AMD’s Ryzen Threadripper PRO is the latest iteration of AMD’s heavy-duty Threadripper processor, based on the same Zen 2 architecture. Multiple generations of AMD’s Threadripper chips have distinguished themselves from the competition for their ability to deliver unbeatable multi-core performance for the price since the chips were first released in 2017.
The new AMD Threadripper PRO retains the base performance of its predecessors but ramps up the memory throughput considerably, giving the PRO range the ability to shuttle substantially more data to and from the processor cores.
This leap forward in memory bandwidth and total addressable memory is a boon for professionals working with very large datasets, be they 3D models or seismic maps of oil and gas reserves. For this reason, the PRO is likely to find its way into workstations used by everyone from visual effects artists and machine-learning researchers to engineers and geophysicists.
So what specifically makes the Threadripper PRO different from its predecessor? As can be seen by the impressive specs on the right side of the table below, it’s all about the data throughput.
Specifications of the AMD Threadripper PRO CPU, Image Source: AMD
The ability of the PRO processors to deliver more data to their cores, thanks to a doubling of available memory channels, helps solve the problem of processors not being supplied with data fast enough to keep up with their execution.
This broader pipeline for feeding the processor with data is combined with improvements to processing efficiency via better branch prediction in the Zen 2 architecture, allowing the chips to achieve superior CPU performance compared to earlier AMD processors. For example, Zen 2 architecture chips are able to handle up to 15% more IPC (instructions per clock) than previous generation processors.
1. Why Corporate AI projects fail?
2. How AI Will Power the Next Wave of Healthcare Innovation?
3. Machine Learning by Using Regression Model
4. Top Data Science Platforms in 2021 Other than Kaggle
The PRO lineup is not just a beast when shuttling data to the CPU, it’s also a natural choice for anyone looking for sufficient data throughput to high-performance storage, GPUs and other devices, with support for 128 PCIe 4.0 lanes, up from 72 in base Threadripper chips, and 64 lanes of PCIe 3.0 on the Intel Xeon W.
The volume of memory supported also takes a big step with the PRO range, up to 2TB, a major increase over the “up to 256GB” support on the non-PRO Threadripper systems.
The architecture that delivers the 4X floating point throughput on the Threadripper PRO. Image: AMD
As mentioned, the Threadripper PRO is designed to power workstations, highly powerful desktop computers capable of carrying out a range of computing workloads at a rate significantly faster than a consumer PC.
There are several key indicators of processor performance:
- the number of cores a chip has,
- the operating frequency of the chip — which reflects the speed at which it handles instructions,
- and the architecture of the processor, with each new revision utilizing new techniques that improve processor efficiency.
The diversity of features found in the PRO chips means power users should be able to find a chip that matches their particular performance needs. If raw processing speed is important, the 12-core PRO 3945WX workstation has got you covered, with its base operating frequency of 4GHz. Or, if pure processor core count is what you need, the PRO 3995WX lowers the operating frequency but is a multi-threaded beast, offering 64 cores able to handle 128 threads.
However, there is also more to performance than just core count, operating frequency, and processor architecture.
On paper, the almost $6,000 flagship Ryzen Threadripper PRO 3995WX is slower than the $3990 Threadripper 3990X, the top-end processor in the non-PRO Threadripper range. While the 3995WX has a base operating frequency of 2.7GHz and turbo speeds of 4.2GHz, the TR 3990X, has a base frequency of 2.9GHz and turbo clock of 4.3GHz, and both processors offer 256MB of Level 3 cache.
However, actual benchmarks of the two processors show the PRO chip often coming out in front, despite appearing slower on paper, with tech site Anandtech finding the PRO processor offered on average 3% better performance across core tests, including for rendering and simulation. The tester put this increase in performance down in part to the increased memory bandwidth available on the PRO system.
AMD Threadripper Pro Shell and Insides, Image Source
The PRO lineup doesn’t just outgun AMD’s base Threadripper, it also handily trounces a range of similar class Intel processors aimed at the workstation market.
AMD pitches the four Threadripper PRO chips as competitors to Intel’s workstation chips, including the W-2200, W-3200, and Xeon Platinum families. Third-party benchmarks of the flagship Threadripper PRO 3995WX found it outperformed Intel’s workstation processors in a range of rendering and other specialist tasks.
The PRO’s other contemporaries in AMD’s processor lineup are the Enterprise EPYC chips, which range up to 64-cores, offer support for eight channels of DDR4–3200 memory, and also support up to 4TB of ECC memory.
AMD recently launched the 3rd generation of EPYC processors based on the Zen 3 architecture, which offers both higher instructions per clock than the Zen 2 architecture found in the Threadripper PRO, as well as lower latency when transferring data between cores.
However, while the PRO chips are targeted at workstations, the EPYC line is typically used within servers.
In terms of raw specs, while the max core/thread count is the same in both lineups, the flagship Ryzen Threadripper PRO 3995WX is on paper more powerful than the best-performing chip in the 3rd gen Enterprise EPYC range: the EPYC 7763, which has a 2.45GHz base clock and 3.5GHz turbo clock.
The performance of Threadripper PRO chips isn’t generally benchmarked against that of EPYC processors, due to their different core markets. However, in tests, the Zen 3 EPYC processor line was found to offer 20–25% improved performance over EPYC processors based on the earlier Zen 2 architecture.
In general the Threadripper PRO chips use more power than the AMD EPYC line, according to the chip’s TDP (Thermal Design Power) rating, a measure that reflects maximum power usage. While the AMD EPYC’s power consumption ranges from 180W-280W, every chip in the Threadripper PRO range has a TDP of 280W.
AMD’s Infinity Fabric used on the Threadripper PRO processor, Image Source: AMD
We think the AMD Ryzen Threadripper PRO is worth the premium over the base model for professionals who routinely work with huge datasets, be they visual effects artists or geophysicists, due to the chip’s prodigious throughput, which helps keep the powerful Threadripper cores fed with data.
You can learn more about our AMD Ryzen Threadripper PRO Workstation options here.