Ryzen 9 5950x Workstation Solver Build (128 GB RAM)
This is our updated workstation solver build for the consumer platform of AMD. In reality this CPU could be classified as enthusiast however, since there is nothing casual about running 16 CPU cores with a boost clock of up to 4.9 GHz.
At this time it’s a little difficult to give a precise number regarding the cost of this build, since the market prices are currently inflated due to supply chain limitations. Prices on either the CPU, GPU or RAM may not be at their equilibrium levels. Ideally this build should set you back approximately $3,000, but at the time of writing may cost as much as $4,000. I will point out which parts may have supply issues below.
A 16-Core CPU with a 3.4 GHz base clock and a boost clock of up to 4.9 GHz. Utilise AMD’s Ryzen Master software for auto overclock features to gain extra performance. The MSRP of this processor is $799. Currently this processor is being sold for around $1,000 in the US and supplies may be limited.
It is important to point out that AMD officially recommends water cooling for their Ryzen 9 5950x CPU. We ran some tests using the Ryzen 9 5950x and the Noctua NH-D15 running Monkersolver 2 (BETA) will all threads utilised. At stock speeds, temperature stabilised at around 61-62 degrees celsius. Running Ryzen Master with “auto OC” enabled, we saw temperatures rise to 67-68 degrees celsius. Both of these temperatures are well below the maximum thermal load of the CPU, which is 90 degrees celsius. If you prefer a liquid cooling solution, we recommend the EVGA CLC 280 liquid cooler.
The Asus Prime X570-Pro represents a good value point and support for the features of this build.
For the Ryzen 5000 series, the value point in terms of price performance lies with DDR4 3600 RAM. RAM may be hard to find given current supplies of memory chips. For this build we recommend a kit of Corsair Vengeance RGB PRO 128GB DDR4 3600 4x32GB RAM, for a total of 128GB.
Storage for solvers is more of a secondary performance metric as most of the time you will not be reading or writing to your hard drive. Therefore we would simply recommend a fast primary drive, such as the Samsung 980 Pro (NVMe) and secondary storage for backup, as needed.
Graphics cards are experiencing the most severe shortages at the moment. This is down to various reasons, but primarily due to a large shortage of semiconductors in the global supply chain. It is important to point out that a graphics card is not required for the majority of poker solvers. We have listed an Asus RTX 2060 here, as it is a good all purpose card to have in a modern consumer system, but it is definitely not required and any graphics card that can output an image to 2-3 high resolution monitors (4k@60) will suffice.
The Fractal Design Meshify 2 is a solid case with good out of the box thermals in an attractive packaging.
The Corsair RM850 provides plenty of power for this build, with room to spare for future upgrades, such as additional storage drives and or a more power hungry graphics card or additional cooling/lighting.
This is a very powerful build for everyday solving needs. The 5950x runs very fast, especially when combined with the correct RAM. There is room in the system for future upgrades. As mentioned, be mindful of the prices of some of the hardware. There may be surge-pricing given the current demand for computer hardware.
Please leave a comment below if you liked this build or you have any questions. Also let us know what other builds you might be interested in seeing in the future.
Disclaimer: We may receive a commission if you use one of our links to purchase any products on a partner site.
Threadripper Pro is now available for custom builds:
But I can only find 128 GB ram modules for it on Newegg and it only has 8 ram slots for the motherboard that seems best suited for it:
Also given monker is mostly ram intensive, why do you prefer this system over an EPYC build, I’m guessing cost? It seems threadripper pro is the good middle ground to me, what do you think?
Thank you for your comment Matt. We do not prefer this system over an EPYC build. There are three platforms that we post builds for: consumer, enthusiast and server. This build is part of the consumer builds. It is top of the list when it comes to “consumer” hardware. More builds for the enthusiast and server platforms are likely to follow. The 3995wx is a very good choice for Monkersolver calculations.
Monkersolver is both CPU and RAM intensive. The quantity of RAM mostly matters if your workload requires a large amount of RAM, otherwise the bottleneck will be the speed of the RAM, which depends on the memory architecture supported by the CPU and the type of RAM installed and also how the RAM is configured.
I found your site today; I am so happy I did because I am looking at a PC for Solver work.
I have a $3500 budget, but I already have SSD storage, a PSU, Case and liquid cooling. I use Simple Post Flop and GTO+, and my current I7-4770K and 32GB RAM build struggles to cope. I leave it overnight to perform the work.
I think the Ryzen 9 5950X could be sufficient as I mainly perform post-flop data. What amount of RAM would you advise? If I wanted to consider Pre-Flop, would a move to a Threadripper be the better option?
Thank you for your comment.
If using the 5950x, I would recommend 128GB RAM (4x32GB kit). This is the effective maximum currently supported by the AM4 platform and should give you enough flexibility for your workload. Keep in mind though that with most solvers, more RAM doesn’t equal faster solving times. So you only really need large amounts of RAM if it suits your workload and/or workflow. By that I mean that the calculations you are trying to perform require a certain amount of RAM or that it is perhaps beneficial for you to run multiple calculations in parallel. Also consider any other background tasks you may be performing at the same time, such as browsing with a large amount of tabs open etc.
I always recommend opening up the software that you use and seeing what the RAM requirements are for various workloads you intend to perform, in order to determine how much RAM you will need. There isn’t a one size fits all solution. However, in the case of the 5950x, since the RAM is capped at 128GB, I think it’s beneficial to just get a kit with 4x32GB from the get go. Especially given that this is more widely supported by AM4 motherboards now.
Threadripper could make sense as an option. If you want to use more than 128GB RAM it makes sense. If it’s within your budget, you could also just go with Threadripper for more CPU speed, but core for core you are paying a lot on top for not that much more performance.
Also keep in mind that Threadripper Zen 3 is scheduled to launch in August this year. It’s probably worth the wait if you are considering Threadripper, but pricing and availability is unclear at this point.
Thanks for your reply. I meant to ask if there any new CPUs in the pipeline, and you mention Threadripper Zen 3 in August. I’ll wait for their release and then decide.