Building a personal workstation: Picking up the pieces

Article By : Brian Dipert

This engineer's gamble on an "open box" AMD high-end desktop PC microprocessor didn't pan out for his planned high-end desktop (HEDT) build.

Assuming you’ve already read last month’s tale of woe (if not, I highly recommend a perusal of it, as well as the two preparatory pieces that preceded it, before continuing), you already know that my gamble on an “open box” AMD high-end desktop PC microprocessor—one that ended up actually being not only “used” but “used up”, i.e., DOA—didn’t pan out for my planned high-end desktop (HEDT) build. Thankfully, though, I eventually got my money back. So, what’s next?

The motherboard is a “sunk” cost, since as I discussed a couple of posts back in this series, I can’t use it with anything other than a third-generation AMD Threadripper CPU. I’ll keep my eyes open for another reasonably-priced sTRX4 socket-based processor, but as may already be obvious, I’m not going to take any more chances on anything “open box” or “used”. As such, I’m skeptical this particular “build” will ever move forward; I’ll probably end up selling the mobo on eBay or donating it to Goodwill for another to carry forward to a hopefully positive outcome.

Note, however, that the Gigabyte “kit” I’d acquired comprised not only the motherboard but also two add-in boards (AIBs), one for Thunderbolt 3 (TB3) support and the other implementing multi-m.2 SSD RAID. I intend to reuse both AIBs in other systems; in particular, the actualization of my TB3 card plans ended up being somewhat convoluted and restrictive, courtesy of OEMs’ proprietary implementations of the supposed industry standard. Stay tuned for a focused-topic piece to come soon, with more details on my travails and solutions.

I’ve got two “mainstream” systems builds planned, which I haven’t yet told you about; actually, one of them is already completed. Four of the eight DIMMs originally intended for the Threadripper HEDT will go in one of them; I’ll also be redirecting the two Sabrent SSDs to it.

I haven’t given up on the HEDT dream, but I’m redirecting my aspirations to an Intel-based solution and saving quite a bit of coin in the process. For the motherboard, I recently came across a highly regarded product from EVGA, the x299 Dark, on Woot for $279.99 plus tax.

As with the Gigabyte Threadripper board’s sTRX4 CPU socket, the x299 Dark’s socket (Intel’s LGA 2026) is also being phased out as Intel prepares to launch its Sapphire Rapids Xeon products (interestingly, Intel, like AMD, is also turning its back on Threadripper-like HEDT CPUs and focusing exclusively on Threadripper Pro-like “true” workstation offerings…but I digress). So why did I seemingly head back down an already-trod path that didn’t lead to my desired destination the first time? Unlike AMD’s one-generation sTRX4 experiment, Intel supported four generations’ worth of HEDT processors on LGA 2026 (and EVGA’s mobo also handles them all), so I haven’t had any trouble sourcing a reasonably priced, robustly featured CPU candidate.

Specifically, I’ve selected the 14-core (28-thread) Core i9-10940X from Intel’s most recent LGA 2026-based Cascade Lake-X product line. I found one brand new for $535 on eBay:

In contrast, Intel’s next (and highest-end) Cascade Lake-X offering, the 18-core Core i9-10980XE, would have set me back well over $1,000. Everything’s relative, of course: either Intel CPU would be notably cheaper than the AMD Threadripper 3960X I was originally planning to use, as is the Intel-intended motherboard versus its AMD-tailored counterpart. Then again, of course, the AMD CPU has a notably higher core count then either Intel alternative, too.

I’ll have more to say about the Intel HEDT system (enclosure, PSU, DRAM, SSDs, graphics, etc.), including upsides and shortcomings versus the AMD precursor, in a dedicated post to come once I actually tackle the build. But I’d like to now take a different tack in wrapping up this post. Were I to acquire another AMD Threadripper 3960X, this one factory-sealed and brand new, it’d conservatively set me back more than $2,000. What else could I do with the $1,500-plus in savings between that and the Intel alternative I bought? More accurately, what have I done?

Long-time readers may recall that mid-last year I mentioned I’d assembled several computers for subsequent donation to Evergreen Christian Outreach (EChO), a local charity that does great work for the financially and otherwise disadvantaged in the community. The systems worked fine, but since they were based on legacy CPUs and TPU-less motherboards, they won’t be upgradeable to Windows 11. I’ve once again done the build-and-donate thing for EChO, in advance of the back-to-school season (with the aspiration to assist kids in low-income families). This time I built three systems (with five more to follow them, ahead of the holiday shopping season…but I’m getting ahead of myself), all Windows 11-compatible. The bill of materials cost was less than $500 per system, translating to well under $1,500 total for the entire project. I’ll tell you more in a blog post series to come. Until then, I welcome your comments!


This article was originally published on EDN.

Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.


Leave a comment