Professional RDNA 3 For The Mid-Vary

As AMD continues to launch their full graphics product stacks primarily based on their newest RDNA 3 structure GPUs, the corporate is now getting ready their subsequent wave {of professional} playing cards beneath the Radeon Professional lineup. Following the launch of their high-end Radeon Pro W7900 and W7800 graphics cards again within the second quarter of this 12 months, immediately the corporate is asserting the low-to-mid-range members of the Radeon Professional W7000 sequence: the Radeon Professional W7500 and Radeon Professional W7600. Each primarily based on AMD’s monolithic Navi 33 silicon, the newest Radeon Professional elements will hit the cabinets a bit later this quarter.

The 2 playing cards, as a complete, will make up what AMD defines because the mid-range phase of their skilled video card market. And like their flagship counterparts, AMD is relying on a mix of RDNA 3’s superior options, together with AV1 encoding help, improved compute and ray tracing throughput, and DisplayPort 2.1 outputs to assist drive gross sales of the brand new video playing cards. That, and as is custom, considerably undercutting NVIDIA’s competing skilled playing cards.

Not in contrast to their high-end counterparts, for this technology AMD has determined to broaden the dimensions of their mid-range professional graphics lineup. Whereas the earlier technology had the only W6600 (and W6400 at entry-level), the W7000 sequence will get each a W7600 card and a W7500 card. Moreover the apparent efficiency distinction, the opposite huge characteristic separating the 2 playing cards is energy consumption. The Radeon Professional W7600 is a full-height video card operating at 130W, whereas the W7500 is explicitly designed as a sub-75W card that may be powered solely by a PCIe slot, coming in at a cool 70 Watts.

Each playing cards are primarily based on AMD’s new Navi 33 GPU, and share quite a lot of similarities in options and bodily card design consequently. Navi 33 just lately launched within the client house as the guts of the Radeon RX 76000, and it brings with it all the performance and idiosyncrasies that we noticed in these playing cards. Whereas not supposed to be a real finances GPU, Navi 33 is the lowest-end chip AMD is making for this technology of processors. Constructed on TSMC’s 6nm course of (in contrast to 5nm used for Navi 31), it’s not precisely cutting-edge when it comes to efficiency of energy effectivity. However it’s low-cost for AMD to make, which helps to maintain prices down – an important aspect when a giant a part of AMD’s professional graphics technique is undercutting NVIDIA.

Radeon Professional W7600 will get a full-fat Navi 33 chip, with all 32 CUs enabled. Typical for Professional elements, that is clocked a bit extra conservatively at 2.43GHz, rendering a complete FP32 compute throughput of 19.9TFLOPS. The cardboard is paired with 8GB of GDDR6 operating at 18Gbps, netting a complete reminiscence bandwidth of 288GB/sec. And, after all, AMD’s on-die Infinity Cache is obtainable as effectively, including 32MB of high-speed cache to the reminiscence hierarchy.

Like-for-like comparisons to its predecessor, the Radeon Professional W6600, are a bit tough as a result of vital architectural modifications in RDNA 3. However at a excessive degree, it capabilities lots like a W6600 with some new methods, proper on all the way down to the an identical TDPs. The RDNA 3 CU overhaul signifies that W7600 has, on paper, much more compute throughput than its predecessor, as long as these further FLOPS may be tapped. However on paper, rendered pixel throughput is definitely down gen-on-gen as a result of decrease increase clockspeed. Alternatively, there’s much more bandwidth than earlier than to feed the overall GPU, due to the change up from 14Gbps GDDR6 to 18Gbps GDDR6.

In the meantime, the W7500 trades in a good bit of GPU efficiency with a purpose to get its TDP down beneath 75W – permitting AMD to launch a brand new sub-75W card for this technology of merchandise. A 28CU configuration of Navi 33 is clocked at simply 1.7GHz, netting 12.2 TFLOPS of compute efficiency. And the cardboard’s reminiscence is equally low energy, benefiting from low voltage GDDR6 operating at 10.8Gbps, for a complete of 173GB/sec of bandwidth. W7500 has no direct predecessor, however for system builders who completely want a sub-75W card, by each metric it’s a great deal extra highly effective than the Radeon Professional W6400 that was beforehand AMD’s finest card on this phase.

Each playing cards are single-slot designs – and, I believe, are utilizing the identical PCB beneath the hood – squeezing in 4 full-sized DisplayPort outputs into the restricted house provided by their kind issue. This has been AMD’s go-to design for 600/500 sequence professional playing cards for a number of generations now, and clearly one which has labored effectively for AMD over time.

As famous earlier, AMD is pushing these playing cards largely on the premise of RDNA 3’s new performance, DisplayPort 2.1 arguably being essentially the most instantly related for AMD’s skilled crowd. All 4 DisplayPorts can drive screens at as much as UHBR10 information charges – 10Gbps uncooked per lane, or 38.7Gbps in whole per port – about 50% extra bandwidth than was provided by the previous-generation W6600 with DisplayPort 1.4. Which, amongst different issues, allowed for 6K video at 60Hz with out compression, or extending 8K video to 120Hz with compression.

Skilled customers will need to make observe, nevertheless, that AMD’s DisplayPort performance is inconsistent throughout the W7000 household. The upper-end W7900/W7800 elements additionally supported even larger DisplayPort switch charges (UHBR 13.5 and UHBR 20), which the W7600/W7500 don’t. This type of show discontinuity with AMD merchandise is uncommon, so professional customers will must be conscious of it, least they get caught off-guard by the diminished capabilities of the mid-range playing cards. That stated, given how few shows there are available on the market that help UHBR information charges over 10Gbps, it’s unlikely too many individuals will likely be making an attempt to hook them as much as something lower than a high-end video card to start with.

Each playing cards additionally get the brand new AV1 encoding capabilities that include RDNA 3. AMD’s video processing block can encode with the cutting-edge codec at as much as 8K resolutions at 60 frames per second. These blocks have sufficient efficiency to deal with simultaneous video encode and decode, however in contrast to the much more highly effective high-end Professional playing cards, encoding is proscribed to a single AV1 steam (or two HEVC streams).

Lastly, on the compute structure entrance, RDNA 3 brings with it new options that enable for higher AI compute efficiency in addition to higher ray tracing efficiency. I’m unsure how a lot the previous will likely be tapped for Professional playing cards, however the ray tracing enhancements needs to be notable for the CAD/CAM/graphics crowd, the place the usage of {hardware} ray tracing is slowly seeping in.

What you received’t see AMD utilizing to push these playing cards an excessive amount of on, nevertheless, is efficiency. Navi 33 is in lots of respects a cost-optimized model of the Navi 22 GPU used within the W6600, and consequently the efficiency of the W7600 is usually not considerably larger than the W6600. The truth is, in some circumstances it’s truly decrease.

First off, kudos to AMD right here for being sincere with their generational efficiency comparability; fairly than cherry-picking outcomes or ignoring the topic solely, they’re publishing a whole information set. However their SPECviewperf 2020 outcomes underscore the inconstant efficiency beneficial properties going from one technology to the subsequent. The W7600 flies in compute-bound workloads due to the cardboard’s higher reminiscence bandwidth and compute throughput, however elsewhere regresses on three of the sub-tests. Whereas I haven’t examined both card personally, primarily based on what I find out about these checks an AMD’s specs, that is doubtless a results of the decrease increase clockspeed of the W7600, resulting in slower efficiency in each the entrance and back-ends of the pixel rasterization pipeline. Based on AMD, the W7600 nonetheless comes out forward total by 19%, although it’s definitely a smaller generational uplift than we’ve seen in different professional elements.

Extra stunning right here might be the W7500. Although it clearly doesn’t win, it doesn’t fall effectively behind the W6600, both, regardless of an virtually 50% decrease TDP. With reference to energy effectivity, it’s clearly working at one thing nearer to Navi 33’s candy spot.

In fact, AMD doesn’t have to outcompete itself a lot because it must outcompete NVIDIA. And whereas I’ll chorus from posting AMD’s comparative efficiency slides, AMD believes they’re in a great place. That is helped in no small half by the truth that NVIDIA hasn’t launched any Ada Lovelace structure professional playing cards for the mid-range market, so in the intervening time AMD wants solely to compete with older Ampere (and Turing) {hardware}.

Wrapping issues up, AMD’s latest Radeon Professional playing cards will start showing available on the market in a while within the present quarter, and in OEM methods in a while within the 12 months. The Radeon Professional W7600 will hit the streets at $599 – $50 cheaper than its predecessor – whereas the W7500 will convey up the rear of the W7000 product stack at $429.