We now have details on what AMD will launch to compete against Nvidia, from an unusual source. Newegg publishes its own graphics and computing guides, including a guide to which GPU you should buy. Included in that guide, at least briefly? Specs for AMD’s upcoming Radeon RX 6700 XT, 6800 XT, and 6900 XT GPUs.
Obviously these specs should be considered preliminary, but they line up with other rumors we’ve heard about the family. Hat-tip to Videocardz for finding the information:
There are some oddities in the chart that are difficult to explain. The Radeon 6700 XT is a marked downgrade from the 5700 XT, with less clock, less RAM, less memory bandwidth, and — admittedly — a very nice TDP of 150W. We can imagine that some of RDNA2’s performance-per-watt uplift will work in AMD’s favor here, but the smaller RAM buffer is a little surprising. AMD may have offset the clock drop and memory bandwidth reduction with more cache and higher efficiency, but there’s no way to offset reduced VRAM.
It’s possible that the 6700 XT will be offered in a 6GB and 12GB flavor, similar to GPUs like the RX 570. Another leak — this one from Reddit — gives some better clock data courtesy of Apple’s macOS 11 beta.
We don’t know if these numbers are accurate either, but they imply a Navi 21a and 21b. 21a has a maximum frequency of 2050MHz, while 22b has a top frequency of 2200MHz. Socket power limits are given as 200W and 238W respectively. This implies it takes 1.19x more power to drive 1.07x more performance on the chip. The fact that the Apple chart refers to socket power could mean this is only the power draw for the chip itself, not the rest of the card. If we assume that, the TDP listings from Newegg and Apple are a bit closer to each other.
Several weeks ago, I wrote that if AMD could increase its core clocks up to 2.2GHz (equivalent to the PS5), it could help the company square off with the RTX 3080. The GPU could pick up 1.22x performance from the additional clock (assuming a 1.85GHz – 2.2GHz jump). We know performance should nearly double from the core count increase, though the jump won’t be perfectly linear. Given that the RTX 3080 is about 2x faster than the 5700 XT in multiple tests, AMD may be able to land a competitive GPU for Ampere in terms of raw performance, though obviously power consumption and thermals will be their own topics. It’s too early to know if the early problems we’re hearing about with GeForce RTX 3080 shipments will affect comparisons when RDNA2 launches or not.
We know that Nvidia is claiming a 1.9x performance/watt improvement over Turing, compared to AMD’s 1.5x improved performance per watt over RDNA. It’s going to be very interesting to see how the two companies compare against each other when cards drop in the not-so-distant future.
- Ampere Unleashed: Nvidia’s RTX 3080 Redefines High-End Gaming
- The Xbox Series X Will Crush an Equivalently Priced Gaming PC
- AMD RDNA2 Launches Oct 28, But How Will It Respond to Nvidia’s Ampere?