Jump to content

AMD Might Only Make 5000 Radeon VII GPUs and No Custom AIB Versions


Guest_Jim_*

Recommended Posts

Among the announcements and reveals from AMD last week at CES 2019 was the Radeon VII, the first consumer-targeting GPU made on a 7 nm manufacturing node. While some may argue it is more prosumer targeting, being repurposed from the MI50 accelerator with 16 GB of HBM2, it is still being described as a graphics card capable of gaming at 4K, but if you the $699 price and the desire for one, you might need to act quickly come February 7, when it launches. Tweak Town is reporting its industry contacts say there will be "less than 5000" made. Not only that, but supposedly AIB partners will not be getting the chips or boards necessary for making versions with custom cooling solutions, and these cards will be sold at a loss.

Technically these are both rumors, but both claims are likely disappointing for some, especially the 5000 number. If that is the case, then we likely will not see the prices drop over time, as the supply will be exhausted and not restored. The lack of AIB custom cooling will also be disappointing because while the new three axial-fan design looks better than the reference designs AMD has used previously, some would expect custom cooling to do a better job still. Time will tell and come February 7 we will learn just how powerful this GPU is for gaming.

Source: Tweak Town



Back to original news post

Share this post


Link to post
Share on other sites

I certainly hope they are Rumors and stay that way. I have been wanting for AMD to come back out of the woods and start competing with Nvidias high end line. Hopefully they have a very successful launch and this is a precursor for things to come. 

Share this post


Link to post
Share on other sites

+1 on successful Vega II launch! 

 

 

They disabled double-precision floating-point on Radeon VII gimping it's FP64 performance, so the VII doesn't negatively impact Radeon Instinct MI50 sales.

 

Despite the gimping Radeon VII should be capable of pumping out 862 GFLOPs of FP64 performance and stay ahead of the Nvidia Titan RTX (509 GFLOPs) and GeForce RTX 2080 Ti (420 GFLOPs). They also disabled PCIe 4.0 on Radeon  VII. The reasoning is that AI workloads can saturate the PCIe 3.0 interface with relative ease, and limiting the available bandwidth would dissuade data center operators from repurposing cheaper consumer gear for those types of workloads. So AMD has left the PCIe 4.0 feature active only for the Radeon Instinct MI50 and MI60.  

Edited by Braegnok

Share this post


Link to post
Share on other sites

I certainly hope they are Rumors and stay that way. I have been wanting for AMD to come back out of the woods and start competing with Nvidias high end line. Hopefully they have a very successful launch and this is a precursor for things to come. 

I have also seen some other rumors that AMD is planning to produce more like 60,000 (20,000 in a first run, 40,000 in a second), so the speculation has become the 5000 number is how many will be available at launch, with more coming after.

Personally, this makes far more sense to me than 5000 total. A number of rumors have stated Navi will first release in the mid-range, which would mean a high-end GPU would be needed, and this might be it, at least until top-tier Navi is ready to release. While it is the mid-range both companies make the most money at, as these products sell the most, having powerful top-tier parts helps sway the mid-range purchases.

I hope Radeon VII has a good launch too, but even if I had the money, I don't think I'd get it. Thoroughly more powerful than I need, unless I also upgrade my monitor, but I'm happy with what I have. Right now, I'd just be running it with Chill enabled all the time and settings turned to max, which is what I'm doing with my Vega 64. Still would be fun, but really a waste on me. Arguably a Vega 56 would be more optimal for my use case, but nothing wrong with going a step above to be ready for more. Plus I got my Vega 64 release week and still think I was lucky to, so waiting to try again with the Vega 56 was not what I wanted to do.

Share this post


Link to post
Share on other sites

Just put up this item: AMD Statement Pushes Back on 5000 Radeon VII Rumor

 

ExtremeTech got a statement from AMD stating it believes it will have the supply to satisfy the demand of gamers. No production numbers were given though, but since this would just be another bin for the Vega20 dies to be sorted into, so long as that total supply is enough, they can probably continue making them.

In theory, the Radeon VII could be sold at a loss so long as enough MI50 and MI60 units, also using Vega20, are sold to make the whole Vega20 die a success. If the dies going into Radeon VII could not be used for MI50/60, then this is a way to make some money, as opposed to destroying them. Then there are the marketing benefits from the Radeon VII disrupting NVIDIA's top-tier dominance (in theory/pending independent reviews) that give it more than just straight monetary value.

Share this post


Link to post
Share on other sites

  • 2 weeks later...

+1 on successful Vega II launch! 

 

 

They disabled double-precision floating-point on Radeon VII gimping it's FP64 performance, so the VII doesn't negatively impact Radeon Instinct MI50 sales.

 

Despite the gimping Radeon VII should be capable of pumping out 862 GFLOPs of FP64 performance and stay ahead of the Nvidia Titan RTX (509 GFLOPs) and GeForce RTX 2080 Ti (420 GFLOPs). They also disabled PCIe 4.0 on Radeon  VII. The reasoning is that AI workloads can saturate the PCIe 3.0 interface with relative ease, and limiting the available bandwidth would dissuade data center operators from repurposing cheaper consumer gear for those types of workloads. So AMD has left the PCIe 4.0 feature active only for the Radeon Instinct MI50 and MI60.  

i got this email earlier today... very interesting !

 

Share this post


Link to post
Share on other sites

It will definitely be interesting to see how the Radeon VII develops in the coming weeks or months or so. Will there be AIB variants (apparently they are free to make them but will depend on supply and demand) and what will be possible once some of the overclocking features are stabilized? I've seen reports that even WattMan is having problems, but at least one place still managed to get an undervolt and, like so many AMD GPUs, there is plenty of room to decrease voltage, improving efficiency. Also seeing what will be possible with overclocking will be cool too. It wasn't easy but at least der8auer saw his reach 2100 MHz, but the only way he found to do it was to cool it differently (he uses dry ice but only cooled it to what he thinks a water cooler could achieve) and use the auto-overclock feature in the drivers; he couldn't directly overclock it.

 

 

Share this post


Link to post
Share on other sites

×
×
  • Create New...