).

More about that and all his news and lotsa pics coming tomorrow the earliest. I don't wanna tell ya more than that as it's saaya's task to do that ;) I am under NDA :hehe:

[Geforce4ti4200]

wont you get less cpu overhead and better performance with one x800xl instead of say two x700pros?

[DevilsRejection]

Hey guys... :D

saaya and me, we've met ATI and the hexus guys at the Cebit.
Hexus seems to have gotten way more info than saaya got (tbh ATI was not telling anything NEW to him iirc).

More about that and all his news and lotsa pics coming tomorrow the earliest. I don't wanna tell ya more than that as it's saaya's task to do that ;) I am under NDA :hehe:

When does your NDA expire? Usually a few hours/days after an NDA expires info is announced to the public.

[EMC2]

Can someone explain to me why not having a physical connector is bad? I'm not making the connection... :slap:

The two GPUs have to communicate and transfer data between each other. Let's say that each is rendering exactly 1/2 of each frame. That means the resulting 1/2 frame has to be transfered from the 2nd GPU to the 1st (that's almost 500MB/sec for 1600x1200, 32-bit, 60fps). Without a physical connection between the two seperate from the PCIe bus, that means that the data has to be transferred across the PCIe bus. This is going to take some % of your available bandwidth (given, PCIe has a boatload of BW, but 500MB/sec is still a lot, about ). It basically ties down 3 lanes of the PCIe bus (out of the 8 each card has in a dual configuration). Plus, since it has to go through the PCIe switch which resides in the MB chipset, it is also robbing bandwidth for other devices and for access to/from memory (for PCIe and PCI devices).

Crystal clear now, right ;)

[sandman]

Wait wait, so you can use two X800XL's then?

When do these become available? I must have one.

[Kanavit]

I love this competition, without both companies, we will still be using intel graphics.