SPECworkstation 3

The best place to start for performance is to confirm that this system does get the best SPECworkstation 3 score ever. For users who have never heard of SPECworkstation, it comes from the same people that have the SPEC benchmark that we often use on new processors. The workstation element comes in because this set of benchmarks are designed to test a number of common workstation workloads, such as 3D rendering and animation, molecular modeling and dynamics, medical, oil and gas, construction and architecture, financial services, general operations, and GPU compute. This benchmark combine 30 workloads and ~140 tests into a single package, and results are given as a multiple of a performance compared to a ‘reference’ machine using an Intel Quad-core Skylake processor running a W3100 AMD GPU. This means that this quad-core Intel system gets a value of ‘1’.

SPECworkstation 3 Test Systems
AnandTech CPU GPU DRAM SSD Price
Fujistu Celsius R970 2 x Xeon 8276 RTX 8000 DDR4-2933 PCIe 3.0 $30000+
Armari Magnetar X64T TR3 3990X RTX 6000 DDR4-3200 PCIe 4.0 ~$14200
TR3 3990X 'Stock' TR3 3990X 2080 super DDR4-3200 SATA -
W-3175X 'Stock' Xeon W-3175X 2080 Ti DDR4-2933 SATA -

The current system at the top of the official SPECworkstation 3 standings is a Fujitsu Celsius R970 workstation (D3488-A2). This is the system that Armari has beaten with the X64T. The Fujitsu uses two Intel Xeon Platinum 8276 processors (28-core each, total 56-corepaired with an NVIDIA Quadro RTX 8000 and 384 GB of DDR4-2933. This system, going on list prices for just these components, already comes to $24538. Add in the rest, and some overhead, and this is easily $30000+. By comparison, Armari’s Magnetar X64T workstation is only ~$14200.

The results are as follows. Here we are comparing the Fujitsu official results to Armari’s official results. We also have included our results with the same system (technically classified as ‘estimated results’ because these haven’t been formally submitted to the results database), and a W-3175X system with an RTX 2080 Ti and PCIe 3.0 SSD.

SPECworkstation 3 Results
AnandTech Fujitsu
+ 2080
2080 Ti
Media and Entertainment 4.72 7.04 6.84 4.79 3.69
Product Development 6.07 10.85 9.95 3.51 3.35
Life Sciences 5.89 8.24 8.11 - 3.72
Financial Services 8.78 10.55 10.45 9.15 6.59
Energy 5.44 9.09 8.73 4.20 2.86
General Operations 2.27 2.53 2.45 1.55 1.59
GPU Compute 5.40 5.75 5.70 4.63 5.01
Geomean 5.17 7.06 6.84 4.08 3.54

*As submitted to SPEC

Within each of these segments, 7-20 sub-tests are performed covering CPU, GPU, and Storage workloads. Our results were a little lower than Armari's, however that can be down to tuning, ambient temperatures, and repeated runs. Our run was within 3%.

Overall, the Magnetar X64T results beat the old Fujitsu results by 37%:

  • CPU: Armari wins by +46%
  • GPU: Armari wins by +12%
  • Storage: Armari wins by +58%

Now, users might wonder how the Armari wins in the GPU tests, given that it has an RTX 6000 compared to the RTX 8000 in the Fujitsu. This is namely down to processor performance – the Fujitsu system processors have a base frequency of 2200 MHz, compared to the Magnetar X64T which can run all processors at 3925 MHz. Even if the Fujitsu was using the CPU in single core mode, and hitting its max turbo of 4000 MHz, the Armari would be using the better IPC of the Zen 2 core against Intel’s Skylake core.

Now each of the above tests are combined scores from sub-tests.

The Intel-based Fujitsu system does have some specific wins in individual tests, such as Maya Storage (+15%), NAMD Storage (+12%) and 7-zip CPU (+75%), however these mostly apply due to the increased memory capacity of the Intel machine.

The AMD-based Armari system has 40 other wins, including Blender CPU (+62%), handbrake CPU (+86%), CFD CPU (+108%), NAMD CPU (+164%), Seismic Data Processing (+230%), LAAMPS storage (+88%), and Creo GPU (+55%).

Full data for the Armari and the Fujitsu systems can be found at these links:

The Armari Magnetar X64T Workstation Rendering Benchmark Performance
Comments Locked


View All Comments

  • WaltC - Wednesday, September 9, 2020 - link

    Very impressive box!...;) Great write-up, too! Great job, Ian--your steady diet of metal & silicon is really producing obvious positive results! I would definitely want to go with a different motherboard, though. Even the GB x570 Aorus Master has received ECC ram support with the latest bios featuring the latest couple of AGESA's from AMD--so it seems like a shoe-in for a TR motherboard. I agree it's kind of an odd exclusion from AMD for TR. But, I suppose if you want ECC support and lots more ram support you'll need to step up to EPYC and its associated motherboards. This is a real pro-sumer offering and the price--well, everything about it--seems right on the money, imo. I really like the three-year warranty and the included service to change out coolant fluid every three years--very nice! If I used water cooling that is the only kind of fluid I would want to use--some of these el cheapo concoctions will eat up a radiator in a year or less! Enjoyed the article--thanks again...;)
  • Makaveli - Wednesday, September 9, 2020 - link

    "Ideally AMD would need a product that pairs the 8-channel + ECC support with a processor overclock."

    But why?

    I maybe wrong here but when you are spending 10k+ on a build isn't stability more important than overclocking?
  • MenhirMike - Wednesday, September 9, 2020 - link

    Overclocking doesn't have to be unstable - and for some workloads, the extra performance is worth the effort to find the limits and beef up the cooling solution.
  • Ian Cutress - Thursday, September 10, 2020 - link

    Had a chat with Armari. The system was built with the OC requirements in mind, and customized to support that. They're using a PBO-based overclock as well, and they've been really impressed with how AMD's latest variation of PBO can optimize the DVFS of the chip to keep the system stable regardless of workload (as long as heat is managed). In my testing, there was zero instability. I was told by Armari that they can build 10, 20, or 50 systems in a row without having any stability issues coming from the processor, and that binning the CPU is almost virtually non-existant.
  • Everett F Sargent - Wednesday, September 9, 2020 - link

    So, such a waste of power, time and money.

    Anyone can build TWO 3990X systems for less then half the price AND less then half the power consumption with each of them easily getting ~90% of the benchmarks.

    That is with a top of the line titanium PSU, top of the line MB, top of the line 2TB SSD, top of the line Quadro RTX 4000 (oh damn paying 5X for the 6000 for only less then 2X the performance, what a b1tch, not), top of the line ... everything.

    So ~1.8X of the total performance at ~1.8X the cost (sale pricing for all components otherwise make that ~1.9X the cost) and ~0.9X of the total power consumption.

    But, you say, an ~15KUS system pays for itself, in less then 1E−44 seconds, even. Magnetarded indeed. /:
  • TallestGargoyle - Thursday, September 10, 2020 - link

    That's a pretty disingenuous stance. Yes, you can split the performance across multiple systems, likely for cheaper, but that ignores general infrastructure requirements, like having multiple locations to set up a system in, having enough power outlets to run them all, being capable of splitting the workload across multiple systems.

    Not every workload can support distributed processing. Not every office has the space for a dedicated rack of systems churning away. Not every workload can sufficiently run on a half-performing GPU, even if the price is only a fifth of the one used here.

    If your only metric is price-to-performance, then yes, this isn't the workstation for you. But this clearly isn't hitting a price-to-performance metric. It's focusing on the performance.
  • Tomatotech - Thursday, September 10, 2020 - link

    Something something chickens and oxen.

    Or was it something something sports cars and 49-ton trucks? I’ll go yell at clouds instead.
  • Spunjji - Friday, September 11, 2020 - link

    Perish the thought that someone might order one of these with a Quadro 4000 and obviate his biggest gripe (not that it would solve the potential problem of fitting a dataset into 8GB of RAM instead of 24GB)
  • Everett F Sargent - Saturday, September 12, 2020 - link

    Well there is that 20% VAT that pushes the actual system price to ~$17K US so dividing by two ~$8.5K per system minus the cost of a RTX 4000 at ~1K US gives one ~$6K US sans graphics card (remember I said build two systems with RTX 4000 at ~$7K US).

    So I have about ~$2.5K per system for a graphics card. Which easily gives one RTX 5000 per build at ~$2K US. Whiich is 2/3 of the RTX 6000.

    But, if the GPU benchmarks are all eventually CPU bound, which appears to be the case here in this review, then yes I get two systems at ~$8K US or ~$16K US for two systems, in other words, I get to build a 3900X system for about ~$1K US to boot..

    So no, I would rather see benchmarks at 4K, mind you (as these were at standard HD) with different GPU options. Basically, I would want to use my money wisely (assuming at a minimum a 3990X CPU at ~$4K US as the entry point).

    Therefore, this review is totally useless to anyone interested in efficiency (which should normally include everyone) and costs (ditto).
  • TallestGargoyle - Monday, September 14, 2020 - link

    That still doesn't take into account the issues I brought up initially with owning and running two systems.

Log in

Don't have an account? Sign up now