$1,599 RTX 4090 Beats $6,800 RTX 6000 Ada in Content Creation

RTX 6000 Ada Generation
RTX 6000 Ada Generation (Image credit: Nvidia)

Given its $6,800 price tag, you probably won't see Nvidia RTX 6000 Ada on any lists of the best graphics cards for gaming. But no one can deny that the RTX 6000 Ada is a powerhouse for professional users and content creators. The RTX 6000 Ada Generation, as it's officially known, features Nvidia's flagship AD102 silicon, which the chipmaker also uses in the mainstream GeForce RTX 4090.

For the RTX 6000 Ada, Nvidia enabled 142 SMs out of the possible 144, so the graphics card is just two SM shy of using the full AD102 die. The RTX 6000 Ada arrives with 18,176 CUDA cores, 568 Tensor cores for AI workloads, and 142 RT cores for ray tracing tasks. With the potential to boost up to 2,505 MHz, the graphics card flaunts an impressive FP32 performance that tops out at 91.06 TFLOPs. There's also 48GB of 20 Gbps GDDR6 memory onboard with a 384-bit memory interface to land the graphics card at a memory bandwidth of up to 960 GBps. Note that the TGP (Total Graphics Power) on the 6000 Ada is capped at 300W, however, compared to 450W for the RTX 4090 Founders Edition.

The previous RTX A6000 delivered 38.71 TFLOPs FP32 performance and 768 GBps of memory bandwidth. Therefore, the RTX 6000 Ada offers 135% higher FP32 performance and 25% more memory bandwidth than the RTX A6000. The GeForce RTX 4090, with 82.58 TFLOPs FP32 performance and 1,008 GBps memory bandwidth, is probably the closest rival to the RTX 6000 Ada.

While Nvidia has ported its Ada Lovelace architecture over to the company's professional offerings, AMD has yet to release any professional-grade graphics card based on RDNA 3. The only competing product currently is the Radeon Pro W6800, dating back to the RDNA 2 days. Unfortunately, the Navi 21-based graphics card's 17.82 TFLOPs FP32 performance and 512 GBps memory bandwidth look rather puny beside the RTX 6000 Ada. But let's see what Puget's benchmarks have to say.

RTX 6000 Ada Benchmarks

Swipe to scroll horizontally
Graphics CardDaVinci Resolve StudioAdobe Premiere ProTopaz AI SuiteUnreal EngineV-RayOctaneRenderRedshiftBlender
RTX 6000 Ada2,9751,498471.274.05,3991,1468712,728
GeForce RTX 40902,9601,505465.592.55,5641,2928512,779
RTX A60002,6481,379445.246.32,7106271595,848
Radeon Pro W68002,1351,017401.624.7N/AN/AN/A1,982

According to Puget Systems' thorough review of the RTX 6000 Ada, professional users can expect up to 2X higher performance from the Ada-powered graphics card compared to its Ampere predecessor. Of course, reducing workload times in half is always a welcome improvement. The results didn't come as a total surprise since the RTX 6000 Ada specifications insinuate as much, but it's still good to have benchmarks that quantify performance.

The GeForce RTX 4090 was superior to the RTX 6000 Ada, outperforming its professional counterpart in six out of eight benchmarks. The GeForce RTX 4090 delivered between 3% to 25% higher performance. The Radeon Pro W6800 wasn't a match for the RTX 6000 Ada, as expected. The latter was about 17% to 542% faster than the former, depending on the workload.

It's beyond dispute that the RTX 6000 Ada is a solid option for professional workloads. Unfortunately, the eye-watering $6,800 MSRP is enough to scare off potential buyers, even if the graphics card will likely pay off in the long run. Note also that none of the above benchmarks benefit from Nvidia's professional drivers — Siemens NX for example typically sees massive performance gains compared to GeForce cards.

The GeForce RTX 4090 is an attractive alternative for the less serious user since it starts at $1,650. You lose the stability and software validation you get from the Nvidia professional drivers and 24GB of VRAM. However, if you don't need a blower GPU and use applications that aren't specifically de-optimized by Nvidia's GeForce drivers, the RTX 4090 ($1,599 MSRP) can lend you its firepower for 76% less than the cost of an RTX 6000 Ada. Little wonder Nvidia doesn't seem to want more compact blower variants of the RTX 4090.

Zhiye Liu
News Editor and Memory Reviewer

Zhiye Liu is a news editor and memory reviewer at Tom’s Hardware. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

  • RichardtST
    But with the massive price difference, why doesn't a 3rd party make a swap-out heatsink? What did you mean by "specifically de-optimized "? Is that a real NGreedia thing? They detect that you're trying to do real work on a gaming GPU and intentionally nerf the thing so that you have to pay 3x more? That's just rude.
    Reply
  • Ravestein NL
    Your very quick with this comparison of cards but the "P" list does not show clock speeds and power levels so of course there could be little difference. Specifically when the RTX 4090 is burning 450+ W of power against the 300 W of the W6000 card.
    What worries me more is your quote "specifically de-optimised" which suggests Nvidia is controlling the GPU market by software restrictions. Could you elaborate on that please?
    Reply
  • Lug
    RichardtST said:
    But with the massive price difference, why doesn't a 3rd party make a swap-out heatsink? What did you mean by "specifically de-optimized "? Is that a real NGreedia thing? They detect that you're trying to do real work on a gaming GPU and intentionally nerf the thing so that you have to pay 3x more? That's just rude.

    All GPU manufacturers do the same and there is nothing abnormal, when you buy a professional card you pay for the professional driver. Just because your PC is capable of running software doesn't mean the software should be free, GPUs are the same.
    Reply
  • crimsonfilms
    RichardtST said:
    But with the massive price difference, why doesn't a 3rd party make a swap-out heatsink? What did you mean by "specifically de-optimized "? Is that a real NGreedia thing? They detect that you're trying to do real work on a gaming GPU and intentionally nerf the thing so that you have to pay 3x more? That's just rude.

    24GB vs 48G. Depending on the tasks and workflow this can be a deal breaker. For large visual processing, large VRAM is a requirement. For a long time, Hollywood CGI was processed in 2K using CPU because GPUs did not have enough VRAM to process it. That is changing but so is the resolution and color fidelity. Thus more VRAM.

    Also, the RTX6000 ADA seems to be more energy efficient. Power and heat are often factors when considering the hardware.
    Reply
  • Firestone
    Are we forgetting about how the 4090 is a 3-slot behemoth enormous gigantic card, while the 6000 is two-slot? And then 24GB vs 48GB, 300W vs 450W, there is really nothing to compare here. No one buying a 6000 seriously expects it to match up to a 4090 in performance, they are buying it because they need 48GB of memory and a smaller form factor and lower power draw.
    Reply
  • kyanwan
    @Firestone - you're totally right + on point.

    4090 "beats" the 6000ADA + 6000 .. and 5500, but - it doesn't really cut it.

    It's the memory & P2P, people. Anything beyond a single 4090 - performance will be (is) lackluster. I am doubtful it will improve - as Nvidia is actively shifting consumer GPUs away from HPC applications. Gaming GPUs look good on paper .. that's it.

    4090 doesn't have what serious MultiGPU creation / HPC needs - and, what did I do with my own 4090s? I sent them back. They're not sufficient for HPC of any sort. I opted to swap from 2x 4090 to 4x A5500. In the meantime, I'm waiting on either A100-40 to get a hair more affordable (* My original plan was 2x A100-40 - 4090 can't compete at all ) .. or for some other option in the A series that beats the A6000 ADA.

    TBH - I need VRAM. Way more VRAM. 96G isn't near enough - My own workstation - before & after disappointing 4090 build. As part of our need for more memory, we get more cards which add more processing power which in turn - absolutely obliterates everything the 4090 has to offer.
    Reply