Original Xbox Gets Hardware Transplant, And Is Very Fast

The original Xbox launched way back in 2001, to much fanfare. This was Microsoft’s big first entry into the console market, with a machine packing a Pentium III CPU, and commodity PC hardware, contributing significantly to its bulk. Modding was a major part of the early Xbox scene, and as the original hardware has grown too feeble to keep up with modern tasks, enterprising makers have instead turned to packing the black box with modern hardware. The team at [Linus Tech Tips] decided that other builds out there weren’t serious enough, and decided to take things up a notch.

The build starts with a passively cooled compact power supply, a Core i5 8400 6-core CPU, and a GeForce RTX2070 to handle graphical tasks. Parts were carefully selected for a combination of performance, packaging, and with an eye to the thermal limits inherent in stuffing high-powered modern hardware into a tight Xbox shell.

All manner of oddball techniques are used to make the build happen. The GPU is connected through a PCI Express cable, which we were surprised to learn was a thing, given the nature of high-speed signals and long transmission lines. The Xbox shell had its original metal insert and plastic standoffs removed, with an aluminium inner shell being CNC cut and bent up on a pan brake to act as a new internal chassis. There’s yet more carnage to come, as the GPU has its extraneous DVI port hacked off with a grinding wheel.

In the end, after much cutting and cajoling, the parts come together and fit inside the case, making the sleeper build a reality. It’s fun to watch the team fiddle with config files and struggle to load and play local multiplayer games, as they realise that there are just some things that consoles do better.

Regardless, it’s an impressive casemod that goes to show what you can pull off with some off-the-shelf parts, a well-stocked workshop, and some ingenuity. If you’re looking for more case mod inspiration, try out this all-in-one printer build. Video after the break.

[Thanks to Keith O for the tip!]


54 thoughts on “Original Xbox Gets Hardware Transplant, And Is Very Fast

  1. Most of what I read on this site is very impressive; this is not… It’s not an Xbox – it is a PC shoved inside an Xbox case. The title is misleading : if it were upgraded hardware that run and operated like an xbox then the title would be correct, but it is not that at all.

      1. Completely agree. I have an dual core atom board that I shoved into an old xbox case that now acts as a music server. So plugging new hardware into an old case is nothing new. It is impressive that they manged to get all that in their without anything melting, but that is the extent that I am impressed.

    1. I think all the xbox performance upgrades have been done and well documented, about the most you can get out of one with a stock board is a 1.2ghz pIII and 128 megs of ram

      still thanks to the decade of homebrew software development and the stout hardware in a og xbox whats available for it is really impressive, my stock board with a softmod and fat hard disk easily runs stuff that things like a recent gen PI has trouble with

      1. 1.2ghz pIII and 128 megs of ram those were the AWSOME upgrades. I never got to that point, all other stuff was really really fun. The Original Xbox still has a special place in my heart.

        The stuff in the article has almost nothing to do with an Xbox. If I was to do such a thing I put my intel NUC inside an empty case I have laying around here somewhere and be done within 15 minutes.

      1. There’s also CPU swaps for higher clock speeds (up to 1.4HGz iirc). You run into major issues with a lot of retail games though since they obviously weren’t designed to cope with anthing but the stock speed.

    2. Because it can’t be done, the chips were standard but the package wasn’t.
      There aren’t any other PIII’s with a BGA package, so unless you want to clip pins, grind the nubs flat, then hollow out divots in the flats to settle balls in before you even begin to even think about reballing the processor ALL for an extra 250 MHz (the difference between the top of the line 1 GHz PIII coppermine and the base XBox processor) be my guest but if that sounds like too much hassle for something that was essentially already a desktop computer stuffed awkwardly in a case with an RF modulator and some proprietary USB connectors slapped on the front plugged into some modded Dreamcast controllers (No really) then welcome to the wonderful world of casemods and emulation.

      Seriously though, PIII coppermine’s are from the bad old days of overclocking, don’t believe me belive Tom https://www.tomshardware.com/reviews/performance-guide,189-4.html

  2. “The GPU is connected through a PCI Express cable, which we were surprised to learn was a thing, given the nature of high-speed signals and long transmission lines.”
    The advantages of differential pair transmission lines.
    These cables have been around for a while. What I’m curious about is how many lanes trained at at what speed.

      1. As long at the PCIe registers say that’s what the card trained at :) I know absurd things can be done with PCIe because the electrical standard revolves around the ‘eye diagram’ for the signal. As long as that stays within spec, engineers can go a little crazy.

      1. This is pretty amazing, what a weird experiment. Good to know in case I ever need to make a very strangely-shaped custom PC.

        God I sure fucking hate the faces literally everybody makes in youtube thumbnails, though. Is that really necessary? Do the dog-ass stupid faces really drive engagement so well that everyone must do them? Will youtube ever escape a world full of lazybearded young men making a face like they’ve just received a heroin enema? The guy in the video in the original article is doing it too. Why?

        1. If it didn’t work for revenue, do you think people would do that?

          The rest of the video (I saw it the day it was released, it’s pretty popular, too) is relatively coherent. If you don’t like it, I’m sorry, you’ll have to go use a different platform.

          Speaking of which, LinusTechTips has another platform, called Floatplane. It’s subscription-based, but the content is about the same (there are exclusives) and the thumbnails are much better. Are you on it?

        2. The reality is that they have to make their video as interesting as possible for people to click on it. That means not only giving the videos clickbait names, but clickbait thumbnail images. I’m not sure who gets excited by these weird open mouth silent scream faces, but apparently enough YouTube users like them that they have a noticeable impact on the viewership numbers.

          I’ve often heard that it boils down the to fact that the most active users on YouTube are children, which supposedly these over the top faces appeal to. But you wouldn’t think this kind of content would be targeted to kids, so who knows.

        3. A bunch of years ago, I saw some embedded equipment that used a card plugged into a PCIe slot to pass the connector through to a plug on the card (it was a passive device IIRC). It was used to connect to a second chassis with PICe slots in it or in one case, to actually connect 2 systems together via a system called non-transparent bridging.

          But chaining regular old extensions together and getting that distance? That’s neat.

        4. There are also theories that the youtube algorithm looks for faces in thumbnails and promotes those, but this is unproven. However, in another video Linus actually digs into why he does these thumbnails, and how they did A/B testing as much as they could and discovered that they got a significant fewer number of views without these types of thumbnails. He even mentioned that he would prefer not to make them (and explain to his kids why he looks silly) but the revenue differences meant that he could give his employees more secure jobs and that made it worth it.

      2. I’m curious as to where the link would have trained down to a lower speed and/or fewer links. It’s totally academic and dependent on the devices on either end, but that is the nature of curiosity :)

      3. So, is it the impedance discontinuation on every connector that limits it to 3m, or are there other transmission line effects/losses or signal delays that start to affect the PCIe performance? Can’t watch the video right now, but i’m wondering if they only did performance evaluation on an application level, or some measurements on the physical layer too.

    1. Yeah, these cables are a staple in the small form factor community. You aren’t going to see any performance issues. Hell, most cards and most applications won’t hit a bottleneck even cut down to four lanes, like when you use a thunderbolt to pcie adapter for an external gpu setup.

      The bandwidth on a pcie bus is absurdly high. If you’re hitting the cap swapping data between the cpu and gpu through it, you’re either doing something extremely intensive like ML or you’re doing something extremely wrong.

      Kinda surprised HaD hadn’t heard of them. I guess it’s more of a MacBook crowd? I dunno.

      1. That’s because most software expects the latency to be piss-poor and loads up all the data into the video ram regardless. When you run out of VRAM on the card and have to swap to main memory, fetch game textures on the go etc., that’s when things go south really fast.

        Then again, that’s what happens anyways.

        1. Point being that GPUs waste a ton of memory to gain access speed. Even with the GPU, you may have the same data copied in 3 – 4 different locations to bypass the DRAM timing considerations (CAS/RAS and precharge). That results in video cards needing gigabytes of memory to handle few hundred megabytes of actual data. If the system has to fall back to swapping data over PCI-e from running out of RAM, it has already failed and cannot operate as expected. These events have to be planned well in advance, and in that sense the link bandwidth really doesn’t matter – games won’t perform worse on 4x than 16x – only the loading times between levels will be a few milliseconds shorter.

  3. Fun little project, building a PC case, but I can’t stand watching this guy. He is marketed (or rather marketing himself) as the mega tech guy, but after a few videos you get the impression that he doesn’t know what he’s talking about half of the time…

    1. I have seen a video* of him explaining why smartphone get slower: this guy is a fraud and his voice is unbearable .

      I don’t understand how a man like Louis Rossmann featured with this Jacky, ok it made his channel’s counter explode, all was not lost.

      HaD, please avoid. I am not against case modding but not from a freaking PewDiePie/shitty/youtuber


      *I was testing a service “neverthinks”(completely matching name) a kind of curated video platform and I discovered this ridiculous video of Linus about smartphones. To throw away.

  4. I watched a couple of Linus’s videos (YouTube recommendation) and was… let’s say unimpressed. I was surprised to see him featured here and decided to read the comments to get a feel of what the general consensus was of his work among the rest of the Hack-a-day community. Judging by the above comments, I am not an outlier.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

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