Sidenote: My Upcoming New PC Build and How I Think About My System Builds

Since I can’t currently sleep and I’m also about 90% sure my plans for a new PC will get to proceed, I figured why not share the process? After reading Naithin’s post about the same topic, I decided it would be fun to share.

For me, PC building is actually almost more of the hobby I enjoy than gaming on said PC. When I show my wife or family my rig running something, it is less about the visual fidelity or smoothness of the game and more of a remark about how cool the system looks, or the size of the monitors, so it makes it sort of fun to chase after good looking hardware and to think about how to properly build things out for visual impact as well as performance.

My current system looks great and I enjoy it very much, but I bought some of the foundational hardware in it at a time prior to some major shifts in the PC market and so I miss some of the features that I can now take advantage of with current games I own, as well as missing some of the smarter performance of the newer generations of graphics cards and the upcoming new generation of CPUs.

But, rather than solely listing off a spec sheet, I want to explain a bit about how I build my systems. I overbuild, for sure – to be clear, you could get a great experience with far less than I am about to show – but again, the hobby of building a PC sometimes compels irrational decisions or overmatched hardware, given the amount of time I spend simply playing WoW or FFXIV, neither of which is actually that intense on a PC.

So, let’s go!

The current build and rationale for change: My current build is a fantastic rig, one I built in 2018 and refreshed very slightly in 2019 with a new CPU and a vertical mount kit for my graphics card. The core of the system is a Ryzen 9 3900x CPU under an NZXT closed loop liquid cooler, with 32 GB of DDR4-3200 memory and a Geforce GTX 1080 Ti Hybrid from EVGA – hybrid because it ships with a closed loop liquid cooler of its own for the GPU and VRAM, while using a heatsink and fan design for the power regulation hardware. All of that is tied to 12 TB of total storage split between a 1 TB NVME SSD boot drive, 2 500 GB SATA SSDs for scratch disk use in editing and rendering work, and 2 5 TB 7200 RPM platter drives, one of which is nearly fully consumed by digital games from Steam, Epic, Origin, and Uplay, and the other which is about half-full with around 15 years of digital detritus, old text files, photos, phone backups, MP3 collections, cracked pirated games I now almost fully own legitimately anyways, and a random collection of more current files I’ve sometimes dumped onto it as the SSDs in my system get close to their redlines.

In my main play games, it works very well. I can play both WoW and FFXIV at maximum settings with little stutter and few dips below the 100 FPS my main monitor supports, and I can play a large amount of my backlog at full framerates and fidelity with no hitches. As I’ve been streaming (*cough*, the system has kept up very well using CPU encoding to maximize the 12 cores available, running them at around 15-20% total utilization while gaming and streaming. Given that then, what is the excuse to improve?

Well, I have a few reasons I’ve mapped out in my head. The first is simple – real-time raytracing support is absent from my rig and with WoW moving to support it in 9.0, I find myself wanting a graphics card capable of it, not to mention the now 10 or so titles that I have that look stunning with it on. The second is that my system struggles with some of the non-gaming workloads I put it through – mostly Blender 3D rendering, which can sometimes take an obscenely long time and has often gobbled all 32 GB of available system RAM. A GPU capable of real-time raytracing can assist Blender’s ray-traced rendering and offers significantly better performance there, while allowing the CPU to assist in a more thoughtful and balanced way. For use in simple stream graphics and transition elements, it works well – but a big thing I’ve been doing lately is moving my hobbyist stage designing into Blender for actual lighting previews and trying to bring in full arenas has been very hit or miss. The mesh data for my newest full arena is 1.5 GB of data and I cannot open it in Blender without it dying, even after some troubleshooting. The third reason is that I have a PCI-Express 4.0 capable CPU that is not being used to its fullest potential, as I originally purchased the parts with a Ryzen second generation CPU and the motherboard vendor was locked out of enabling PCI-E 4.0 on the system, despite the fact that it does work if BIOS allows it. The fourth reason and arguably best, is that my wife is getting into more heavy media editing work for projects we are working on together and having a system that can manage that fully without being my much-older hand-me-down hardware would be nice (as she would get a newer hand-me-down)!

So I’ve established my reasons for why I want a new system and think one is viable and useful to my household (or at least, run down the excuses I’ve made for it!). How, then, do I select my hardware? Well, a few goals come to mind.

Beat the new consoles as much as possible: I am, first and foremost, a PC gamer. I vastly prefer playing on PC with the options that allows for, even while I grew up as a console fan first. With Microsoft doing day and date PC versions of first-party Xbox Series X/S titles, and my current console choice being the PS4, it logically tracks how my house will go – PC for me to play the MS stuff I like and would consider an Xbox for, PS5 to replace my 7-year old PS4 and to allow for Sony exclusives (when there are some I want to play, as I am absolutely not getting a PS5 day one, by choice).

If my PC is to match or exceed the Xbox Series X and PS5, a few things need to be lined up.

Zen 2 CPU with at least 8 cores: Already here with the Ryzen 9 3900X being a 12-core Zen 2 CPU, but whatever I replace it with must meet this standard at a minimum. All the Ryzen desktop parts since 2018 beat the consoles on clock speed, so that is less of a concern, but any overhead can help as the consoles do not have the overhead of a Windows PC.

PCI-E 4.0 NVME Storage: This one comes with caveats past just the interface, though. Firstly, both systems are using PCIE 4.0 interfaces of some sort – Microsoft is confirmed to be using PCIE 4.0 x2 for both the internal SSD and the expansion slot SSDs, so a full PCIE 4×4 as is standard for PC will provide enough bandwidth for that platform. However, both platforms make use of aggressive compression and decompression to accelerate performance drastically, and do so using custom logic blocks rather than dumping it onto their CPUs. This means a few possibilities for ports from these consoles: games could dump compressed archives of data they may need into system memory so they are accessible much faster at the cost of CPU time unpacking the data and extra system RAM being eaten alive by it, they could make use of new technologies like RTX IO or the larger Direct Storage API that rolls out next year, or they could simply require or run best under hardware situations that get you as close to the console performance as possible, with some combination of the other two points to bring performance even closer to the consoles (well, mostly the PS5).

Real-Time Raytracing: A recurring theme in this post – both consoles support it, and between that and graphics cards on PC from both Nvidia and AMD doing the same, games will start to move towards raytracing slowly, albeit faster than they did in the wake of the launch of Turing in 2018.

A Fast GPU in General: Both consoles have support for newer features my GTX 1080 Ti doesn’t have, like variable rate shading, stronger asynchronous compute and command queuing with concurrent INT and FP execution, raytracing (again), and Direct X 12 Ultimate featuresets. Lastly, while I won’t be upgrading my monitor to my coveted Samsung Odyssey G9 immediately, I want a GPU and display combo that can support HDR rendering and do so at smooth, high framerates.

Given all of this, then, where do I land?

The New Build:

Ryzen 9 5900X/5950X
64 GB DDR4-3600 RAM
X570 (maybe 670?) Motherboard
GeForce RTX 3080 (holding out for the rumored 20 GB card, 10 GB if not)
Watercooling loop from EKWB with CPU and Graphics Card blocks and dual 360mm radiators
1000w Power Supply

Full list link here (the CPU is a placeholder).

The CPU is a choice I am not decided on yet because the announcement from AMD is next week. Early on, I thought it was a no-brainer to upgrade to 16 cores and the 5950X. However, rumors and leaks are stirring suggesting two things that make me contemplate against that – first, that the 16-core will once again not be a day 1 launch part, and two, that the 12-core will be the flagship gaming CPU with a 5 GHz maximum boost clock. There were people suggesting Ryzen 3xxx parts could hit 5 GHz out of the box and that was wrong, so I’ll believe it when I see it and adjust my build accordingly, but if the 12-core can hit a higher boost clock and sustain it longer, then it probably wins over the ability to boast about 16 cores. If the 16-core clocks similarly and will be out prior to year end, however…it wins.

You might think “why Ryzen?” and my answer is simple – owning 3 different Ryzen rigs in our house over the last 3 years has sold me on it for multi-tasking, and a benchmark leak suggests that the fourth-gen CPUs are going to also win in gaming. Plus, with the consoles being Zen-based and Microsoft doing same-day release on PC ports from XSX, it should mean that optimizations made for the console CPUs will carry over to a Ryzen PC easily, offering further improvements. The early benchmark leak I saw shows as much as a 20% single core IPC uplift and a 15% multi-core improvement, which is a strong result that, if coupled with a 5 GHz or really close clockspeed, should mean that this era of AMD will win completely over Intel (without even contemplating Intel’s myriad security flaws which remain in-hardware until they move from Skylake-derivative core designs).

The RAM quantity is a hedge for two things – Blender’s insane appetite with my models and a bet that some next-gen game engines will use main system memory as a dumping ground for not-yet-needed data to keep parity with the console SSDs, especially the PS5. The faster speed is to eke out a small additional performance improvement, and because the sweet-spot on RAM cost keeps climbing speeds thanks to COVID-related overproduction of DRAM causing prices to fall. With a dual-chiplet Ryzen CPU in store, the faster RAM will also speed up the Infinity Fabric connecting the chips, which should reduce latency and keep things smooth and responsive.

The GPU is one I am actually flexible on, to a point. After firmly getting off of the RTX 3090 hype train, the 3080 is a damn good card, albeit with low (current) availability. As I am likely not building until November, this is fine, since the 3070 launch will hopefully ease demand a bit and make getting a 3080 more realistic. My ideal state is that this wait will also prompt Nvidia to launch the oft-rumored 20 GB version of the card. I want that for a couple of reasons – Blender yet again, a bet that RTX IO will help accelerate my SSD to the level of the PS5 but will need a not-insignificant chunk of VRAM to do so, and mentally, because going from 11GB of VRAM to 10 GB really bothers me internally in a way that is silly. As a darkhorse pick, I am also flexible to getting a Radeon RX 6900 XT (or whatever name the flagship Radeon card takes when announced on 10/28/2020). While my experiences personally with AMD drivers have left me a bit wary (and my reading of the R/AMD subreddit where reports of a myriad of instabilities are often dominating on the front page at all times), if they bring a strong performer out, I could be convinced. Realistically, such a card would need to beat the RTX 3080 by enough of a margin to make up for the loss of CUDA acceleration in Adobe suite apps, to offer compatible raytracing that comes at a similar or less loss of performance compared to the Nvidia offerings, and to offer something similar to the promises of RTX IO. My expectation is that I’ll end up with an RTX 3080 because the Asus Strix card I want is a good, strong card and EKWB already has a waterblock design for it, but I remain open to the possibility of a far more competitive AMD, given that they have put up a tremendous fight against Intel on the CPU front and done things most did not expect after the FX disaster.

A PCIE4x4 SSD is…fast? Not much more to say there – I’m holding to see what current-year offerings like the Samsung 980 Pro will cost in 2 TB sizes, as the reviews of the 1 TB version show a substantial improvement in performance even over last year’s 4×4 models (around 2 GB/s sequential read uplift and a similar amount of sequential write improvement with smaller random IOPS performance increases), but the Sabrent Rocket Q4 is well-reviewed and one I would be happy to take.

Custom watercooling is a project I wanted to undertake this year regardless of the host system, and the new system build offers a great chance for that. While I am not particularly noise-averse and have often run fans in my PCs at full speed for the sake of performance, I find myself enamored with the challenge of getting a near-noiseless system running. I also find watercooling visually appealing, especially with fun fluids and well-planned cable and tubing runs. From a performance perspective, getting a proper water system with copper radiators and stronger cooling capabilities should help maintain higher boost clocks on the CPU and GPU, while running fans quieter and slower, extending their usable lives.

The 1000w PSU is an overbuy, but one I feel comfortable with. It offers multiple GPU rails to meet the three 8-pin connectors on the Strix graphics card I’ve been eyeing for the RTX 3080, has dual CPU connectors to feed the motherboard, and is fully modular which I really like, even though my selected case very easily hides the PSU and cable clutter from it. Between a rumored 5 GHz Ryzen TDP of 150w, the Strix card’s capability to draw 400w+, and the drives, water pump, and RGB LEDs in the system, I could use most of the capacity and have some headroom for crazier overclocking if possible.

With that, I’ve summarized my thought process which will, hopefully, lead me to an absolutely incredible upgrade in a couple of months time, or made my crazy hobbyist obsession with gaming hardware more clear, at least.

One thought on “Sidenote: My Upcoming New PC Build and How I Think About My System Builds

  1. Ahh all the best for this! Looking forward to seeing how it all comes together and if much changes for you post announcements of detail, as it did with the 3090.

    I spec’d mine largely under yours, but since productivity use-cases for me entail office work — perhaps the odd oversized excel data set at a stretch — making gaming my primary demand on the system, I’m OK with that. I won’t lie — it was still tempting to go further than I did ‘just because’. 😉

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.