[News] AMD Giving Up on CrossFire with RX Vega

[StyM]

https://www.techpowerup.com/235699/a...e-with-rx-vega

AMD is reportedly scaling down efforts on its end to support and promote multi-GPU technologies such as CrossFire, and not in favor of open-standards such as DirectX 12 native multi-GPU, either. Speaking to GamersNexus, an AMD representative confirmed that while the new Radeon RX Vega family of graphics cards support CrossFire, the company may not allocate as many resources as it used to with older GPU launches, in promoting or supporting it.

This is keeping up with trends in the industry moving away from multi-GPU configurations, and aligns with NVIDIA's decision to dial-down investment in SLI. This also more or less confirms that AMD won't build a Radeon RX series consumer graphics product based on two "Vega 10" ASICs. At best, one can expect dual-GPU cards for the professional or GPU-compute markets, such as the Radeon Pro or Radeon Instinct brands.

[StAndrew]

I'm curious why this is. I've been a huge fan of SLI; whats the drive diverge from multi GPU's?

I figured NVIDIA did it with Pascal for a few reasons: a) Pascal doesn't need it with current games and b) AMD wasn't competing well and c) SLI tended to cut into flagship sales as two mainstream cards tend to be a better option than one flagship. Building SLI profiles tend to take a lot of work and man hours; why not refocus those resources towards AI and other emerging tech trends?

But AMD, not competing with NVIDIA's top two (three) cards, might want to reconsider??? Although with +300w per GPU, maybe crossfire isn't the best idea for the end user

[chew*]

Its rather simple. They put a bunch of time into driver coding. Consumer buys 2 cards thats total costs less than flagship. Gets equal or slightly better performance.

Basically nvidia and AMD want to sell flagships.

Im fairly certain the game designers also play a "lazy" role in this.

Perfect example of this is i bought 2 used 290x for much less than a single fury x. I have compared in almost everything. Except some "batman" title the fury x got smoked...

[StAndrew]

Its rather simple. They put a bunch of time into driver coding. Consumer buys 2 cards thats total costs less than flagship. Gets equal or slightly better performance.

Basically nvidia and AMD want to sell flagships.

Im fairly certain the game designers also play a "lazy" role in this.

Perfect example of this is i bought 2 used 290x for much less than a single fury x. I have compared in almost everything. Except some "batman" title the fury x got smoked...

I absolutely understand why Nvidia limited SLI support. When you completely dominate your competitor it makes sense (just ask Intel).

Why AMD is dropping functionality when they have a chance to one up Nvidia is a mystery to me. Especially when you consider Vega's poor (in comparison) performance, if they can launch a cheap card that, when in crossfire, can match or even beat a more expensive TI, AMD may have a winner?

Or is there a trend away from multi GPU's I'm not seeing? Is game coding becoming more complex and harder to code drivers for? Otherwise I just don't understand why and underdog would not invest into a potentially advantage.

With the release of Threadripper and 3-4 way GPU computing with PCI-E lanes to spare, 3-way or 4 way vega crushing Nvidia at 4k or even 8k (probably not) might be a selling point. Assuming they can manage power and heat from multiple 300-400W GPU's...

[JAWS]

I don't want to sound like a broken record, but I would never spend more than $300 on a gpu. If they are doing this because they want to sell $500-$1000 gpu's....good luck.

There's a reason why I don't buy Nvidia. too expensive. Why would AMD start down that path? I've always been an AMD supporter, but some of these prices on their new products have me shaking my head. (cpu and gpu)

sorry maybe not the thread for this.

[Charles Wirth]

WB Justin, good to see you dropped in to comment.

[zanzabar]

there was the break down on the steam hardware survey a couple years ago with what cards were used in multi gpu and it was almost all high end cards. amd and nvidia also both said that they did not recommend or see people using two cards as an upgrade path for mid/cards, but who knows if that is true or they just want to annualiz mid range.

I don't want to sound like a broken record, but I would never spend more than $300 on a gpu. If they are doing this because they want to sell $500-$1000 gpu's....good luck.

There's a reason why I don't buy Nvidia. too expensive. Why would AMD start down that path? I've always been an AMD supporter, but some of these prices on their new products have me shaking my head. (cpu and gpu)

sorry maybe not the thread for this.

the new rx vega has around $200 in HBM2 and likely has more than $50 in lithography per part. sure NV is charging $400+ for mid range gp104 parts like the 1070/1080 but they also are putting proper vrm, coolers, and memory in them too so they do clock better now compared to the 100/102 unlike their older 104 parts.

[iddqd]

It wouldn't surprise me if Crossfire support is there - but broken. There are so many bugs in this ASIC, absolutely nothing works right.

[nunchucka]

or NAVI is going to be multi cores with infinity fabric on die so why bother with crossfire anymore?

[iddqd]

Who started this rumor? There's 0 evidence that Navi is going to be anything other than a monolithic design. Vega also uses infinity fabric, by the way.

[StAndrew]

Who started this rumor? There's 0 evidence that Navi is going to be anything other than a monolithic design. Vega also uses infinity fabric, by the way.

Nvidia.

https://www.extremetech.com/gaming/2...future-designs

[AliG]

If you can get the interconnect properly working, there's a lot of benefits to doing an MCM. Back when Intel launched their first dual cores, they tremendously improved their yields by deciding to glue together 2 single core chips instead of a monolithic die. AMD is now doing the same thing with Ryzen, where they designed 1 8 core CPU (which is really 2 quadcores internally connected) and has used that to fit basically their entire market segment. That allows them to consolidate R&D and greatly reduce costs.

This isn't to say that monolithic designs don't tend to be more efficient. There's less latency between cores and less risk of mishandled data packets. But if we're talking a performance decrease of a couple percent for a 20% reduction in MSRP, I'm all for it.

[screwtech02]

Well this just sucks. Been a crossfire user since X1900xt days when Nvidia wouldnt allow SLI on 975 chipset.... Seems "both" teams are now going to "force" the buying populace to purchase what "the man" thinks they "need"........

[nunchucka]

Who started this rumor? There's 0 evidence that Navi is going to be anything other than a monolithic design. Vega also uses infinity fabric, by the way.

from Raja Koduri :

"Infinity Fabric allows us to join different engines together on a die much easier than before. As well it enables some really low latency and high-bandwidth interconnects. This is important to tie together our different IPs (and partner IPs) together efficiently and quickly.

It forms the basis of all of our future ASIC designs.

We haven't mentioned any multi-GPU designs on a single ASIC like Epyc, but the capability is possible with Infinity Fabric."

so it is "possible" that is why they are stopping xfire.