Running A Modern Graphics Card In A 33 MHz PCI Slot

If you ever looked at a PCI to PCIe x16 adapter and wondered what’d happen if you were to stick a modern PCIe GPU in it, the answer apparently is ‘it works’ according to an attempt by [Circuit Rewind]. As long as you accept needing to supply external power with even a low-end GT 1030 card – as the PCI slot cannot provide enough power – and being limited to a single PCIe lane. This latter point isn’t so much of an issue as a single PCIe lane offers more bandwidth than the (shared) PCI bus anyway.

Despite the somewhat improvised setup, the GT 1030 card provided a decent 1080p experience in a range of games, after removing half of the 8 GB of system RAM before the configuration would work, probably due to VRAM mapping issues. Since the mainboard used also offered PCIe, the same card was run in a PCIe x4 slot, as well as in an x1 configuration, both with noticeably higher performance and putting the ‘why’ in ‘try’.

Perhaps unsurprisingly, a RTX 3080 also booted fine with external power and only 4 GB system RAM installed. Despite the PCIe x1 link, the system was able to finish a 3D benchmark and play Doom 2016, but with only 4 GB of system RAM and an old Athlon quad-core CPU, it was a terrible experience. Perhaps the most fascinating lesson to learn from this is that PCI and PCIe are amazingly compatible with only a simple translation bridge, even if high-performance graphics aren’t quite what PCI was meant for. After all, that’s why we got cursed with AGP for many years.

Continue reading “Running A Modern Graphics Card In A 33 MHz PCI Slot”

PCIe For Hackers: Our M.2 Card Is Done

We’ve started designing a PCIe card last week, an adapter from M.2 E-key to E-key, that adds an extra link to the E-key slot it carries – useful for fully utilizing a few rare but fancy E-key cards. By now, the schematic is done, the component placement has been figured out, and we only need to route the differential pairs – should be simple, right? Buckle up.

Getting Diffpairs Done

PCIe needs TX pairs connected to RX on another end, like UART – and this is non-negotiable. Connectors will use host-side naming, and vice-versa. As the diagram demonstrates, we connect the socket’s TX to chip’s RX and vice-versa; if we ever get confused, the laptop schematic is there to help us make things clear. To sum up, we only need to flip the names on the link coming to the PCIe switch, since the PCIe switch acts as a device on the card; the two links from the switch go to the E-key socket, and for that socket’s purposes, the PCIe switch acts as a host.

While initially routing this board, I absolutely forgot about one more important thing for PCIe – series capacitors on every data pair, on the host TX side of the link. We need three capacitor pairs here – on TX of the PCIe switch uplink, and two pairs on TX side of the switch – again, naming is host-side. I only remembered this after having finished routing all the diffpairs, and, after a bit of deliberation, I decided that this is my chance to try 0201 capacitors. For that, I took the footprints from [Christoph]‘s wonderful project, called “Effect of moon phase on tombstoning” – with such a name, these footprints have got to be good.

We’ve talked about differential pair calculations before in one of the PCIe articles, and there was a demo video too! That said, let’s repeat the calculations on this one – I’ll show how to get from “PCB fab website information” to “proper width and clearance diffpairs”, with a few fun shortcuts. Our setup is, once again, having signals on outer layers, referenced to the ground layer right below them. I, sadly, don’t yet understand how to calculate differential impedance for signal layers sandwiched between two ground planes, which is to say – if there’s any commenters willing to share this knowledge, I’d appreciate your input tremendously! For now, I don’t see that there’d be a tangible benefit to such an arrangement, anyway.

Continue reading “PCIe For Hackers: Our M.2 Card Is Done”

PCIe For Hackers: An M.2 Card Journey

I’ve designed a few M.2 adapters for my own and my friends’ use, and having found those designs online, people have asked me for custom-made adapters. One of these requests is quite specific – an adapter that adds one more PCIe link to an E-key M.2 slot, the kind of slot you will see used in laptops for WiFi cards.

See, the M.2 specification allows two separate PCIe links connected to the E-key slot; however, no WiFi cards use this apart from some really old WiGig-capable ones, and manufacturers have long given up on connecting a second link. Nevertheless, there are some cards like the Google Coral M.2 E-key dual AI accelerator and the recently announced uSDR, that do indeed require the second link – otherwise, only half of their capacity is available.

It’s not clear why both Google and WaveletSDR designed for a dual-link E-key socket, since those are a rare occurrence; for the Google card, there are plenty of people complaining that the board they bought just doesn’t fully work. In theory, all you need to do to help such a situation, is getting a second PCIe link from somewhere, then wiring it up to the socket – and a perfect way to do it is to get a PCIe switch chip. You will lose out on some bandwidth because the uplink PCIe connection of the switch can only go so fast; for things like this AI accelerator, it’s not much of a problem since the main point is to get the second device accessible. For the aforementioned SDR, it might turn out useless, or you might win some but lose some – can’t know until you try! Continue reading “PCIe For Hackers: An M.2 Card Journey”

Picture of the PCIce card with a fan attached

Server Network Cards Made Extra Cool

Using cheap and powerful server expansion cards in your desktop builds is a tempting option for many hackers. Of course, they don’t always fit mechanically or work perfectly; for instance, some server-purpose cards are designed for intense amounts of cooling that servers come with, and will overheat inside a relatively calm desktop case. Having encountered such a network card, [Chris] has developed and brought us the PCIce – a PCIe card that’s a holder and a controller for a 80mm fan.

The card gets fan 12V from the PCIe slot, and there’s an ATTiny to control the fan’s speed, letting you cycle through speeds with a single button press and displaying the current speed through LEDs. There’s a great amount of polish put into this card – from making it mechanically feature-complete with all the fancy fasteners, to longevity-oriented firmware that even makes sure to notice if the EEPROM-stored settings ever get corrupted. At the moment, the schematics and the ATTiny firmware are open-source, [Chris] has promised to publish hardware files after polishing them, and has also manufactured a batch of PCIce cards for sale.

When it comes to making use of cheap server-purpose cards, a cooling solution is good to see – we’ve generally seen adapters from proprietary form-factors, like this FlexLOM adapter from [TobleMiner] to make use of cheap high-throughput network cards with slightly differing mechanical dimensions and pinouts. Every batch of decommissioned server cards has some potential with only a slight hitch or two, and it’s reassuring to see hackers make their eBay finds really work for them.

PCIe For Hackers: Extracting The Most

So, you now know the basics of approaching PCIe, and perhaps you have a PCIe-related goal in mind. Maybe you want to equip a single-board computer of yours with a bunch of cheap yet powerful PCIe WiFi cards for wardriving, perhaps add a second NVMe SSD to your laptop instead of that Ethernet controller you never use, or maybe, add a full-size GPU to your Raspberry Pi 4 through a nifty adapter. Whatever you want to do – let’s make sure there isn’t an area of PCIe that you aren’t familiar of.

Splitting A PCIe Port

You might have heard the term “bifurcation” if you’ve been around PCIe, especially in mining or PC tinkering communities. This is splitting a PCIe slot into multiple PCIe links, and as you can imagine, it’s quite tasty of a feature for hackers; you don’t need any extra hardware, really, all you need is to add a buffer for REFCLK. See, it’s still needed by every single extra port you get – but you can’t physically just pull the same clock diffpair to all the slots at once, since that will result in stubs and, consequently, signal reflections; a REFCLK buffer chip takes the clock from the host and produces a number of identical copies of the REFCLK signal that you then pull standalone. You might have seen x16 to four NVMe slot cards online – invariably, somewhere in the corner of the card, you can spot the REFCLK buffer chip. In a perfect scenario, this is all you need to get more PCIe out of your PCIe.

Continue reading “PCIe For Hackers: Extracting The Most”

PCIe For Hackers: Link Anatomy

Last time, we looked over diffpairs, their basics, routing rules and the notorious tolerances of PCIe when it comes to diffpairs. Now, let’s take a look at the exact signals that make PCIe tick, as well as give you an overview of which sockets you can get PCIe on.

I separate PCIe sockets into two categories – desktop sockets, which are the usual 1x, 16x, or perhaps x4 PCIe sockets we see on desktop motherboards — and mobile sockets: mPCIe and M.2. There are also sockets like ExpressCard that are still found on some older laptops, but they have definitely fallen out of relevance. On mobile, M.2 is the way forward when it comes to PCIe – if you’re interested, I’ve written a short series talking about everything M.2 before.

On these sockets, most signals will be optional and some signals will be socket-specific, but there are some signals required in any PCIe device. Of course, the first group of required signals is the PCIe diffpairs themselves.

Continue reading “PCIe For Hackers: Link Anatomy”

PCIe For Hackers: The Diffpair Prelude

PCIe, also known as PCI-Express, is a highly powerful interface. So let’s see what it takes to hack on something that powerful. PCIe is be a bit intimidating at first, however it is reasonably simple to start building PCIe stuff, and the interface is quite resilient for hobbyist-level technology. There will come a time when we want to use a PCIe chip in our designs, or perhaps, make use of the PCIe connection available on a certain Compute Module, and it’s good to make sure that we’re ready for that.

PCIe is everywhere now. Every modern computer has a bunch of PCIe devices performing crucial functions, and even iPhones use PCIe internally to connect the CPU with the flash and WiFi chips. You can get all kinds of PCIe devices: Ethernet controllers, high-throughput WiFi cards, graphics, and all the cheap NVMe drives that gladly provide you with heaps of storage when connected over PCIe. If you’re hacking on a laptop or a single-board computer and you’d like to add a PCIe device, you can get some PCIe from one of the PCIe-carrying sockets, or just tap into an existing PCIe link if there’s no socket to connect to. It’s been two decades since we’ve started getting PCIe devices – now, PCIe is on its 5.0 revision, and it’s clear that it’s here to stay.

Continue reading “PCIe For Hackers: The Diffpair Prelude”