Ryzen 7 2700x Workstation Solver Build
It’s been a while since we’ve posted a consumer platform build for working with solvers. AMD’s Zen architecture was initially released in March of 2017, with Zen+ being released in April of 2018, so plenty of time has passed to allow people to figure out how to get the most out of AMD’s new platform.
Our aim with this build is to show you a current build that will get you the most performance out of your solver, whilst allowing for some future upgradeability with AMD’s new generation of Ryzen 3000 CPUs (Ryzen Matisse) that are anticipated to launch mid 2019, and rumoured to feature CPUs with up to 16 cores and 32 threads. AMD has guaranteed that this new platform will still be compatible with the current generation AM4 socket.
Previously we aimed for a build that would cost around 1,000 USD. We will not have a budget restriction for this build (and are therefore not labeling it as “budget”). Instead we will simply go with what we feel will deliver the currently best performance on the Zen+ platform. That being said, where applicable we will also show you some alternative configuration options that will offer more options for different budgets. The approximate price of this build at current market prices is 1675 USD.
For whom is this build?
I would recommend this build to anyone who is looking to build a multipurpose computer that will perform well as a general workhorse for grinding, and studying with different types of poker software beyond just solvers. If you are looking for a machine that you would want to purely solve poker simulations with 24/7, I’d recommend either building a machine dedicated to that using AMD’s or Intel’s HEDT (high end desktop) or server platforms or renting a server with a server hosting company. If you would like to know more about said builds, an example of a current HEDT build can be found here and a current server build here. For an entry level hosted server, we would recommend our Contabo Linux setup guide.
AMD’s current flagship on the consumer end. It features 8 cores and 16 Threads, with an out of the box boost clock of 4.3 GHz. It supports AMD’s new PBO (Precision Boost Overdrive) technology, which is a feature that helps automatically overclock your CPU for the best performance. This CPU is currently available for around 290 USD. The 2700x also comes with it’s own Wraith PRISM cooler, so you don’t need to purchase one additionaly. The noise levels the cooler produces under load may be too loud for some. If this applies to you, you may want to consider upgrading to a third party cooling solution like the NZXT Kraken X52 for example. A 3rd party cooling solution will also help you push the CPU a bit further through either manual overclocking or AMD’s aforementioned PBO technology.
Another, cheaper option for this build would be to pick up a last generation Ryzen 7 1700. These are currently available for roughly 160 USD. You will still get 8 cores and 16 threads, but with a lower boost clock of 3.7 GHz and no PBO.
A high-end motherboard that supports AMD’s precision boost overdrive overclocking technology. Relevant features include support for up to two NVMe drives that can be configured to run in a RAID configuration, which can be useful for usage with SQL databases, but also for load and save times when working with large solver files. This motherboard works well with the Ryzen 7 2700x thanks to it’s feature set. It’s large VRM (voltage regulator module) provides ample and will be useful when considering potentially upgrading to a 3rd generation Ryzen CPU later in the year. An additional bonus is that this motherboard comes with WIFI for those who need to use a wireless internet connection.
Corsair Dominator Platinum 64 GB DDR4 3200
RAM is a very important part of this build. Not only will faster RAM improve the performance of any solver we use, but the Ryzen 7 2700x itself benefits from faster RAM and you’ll be able to eek out extra performance from your CPU with this DDR4 3200 MHz kit from Corsair. In terms of value for money, 3200 MHz kits are the sweet spot, but faster RAM kits may give you a small additional boost. When purchasing RAM kits, it’s important to know that both the frequency and the timings of the RAM affect the effective speed. Lower CAS latency will be preferable. This Corsair kit has a CAS latency of 15. Alternatively, you can start off with a 2×16 GB RAM kit of the same RAM and upgrade later if you prefer, but we would recommend using at least two RAM sticks to take advantage of the dual channel memory configuration option.
The 970 EVO Plus is one of the fastest consumer grade SSDs available and performs very well with 4K mixed read/write operations. This is useful for work in software like Poketracker and Holdemmanager. On top of that, you will see fast load and save speeds when working with large files with a solver. If you wanted to further up your SSD performance, you could purchase a second one of these M.2 Sata drives and run it in a RAID 0 configuration, which is supported by the x470 Taichi motherboard. For additional storage, you may want to consider also adding something like a Toshiba X300 hard disk drive.
The Ryzen 7 2700x doesn’t come with an integrated graphics card, so we will need to pick up a graphics card for this build. Keep in mind that this will not have any impact on the performance of your solver, so what graphics card you choose will depend on what other use cases you would like to cover with this build. For example, you may want to also use this machine for video editing, or gaming. In those cases you may decide to use something more powerful than what we have chosen here in accordance with your personal budget and preference. We’ve selected the Gigabyte GT 1050 since it can output to two 4K displays without effort (or one 8k display) and is reliable. For something more beefier that won’t break the bank, you may consider upgrading to Nvidia’s recently released GTX 1660 Ti, which is a very capable mid-range graphics card and currently available for around 280 USD.
For cases you will have multiple options, just keep in mind that the case has to be able to fit the x470 Taichi motheroard and any cooling solution you use. We are currently using this case for our in-house Ryzen consumer build and have found it easy to build in, as well as aesthetically pleasing. It is compatible with the ASRock x470 Taichi ATX motherboard, and also allows for easy installation of AIO water coolers should you choose to add that to your build. It is also available in different colours and accents according to your taste.
PSU (Power Supply)
Seasonic G Series 650 650W 80+ Gold
High quality PSU by a trusted brand. Rated at 80+ Gold for high efficiency, which will save money over time.
This is the entire build as we would currently recommend it. If you have any suggestions or comments, please post below. We hope this was useful to you. Subscribe if you would like to be notified of future posts like these.
So it looks like it’s a way better investment than actually renting a server.
I have a few questions and would appreciate your feedback:
I just had a phone call with a computer store and they said that no cpu really supports more than 2666 mhz RAM. So even if you plug in 3000+ mhz RAM, it still won’t give you any speed advantage. We actually did rent a contabo server and noticed that our iMac loads sims way faster than the contabo server that has 256 GB RAM. But Contabo uses 1600 or 1800 mhz RAM sticks and the iMac has 2400 ones, so there must be a difference?`
So does it make sense to get the fastest RAM out there in order to load and run sims faster or is there a point where no difference will be seen?
Other than that, what’s the difference between the Ryzen and the Threadripper CPU? How much of an impact does the CPU have on loading times of the sims or is it rather only important when running sims?
Thanks in advance and Kind Regards,
Hello Jason. The speed increase with RAM depends a bit on the CPU being used, with the AMD Ryzen platform being more sensitive to RAM speeds than Intel. It’s probably accurate to say that Intel won’t benefit from more than 2666 MHz speeds much, but there is strong evidence to support Ryzen benefitting from speeds higher than 2666 MHz. Just search ‘Ryzen 2666 MHz vs 3000 MHz’ and you will see some benchmarks for various software. If your goal is to optimize load times of sims, the first place to look would be how fast you can read from your storage device. It’s the slowest component when comparing CPU, RAM and storage, so that’s where you should invest in to improve load times. After that, RAM would be the next piece of hardware I would optimize for.
Threadripper CPUs are more powerful (faster), with support for more RAM (currently 128 GB RAM versus 64 GB RAM). If you are contemplating a Ryzen 2x build, you may want to wait for Ryzen 3x, which will be released a month from now on July 7th. We will publish a build once the hardware is available for purchase.
Thank you for your reply.
We actually did rent a contabo server in order to run sims and all that stuff, figuring because it has 256 GB RAM and 20 core CPU it must be fast. However, that was not true. In fact, my iMac 2017 with 24 GB ram and the basic CPU is at least 2-3 times faster.
At first, we thought it must be the SSD, that’s not as fast on the server so I ran a test. And guess what? My iMac loads the sims at the same speed regardless of where I have them stored. I tried the internal SSD (2666mb/s), external SSD (522 mb/s) and a very old USB-flashdrive (220 mb/s). So I guess maybe what matters is where Monker is installed but not where the sims are loaded from?
What we did notice is that once I upgraded my RAM in the iMac from 24 GB to 64 GB, loading times decreased so I suppose in the end it’s about super fast RAM for the most part.
I will order the parts as recommended right here and let you know how well they perform.
I am not able to replicate your results with our hardware. I ran a quick test on some office hardware with an NVMe SSD versus an external SSD, and the load time on the external SSD took more than twice as long.
I ran this test on Windows 10, but I doubt that the OS could have such a dramatic impact on the way Monkersolver operates.
Did you retest the load times from various sources, after you upgraded your RAM?
We are running WIN 10 through Bootcamp on the iMac. The results we had on the contabo server were also made on WIN 10.
We even upgraded to the Intel Optane 900P SSD on contabo and it was still fairly slow.
What sims did you test, how big were they and how many streets did you load? Have you guys ever tested the same sims on the iMac or the iMac pro?
It seems you are saying that upgrading to a faster SSD on your server did improve your load times.
Based on what you have written, am I right to assume that you are loading the full game tree from a compressed file? If so, it makes sense that this operation would be heavier in CPU/RAM than storage. With a limited budget then I wouldn’t spend so much on storage, and instead make sure I have a faster RAM kit.
One more question regarding your workstation right here. I was about to purchase something similar and was told that this RAM needs additional water-cooling such as Corsair H115i RGB. Or is your mainboard perhaps already cooling the RAM itself?
No RAM should require water-cooling to be ran. Sufficient air flow in your case is enough to be able to operate at stock speeds, and even moderate overclock speeds. It may be possible to achieve higher overclocks with additional cooling, but this isn’t a requirement, and it’s unclear how much more performance you might gain from that.
Thanks for this. Do you know if this same build would work with 3900x?
Yes, This build is compatible with the 3900x. We will also be publishing a build specifically for this CPU soon, so stay tuned for that.
Here is our 3900x build.
Hi, wouldn’t be the threadripper 1920x a better option (since the price is about 40 € difference) having 12 cores at 4.00 GHz in turbo mode? Thanks!
This is an AMD x470 build. A Threadripper CPU requires a Threadripper motherboard.
hello! when is this build coming with the 3900x?
Our 3900x build is now available for reading here.