Cycle Manufacturer Gets Personal Supercomputer — and Its Off to the Races

Predator Cycling uses the Lenovo ThinkStation P620 workstation for simulation and rendering.

Aram Goganian, CEO and co-owner of Predator Cycling does CFD simulation. (Picture courtesy of Lenovo.)

Aram Goganian, CEO and co-owner of Predator Cycling does CFD simulation. (Picture courtesy of Lenovo.)

Computing seems to be moving inexorably to the cloud. The concerted media industry —us included—has jumped on the cloud bandwagon. Because its benefits are obvious: capacity as needed, cores galore for simulation, pay only for what is used without the bother of maintenance or the cost acquisition of an expensive workstation, use of any device and more. To question the inevitability of the cloud as the future of engineering computing would seem so backward.

Lenovo, which must not have been paying attention to cloud praises, delivers a workstation that bucks the trend of remote computing. Its ThinkStation P620 workstation, introduced last fall, is a local powerhouse like we thought we would never see again.

Supercomputer, Anyone?

Aram Goganian, CEO and founder of Predator Cycling, scored a Lenovo P620 workstation—and it was off to the races. (Picture courtesy of LinkedIn.)

Aram Goganian, CEO and founder of Predator
Cycling, scored a Lenovo P620 workstation
—and it was off to the races.
(Picture courtesy of LinkedIn.)

We meet Aram Goganian, co-owner and CEO of Predator Cycling LLC, who has done an about-face on workstations.

“A couple of years ago, we ran our company of these little network PCs and small laptops,” said Goganian. “We didn’t find that limiting us. Then Ansys opened our eyes. And then KeyShot.”

Goganian was able to get renderings done locally in less than a minute on the P620—less time that it took to upload a model to the cloud and download the result. Ansys Discovery, which does real-time simulation using GPUs, flowed as smooth as silk on the P620’s pair of NVIDIA graphics cards.

How does one small manufacturing company score what could be a $50,0000 workstation (more on the cost later)?

“Lenovo asked me if I wanted a supercomputer on my desk,” explained Goganian.

The Making of a Predator

Aram Goganian began racing bikes at 13. He started Predator Cycling when he was 14. Now in his early 30s, he takes the lead in applying the latest design and simulation technology to his company’s products. Goganian has no formal engineering degree. In our conversation, he managed to present the best case we’ve heard for not needing one.

Predator bike frames are made with carbon fiber and are analyzed for deflection and stress. They are streamlined using computational fluid dynamics (CFD). While CFD simulation is almost always learned on the job, a course in composites is arguably the toughest elective a mechanical engineering student can select. [Editor’s note: The author vividly remembers boards filled with differential equations and matrices.]

Predator’s radically thin handlebars (below) are a case in point. Goganian has no trouble explaining that the handlebars needed to have the fibers lengthwise along the top and bottom (where there is maximum stress) and how he would set up Ansys composite elements for the simulation.


Wings or handlebars? Predator’s Major Pilot track handlebars streamlined with CFD and strengthened with FEA. (Picture courtesy of Predator.)

That it’s balanced when stationary may be the only clue that this bike is not for real. Predator renders all product shots. No more photo shoots. The rendering is completed in 40 seconds or less, thanks to the Lenovo P620 and the NVIDIA RTX 6000 graphic boards. Shown is the company's RF40 frame. (Picture courtesy of NVIDIA.)

That it’s balanced when stationary may be the only clue that this bike is not for real. Predator renders all product shots. No more photo shoots. The rendering is completed in 40 seconds or less, thanks to the Lenovo P620 and the NVIDIA RTX 6000 graphic boards. Shown is the company’s RF20 frame. (Picture courtesy of NVIDIA.)

Changing the Workflow

The company’s RF20 frame is borne of fluid and structural simulation, all of it done with the Lenovo
ThinkStation P620 equipped with a pair of NVIDIA RTX
A 6000
GPU filled cards. The power of the P620 was enough for Goganian to almost quit the cloud completely.

“We can solve the same model with more accuracy, or we can cut our time in half from our previous workstation,” said Goganian.

How’s the P620 fare with generative design, which simulation done repeatedly as a shape is optimized?

“We used to use Fusion 360, but that does not let us optimize material, so we are doing generative design with Ansys,” said Goganian.

A whole bike has fewer parts than any motorized, computerized transportation—unless you model like Predator does.

“We model everything on the bike, down to the links in the chain and the links holding them together. Still the model renders in 40 seconds,” said Goganian, who saves a lot of time simply by not having to defeature (simplify) the model.

“We used to take out the chain links, pins, the bearing assemblies.… We wouldn’t put those things in because the extra details would make our CFD meshes huge. Once we got the RTX 6000, we stopped simplifying. We threw the entire frickin’ bike at it, and it didn’t matter. It would just solve it.”

Running simulations does not render the P620 useless for other tasks. While the P620 workstation was running a CFD simulation, Goganian was able to use the computer for other tasks—even other simulations—a credit to the workstation’s architecture.

“I’m now running Discovery simulation on one screen and running a composites solution in the background. The workstation is literally multitasking. It’s really trippy.”

It’s a sweet change from his water-cooled, overclocked workstation with NVIDIA Quadro Pro graphics.

NVIDIA’s Role in Computing and Rendering

Goganian attributes much of the increase in processing speed he now enjoys to the NVIDIA RTX graphics, where a lot of the CFD solution happens.

“We were very fortunate to be one of the early testers of the NVIDIA RTX 6000.

“We threw everything at it, including Ansys Discovery [for structural FEA] and CFD. Our computational time was drastically reduced. It dropped by a factor of three or four with models with the same mesh size. We were able to run meshes that were almost 200 percent bigger in the same amount of time. Discovery was used and done on the entire frame at once.”

The P620 has also come in handy for mold simulation, which has enabled Goganian to shorten the manufacturing process.

“I used to have to design plugs that we had to load into our final mold. I had to design, produce the plugs, and then have to machine them out of the frame. I don’t have to do that now. Plus, with mold flow simulation, I know where to put the heaters and the cores.

“We use KeyShot for our visualization and that has been another big-time saver. We used to manufacture the frames, do a paint scheme and send [them] to be photographed. We had to build the bike to get the visual. We don’t do that anymore. We render it. You can render with your CAD software, but the customer can tell it’s a rendering. With KeyShot and the GPU renderer, it’s ridiculous. You get photorealistic renderings in 30 seconds, 40 seconds. They are beautiful. Our entire website is filled with KeyShot images. And for all the 500 some renderings, the longest time it took was 40 seconds. It’s amazing.”

The Personal Supercomputer

The Lenovo P620 begs for a new category of computer: the personal supercomputer. It was the first to use the AMD Ryzen Threadripper PRO microprocessor, can harbor multiple graphic boards, has capacious memory (up to 512 GB), and can accommodate vast amounts of storage (up to 20 TB), either with the preferred solid-state drives or traditional spinning disk drives.

The new AMD chip is a beast. Tom’s Hardware declares it the most powerful workstation chip available. The most powerful of the Threadripper PRO lineup, the 3995X has a whopping 64 CPU cores all on one chip (compared to Intel’s best of 28 CPU cores) that can manage 128 threads, run as high as 4.2 GHz and draw 280 watts of power (over a quarter of the P620’s power supply). Before the Threadripper PRO, the maximum that a single CPU system could handle was 18 cores. A dual-processor system could bump the total to 56 cores. AMD is the first, and may still be the only professional workstation processor, to support PCIe 4.0, which relieves the I/O bottleneck with twice the I/O performance as its predecessor, PCIe 3.0.

The Threadripper PRO was introduced simultaneously with the Lenovo P620 and, at the time of this writing, Lenovo is still the only one of the Big Three computer manufacturers to run with it.
AMD compares the Threadripper to Intel’s top-of-the-line processors and finds it superior for Revit use. (Picture courtesy of AMD.)

AMD compares the Threadripper to Intel’s top-of-the-line processors and finds it superior for Revit use. (Picture courtesy of AMD.)

How Much?

How much does a near supercomputer like the P620 cost? We consulted Lenovo’s website and came up with a $30,000 price tag for a system that was only near maximally loaded. It had the top-of-the-line 64-core Threadripper PRO 3995WX, the maximum of 64 GB of ECC RAM, but we opted for a pair of NVIDIA Quadro RTX 6000 graphic cards. The top-of-the-line NVIDIA Quadro GV100s would have set us back another $20,000.

When you pick yourself off the floor, we can tell you the base price of the P620 is $3,600, which gets you a 12-core processor, 16 GB ECC RAM, one NVIDIA Quadro graphics card and 256 GB solid-state drive. At the time of this writing, Lenovo was offering a specially priced P620 with a Threadripper PRO 3955WX (16 cores), more memory (32 GB ECC RAM), more storage (1 TB SSD) and one NVIDIA Quadro RTX 4000.

The Pendulum Swings Back?

Does Lenovo’s near supercomputer workstation signal a return to the desktop workstation? Will engineers who cleared their literal desktop for a lithe little laptop to embrace the cloud be eager for the return of a huge hulking workstation?

The Lenovo ThinkStation P620 will not be for every engineer—and with its price, it cannot be. But for those who are doing many simultaneous compute-intensive tasks and for those whom the cloud has not lived up to its promise—or simply cannot be used—the workstation could make a lot of sense.

But you have to ask yourself, are you worthy?