Nvidia has unveiled eight new professional graphics cards based on its new Ampere GPU architecture.
The RTX A5000 and A4000: new mid-range Ampere workstation GPUs
The latest cards to use the Ampere architecture, the RTX A5000 and RTX A4000 flesh out the mid range of Nvidia’s new generation of workstation GPUs, launched with the high-end RTX A6000 last year.
Unlike with the previous Turing architecture, and its dedicated RT ray tracing cores, Nvidia hasn’t introduced any new hardware types ths time round, though the designs of the existing cores have been updated.
Nvidia describes the Ampere cards’ second-gen RT cores as providing “up to 2x the throughput” of their previous-gen equivalents.
|Nvidia workstation GPUs
Ampere vs Turing
|19.2 Tflops||7.1 Tflops||27.8 Tflops||11.2 Tflops||40.0 Tflops*||16.3 Tflops|
|Memory bandwidth||448 GB/s||416 GB/s*||768 GB/s||448 GB/s*||768 GB/s||672 GB/s|
|Graphics bus||PCIe 4.0 x16||PCIe 3.0 x16||PCIe 4.0 x16||PCIe 3.0 x16||PCIe 4.0 x16||PCIe 3.0 x16|
|Display connectors||4 x DP 1.4||3 x DP 1.4*||4 x DP 1.4||4 x DP 1.4*||4 x DP 1.4||4 x DP 1.4|
More bang for the buck than their Turing counterparts
Since both of the new Ampere cards have more CUDA and RT cores than their Turing counterparts, each outpaces its predecessor when it comes to raw compute performance.
However, they also outstrip the Turing cards in most other key metrics: more GPU memory, higher memory bandwidth, PCIe 4.0 instead of PCIe 3.0, and lower maximum power consumption.
Expected street pricing remains similar to the previous-gen cards: up slightly in the case of the RTX A4000; down slightly in the case of the RTX A5000.
Up to ‘5x faster’ than the equivalent Tesla cards in selected GPU renderers
According to Nvidia, those new specs translate into a significant performance over the previous-generation cards when working with compatible GPU renderers.
Robert Cervellione, studio leader for design technology at architectural practice Woods Bagot is quoted as being “shocked” by the performance gap between the RTX A5000 and the equivalent previous-card card in applications like Chaos Vantage and Nvidia’s own Omniverse, although no exact figure is given.
New Ampere server and laptop GPUs
Of the other Ampere GPUs announced at GTC 2021, Nvidia describes the new A10 server card (above) as the “little brother” of the A40, its first Ampere data center GPU, announced last year.
It’s a single-slot 24GB card, with performance described as midway between that of the A4000 and A5000.
Its sibling, the A16, is a high-density dual-slot card that packs four 16GB GPUs onto a single card.
The four laptop GPUs – the RTX A5000, A4000, A3000 and A2000 – provide similar improvements over the equivalent Turing cards as their workstation counterparts.
We don’t usually cover mobile GPUs on CG Channel, but you can find a detailed comparison table here.
Pricing and release dates
The new workstation and server GPUs are due to ship later this month. The RTX A5000 has an expected street price of $2,250; the RTX A4000 will cost $1,000. Nvidia hasn’t announced prices for the A10 and A16.
The laptop GPUs will become available in Q2 2021 via third-party laptop manufacturers.
Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,