I review CPUs for 3DTested. I don’t review products in a dozen different categories, cosplaying as some expert in each of them; I review CPUs. As you can imagine, my day is filled with cleaning TIM, unmounting coolers, swapping CPUs, and darting back and forth across different platforms to gather the data I need for reviews, comparisons, and just about anything else in the world of CPUs that’s worth taking a closer look at.
We do a lot of benchmarking here at 3DTested, but CPUs and GPUs require even more than other categories. A full pass of our test suite can take upwards of 10 hours for a single chip, and that doesn’t include the time it takes to swap chips and get benches prepped for another pass. It should go without saying, but efficiency is king for benchmarking CPUs here. We don’t want to (and won’t) sacrifice the abundance of data we gather for reviews, even when going across a dozen or more SKUs for a single review.
I’m not focused on the hardware itself inside the benches. You can find that detailed in any of our CPU reviews. Given the catch-22 of PC hardware pricing right now, it’s hard to justify a new build anyway. Instead, I’m focusing on the tools I use nearly every day to streamline CPU testing, including some tips I’ve learned after years of flipping chips.
Alamengda Open Computer Case
The best test bench around is probably the Open Benchtable, with the Praxis WetBench and newer Thermal Grizzly Der8enchtable coming in close behind. The problem is that they’re ridiculously expensive. The Open Benchtable is $200. The WetBench reaches all the way to $300 for the SX version, and the Der8enchtable is $270 — though, admittedly, the Der8enchtable does a lot more with its active PCB. I have three benches running at any given time, so shelling out $600 to $900 just wasn’t an option. Let me introduce you to the Alamengda Open Computer Case, which is around $25.
It may seem innocuous, but finding a test bench with this layout at this price is surprisingly difficult. For starters, it’s aluminum. There’s no shortage of cheap acrylic benches you can pick up on AliExpress or any other provider of direct Chinese goods, but I wanted aluminum. It’s also flat, while many similar inexpensive benches are set up so your motherboard is oriented vertically.
The critical perk of the Alamengda benches I have, however, is that there isn’t a shroud for the PCH. Several flat, inexpensive benches have a single piece of aluminum for the PCIe brackets, PCH, and power supply, such as this one and this one. During benchmarking, I have a single 120mm fan sitting on top of the PCH to help keep everything cool, including the NVMe SSD I have installed. These Alamengda benches allow me to easily place a fan on top of the PCH, as well as move it out of the way when I’m flipping chips.
I’m using the ATX version, but Alamengda offers E-ATX and ITX variants, as well. For the ATX version, you get an optional metal rail for installing up to three 120mm fans (or up to a 360mm radiator), as well as a bottom bracket that can house two 2.5-inch drives. There isn’t space for a 3.5-inch drive. In addition, the bench comes with feet so it’s slightly elevated off your desk, as well as a metal carrying handle if you need to move your test rig around.
Arctic MX Cleaner wipes and MX-4 TIM
I’ve kept a bottle of 97% isopropyl alcohol and a stack of coffee filters on my desk for years, but I recently discovered Arctic’s MX Cleaner wipes, and I’m never looking back. After a day of benchmarking, my hands are covered in TIM, and usually my keyboard, mouse, and anything else I happen to touch has some stray splotches, as well. Arctic’s MX Cleaner wipes solve that problem because they’re specifically made for cleaning TIM.
In other words, they aren’t alcohol wipes. Instead, the wipes are soaked in Limonene, an organic compound often extracted from citrus rinds and used widely in essential oils. Arctic has sold a similar cleaning kit in non-wipe form under the ArctiClean name for years, but ripping a wipe open and swiping over a chip is much easier for my purposes.
These wipes have quickly become a staple for me when benchmarking because they work in a single swipe. The TIM almost sticks to the wipe wherever you apply pressure, so you don’t get the smearing you typically get when using isopropyl and coffee filters. They keep my hands clean, so everything in my office doesn’t become a TIM deposit point after I’m done benchmarking for a review.
Although I love these wipes, they aren’t for everyone. First, the smell. I personally like the smell, which leaves a lingering lemon scent in the air even after you’ve discarded the wipe. Reading around online, there are some folks who hate the smell. It’s not as pungent as I’ve seen many commenters suggest, but it certainly lingers for a long time. If you’re worried about the smell, I recommend picking up something like this Arctic MX-6 kit, which includes a 4g tube of TIM and six wipes before buying the 40-pack (which itself is only $9).
Additionally, these wipes don’t use alcohol, so the liquid doesn’t evaporate in seconds, like 90%+ isopropyl does. I still keep some coffee filters on-hand in case any moisture is still present while I’m flipping chips. It’s gone in about a minute, so I don’t normally run into issues, but it’s something to keep in mind.
And the TIM I’m cleaning up is Arctic’s MX-4. It’s an oldie at this point, previously ranking among the best thermal paste before being bumped by the newer MX-6. But our testing shows that MX-4 is able to maintain a temperature of 63.3 degrees Celsius on the Core i9-14900K with a 360mm AIO and 300W power limit in Cinebench R23 nT. That’s 0.6 degrees off MX-6, 1.5 degrees off of Noctua NT-H2, and less than a degree off Thermal Grizzly Kryonaut. In other words, it’s still great TIM, despite hitting the market more than 15 years ago.
Performance always comes first, but I use MX-4 over other, newer TIM because you can buy a boatload of it at once. Arctic offers MX-4 in a 45-gram tube for about $20. A 1g tube of Thermal Grizzly Kryonaut will run you $9 (11g comes in at $32), while a 3.5g tube of Noctua NT-H2 clocks in at $15. TIM like MX-4 usually isn’t solid in higher capacities like 45 grams, forcing you to either buy several smaller tubes or settle for TIM that’s questionable at best.
Hoto Electric Screwdriver Kit
Regular 3DTested readers will have undoubtedly heard multiple staffers gush about the Hoto NEX O1 Pro, which Hoto now offers as the “3.6V Electric Screwdriver Kit Pro,” but regardless of what you call it, it’s incredible. It’s an electric screwdriver with a 1500mAh battery and USB-C charging, and it includes 25 S2 bits (a hardened steel alloy), including Philips, Torx, flat, and hexagon bits. The main PH2 (Philips) bit is also magnetic, which is essential for any PC building-related task.
For my purposes, there are two things that stand out about Hoto’s screwdriver. First, the ring light. Around the bit, there’s a faint light that glows when you’re using the screwdriver. It’s fairly dim, but when you’re reaching your hand deep into a rig trying to get that last motherboard screw by the EPS connectors, you’ll be thankful you have the extra bit of light.
In addition, Hoto includes a sensor in the screwdriver that stops the head when you release the button. That may sound small if you haven’t used a lot of electric screw drivers, but it isn’t. Most electric screwdrivers have a small window after you’ve released the button where the bit will continue to spin, making it very easy to strip the often tiny screws involved in building a PC. With the Hoto screwdriver, the moment I meet resistance, I let go of the button, so there are no stripped screws.
My only issue with the kit is the bits themselves. They have a hardness of 60HRC on the Rockwell scale from the S2 steel, so they’re certainly quality bits. Only the long PH2 bit is magnetic, however. The other bits aren’t, or at the very least, the magnetism is so weak that it can’t even pick up a #6-32 screw. Thankfully, Hoto uses a magnetic 1/4-inch bit connection, so you can throw in just about any bit you want.
A long screwdriver (or a power button)
In one of my first meetings with 3DTested Editor-in-Chief (and former CPU reviewer) Paul Alcorn, he pulled out a long, Philips-head screwdriver and said, “This thing is awesome.” I turned around, dug around in a shelf under my desk, and pulled out a similarly long, Philips-head screwdriver, and responded, “I know.” It was a clear “if you know, you know” situation, and it retroactively justified a decade of me using this ridiculous 12-inch screwdriver to turn on my test benches.
Something a bit cleaner (and a change I’ve recently made) is to invest in a cheap power button. “Invest” is overselling it a bit, though; the external power buttons I use for my test benches are $8 each. There are a few common variations of these buttons you’ll find, some of which use a mechanical keyboard switch and give you a reset button, and others that even work wirelessly. But the bog-standard button you can see below is what I use.
I switched over to a few of these buttons because I’ve centralized my test benches to a single monitor, keyboard, and mouse (more on that next), so I can place all the buttons right in front of my keyboard and easily turn machines on and off. I also sprung for this inexpensive Nelko label maker, so I know which machine I’m turning on. It’s less than $20 and connects to your phone via Bluetooth, and it’s been rock-solid so far.
◼️ UGreen HDMI KVM
My testing setup involves three rigs hooked up to the same monitor, using the same keyboard and mouse. I can swap between my rigs at the press of a button, leaving automated tests to run unattended on one machine while I run manual tests on another, and that’s all enabled by a KVM. With the advent of USB-C in monitors (including power delivery), you’ll find a ton of monitors with a KVM built in. I’m not using USB-C with my test benches, however, so I needed a dedicated, external KVM.
If you aren’t familiar, KVM stands for Keyboard, Video, Mouse, and it’s a device that reroutes your display and peripheral inputs to different outputs, essentially allowing you to use the same setup across multiple rigs without swapping over cables. The particular KVM I use is this four-port model from UGreen that you can see below. It’s an HDMI KVM, unfortunately, but it hits on a few key areas that are important for my testing.
First, it supports up to four PCs. Many KVMs only support two PCs, with more expensive versions supporting two PCs with multiple monitors. This one trades the support for extra monitors for more inputs. It also supports HDMI 2.1. I could settle for 4K@60Hz and use a cheaper KVM, but it’s jarring constantly swapping between 120Hz on my main workstation to 60Hz on my test benches.
Outside of the display, the KVM includes four 5Gbps USB ports, split across three Type-A connections and a single Type-C. You don’t need fast USB ports for a keyboard and mouse, but I like having a couple of extra ports, in the event that I need to plug in an external hard drive. If I install a new game for testing through Steam, for example, I can toggle over to the other machine and transfer the files off an external SSD, so I don’t have to wait through a download three times for each new title I want to test.
Still, it’s not a perfect solution. Critically, it doesn’t support EDID (Extended Display Identification Data) emulation. EDID emulation basically allows the KVM to spoof a video signal when you switch to a different input, so your resolution, refresh rate, and window position all remain the same when you’re swapping between inputs. I’ve been eyeing this four-port KVM from Angeet that has EDID emulation, so hopefully I’ll be able to move on from the UGreen KVM soon.
◼️ ElemorLabs PMD2
I don’t want to leave this article without talking about at least some hardware, so let me introduce you to the ElemorLabs PMD2. If you haven’t heard of it, it’s basically a power meter. For my purposes, I run both 8-pin EPS cables and the 24-pin ATX cable into the PMD2, and run the same cables out the other side into the machine. It intercepts the power traveling from the power supply to the motherboard to give you an accurate reading of how much power your PC is using.
It’s not just for CPUs — the PMD2 includes three 8-pin PCIe connectors and a single 12VHPWR (12V-2x6) connector, so you can measure PCIe power, as well. Once it’s set up, you run the USB-C output to your PC, and you can monitor and log your power usage with ElemorLabs’ own software. It supports integration with HWinfo, as well, which is where I actually do the logging for reviews.
Software can provide a good estimate of power consumption, but it’s not enough for a proper CPU review. For starters, software can have bugs; it just takes one bunk update to throw off hours of dedicated power testing. More importantly, however, is that software just provides an estimate. It’s seeing and reporting power usage after that power has already gone through several layers of regulation.
Voltage sensors on your motherboard, which are visible to monitoring software, offer an approximation of power consumption based on what the sensor sees. Other solutions, like a P3 Kill A Watt, can measure total power usage from the wall. The PMD2 does something different. It allows you to drill down on exactly how much power is traveling from the power supply to the board itself, and provides a more accurate representation of power consumption for CPUs and GPUs, in particular.
Follow 3DTested on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
