Last week we put together a compact little PC to test out the integrated graphics on the Ryzen 5 2400G, but this Metalfish case does have just enough space to squeeze in a compact graphics card, and I do hate to leave a hole unfilled. The question is, what the best choice to plug this gap, and is it really worth doing?
Welcome to Esk Computers, I’m Scott and today on the desk of Esk, let’s answer that, shall we?
The sales blurb for the case intimates that this slot should be used to slap in a PCIe to Oculink card, for using with an external GPU box. Maybe there’s a fringe case where than makes sense, but I expect most people would be better served buying a slightly bigger case that can hold a full sized graphics card, like this barely bigger case also from Metalfish, the T60, used in a few builds prior on this channel.
For the same reasons, trying to find a suitable graphics card here is a bit of a fool’s errand, but viewers, I am ever your fool, particularly for oddball tech challenges. So, the constraints we’re working to is a single slot graphics card, which vastly limits our contenders, a half height restriction, which further limits our contenders, and a length limit of 210mm, which, say it with me, limits our contenders, that’s right, well done, have a cookie.
Trawling the web does give us a few options. I expect the single most powerful card that fits the bill is the RTX A1000, but the cheapest I’ve seen that going for lately is £250 and it’s more commonly north of £350, so, well, no. Too rich for my blood.
I’ve picked up this Sparkle Intel Arc A310 ECO for £83 a little while ago for another project, which I’m sure I’ll bore you with soon, but this would struggle without Resizable Bar support, so isn’t a great fit for this system.
AMD’s Radeon RX 6400 or the workstation version the Radeon Pro W6400 might be a worthwhile upgrade even if it’s going to be performance limited by the older PCIe generation on the 2400G. However the lowest price I’ve seen for these is around £150, which is more than I’d like to spend, well, at all, let alone on something that’s going to be bottlenecked out of the gate.
There are some single slot variants of the RTX 3050 6GB which should be a great contender, but they’re rare enough that the cheapest I’ve seen them is about £200, and more normally £300+. Again, too rich for my blood.
There’s a suitable single slot variant of AMDs RX550, but this is also somehow going for about £150 and that seems like a bad deal price to performance wise. On a good day you might find it for half that, but even so, it’s not likely to be much faster than the 2400G’s integrated graphics, and that’s also the case for a number of older generation cards.
So, after discounting that lot, we’re left with two contenders, both from Team Green, the Quadro P1000 and it’s successor the T1000. I had settled on this 4GB P1000, which I’ve picked up on eBay for about £45, because it seems like it should be a halfway decent performer at a reasonable price.
The P1000 came out in early 2017, using the same GP107 die found in the 1050 and 1050Ti, albeit with a trimmed shader count of 512 vs the full fat 768 units, and it ought to be running at 1354 MHz, and has 4GB of GDDR5.
Its successor, the T1000, appeared in 2021, and similarly it uses the TU117 die from the 1650 and 1650Ti, again cut down to 896 shader units over the full 1024 of the 1650Ti, with a base clock of 1065MHz. Both this and the P1000 have a max power draw of roughly 50W, so no separate power connection needed.
However, the T1000 seemed to command too much of a price premium, particularly the 8GB GDDR6 variant which is another of the barely justifiable £300+ club, and the 4GB variant is often going for north of £150. I had discounted it, but when one 8GB card popped up for about £100, I simply could not resist the prospect of deathmatching them against each other.
The elephant in the room here is the temperatures, as there’s a distinct lack of airflow in this case. We’ll come back to that, but let’s see what we’re getting as we compare the hell out of them, and the Vega 11 integrated graphics in the 2400G.
Taking a look at the synthetics first, in 3DMark’s Fire Strike the P1000 gets 5107 points, a 71% gain on the integrated graphics 2990 result. The T1000 gives us 7333 points, which is 44% more than the P1000 result and a 145% gain on the integrated graphics.
Moving over to Time Spy, the P1000 gives us 1732 points, 40% more than the base 2400G result of 1238. The T1000 gives us 2936 points, that’s a 70% jump from the P1000 and 137% from the integrated graphics result.
Superposition at 1080p High Settings next, and the P1000 gives us 2701 points, which is a 81% gain from the 2400G result. The T1000 boosts that to 3960 points, a 47% jump from the P1000 and 166% more than the integrated graphics alone.
Shadow of the Tomb Raider’s built in benchmark at 1080p High Settings gave us 15 average fps with integrated graphics alone. The P1000 increased that by 73% to 26 fps, and the T1000 increased that number by a further 73% to 45 fps. That’s a 200% improvement over the Vega 11 base result.
Counter-Strike 2 at 1080p Very High Settings in the fps Heaven benchmark gave 25 fps with the 2400G alone. Adding the P1000 takes that up 92% to 48 fps, and the T1000 increased that another 42% to a very nearly nice 68 fps. That’s 172% more than the base figure.
Moving down to 1080p Competitive settings, the P1000 gives us 123 fps, that’s 64% more than the 2400G’s 75 fps. Oddly not so much of a gain going to the T1000, it only managing a 15% improvement to 141 fps. Still, that’s an 88% improvement over the integrated graphics.
Cyberpunk 2077’s benchmark at 1080p Ultra settings is a bit beyond all of the setups, although for science, the P1000 gives 13.5 fps, which is 35% more than the 2400G’s 10 fps. The T1000 gets a sizeable 93% uplift to 26 fps, which is 160% more than the 2400G baseline, but still, not something you’d want to play a game like this at.
Dropping down to 720p Low Settings, and the benchmark gives us 31 fps with the 2400G, increasing only 26% to 39 fps with the P1000. There’s a 72% jump from that to 67 fps when moving to the T1000, an overall 116% increase from the Vega 11 figures.
Testing some actual Cyberpunk gameplay with a run around Kubuki at that 720p Low Setting but with High crowd density, the integrated graphics alone gave 23 fps, and the P1000 increased this by 42% to 32.6 fps. The T1000 could only give a modest 16% uplift to 37.8 fps, that’s a mere 64% more than the base 2400G alone could give. Looking at the CPU usage figures reveals that there’s a bottleneck there, so certainly the T1000 has more to give here.
Black Myth Wukong’s Benchmark at 1080p High Settings next, no framegen or upscaling, and oddly, there’s no improvement over the 2400G’s 6 fps by moving to the P1000. There’s a massive 117% boost by moving to the T1000, although that only gets us to 13 fps. Out of interest, dropping to 720p, adding in 66% FSR upscaling and framegen gets us just about playable results out of the P1000 at 47 fps, and a certainly playable 83 fps on the T1000.
More Unreal Engine 5 shenanigans next with Marvel Rivals at 1080p Native resolution, Medium settings, where the in-built benchmark gets 6 fps on the Vega 11, the P1000 increasing that 33% to a whopping 8 fps, and the T1000 giving a massive 188% boost to an almost cinematic 23fps, 283% more than the base result.
We get closer to playability by dropping to 720p Low Settings and adding performance FSR and framegen, at least with the P1000’s 54 fps and the T1000’s 65 fps, although personally I don’t rate frame generation as being all that great an experience on multiplayer shooters.
Assassin’s Creed Mirage benchmark at 1080p High Settings gave 27 fps on the 2400G alone, and the P1000 could increase that to 42 fps, a 56% improvement. The T1000 gave a further 36% boost to 57 fps, for a total improvement of 111% over the Vega 11 result.
Horizon Zero Dawn at the 1080p Favour Performance Settings saw some decent gains, from the clearly not playable 19 fps of the Vega 11 baseline, up 84% to 35 fps with the P1000, then up another 63% to 57 fps with the T1000, a total improvement of 200%.
Total War: Warhammer III at 1080p Ultra Settings next, and using the inbuilt Mirrors of Madness map benchmark gave 8 fps on the 2400G. There’s a 50% boost to 12 fps by changing to the P1000, and moving to the T1000 gives another 33% improvement to 16 fps. Not quite where I’d hope to be, even if that is a 100% improvement over the Vega 11.
Dropping to 720p Low settings, and we might be getting closer to playable levels, maybe not with the integrated graphics results of 26 fps, but there’s a welcome 100% boost to 52 fps by dropping in the P1000, and further 60% goosing to 83 fps with the T1000, a total improvement of 219% over the 2400G alone.
Civilization VI’s Gathering Storm AI Benchmark at 1080p highest settings gets a bit of a boost by letting the CPU concentrate on the game logic, it seems, with the integrated graphics setting a baseline turn time of 72 seconds that speeds up to 63 seconds when using the P1000, 13% faster. However there’s only a further 3% drop to 61 seconds when moving to the T1000, so that’s a mere 15% total improvement over the 2400G.
We”ll round off with a trip to Baldur’s Gate 3, wandering around the Lower City in Act 3 at 1080p Ultra Settings giving a 7.9 fps baseline with the integrated graphics, improving 87% to 14.8 fps with the P1000, and a further 89% jump to with the T1000, a total improvement of 254%.
Dropping the settings to 720p Low, and the 2400G sets a baseline of 19 fps which the P1000 improves on by 72% to 32.7 fps. There’s only a modest 28% improvement to 42 fps when moving to the T1000, and like Cyberpunk, we’re being bottlenecked by the CPU here. Better to stay with 1080p Ultra in my opinion.
The problem with this specific set of data is that both cards are putting out more heat than this case can handle. That’s not a fault of the case, it’s simply not sanctioned for the type of buffoonery I’m inflicting on it. As a result both are thermally throttling themselves to about 80°C, leaving performance on the table. I think I can improve this by bodging a little Noctua fan in here, but that’s another day’s video.
So, with the proviso that this wasn’t a sensible idea in the first place, if you find yourself in this situation perhaps by having a similarly specced and restricted office PC that you’d like to game on, what’s the best choice?
Well, just taking it on face value, there’s a few occasions where the T1000 is being held back by the 2400G, and that’s only going to get worse if it’s properly cooled. I’m reluctant to recommend a system that unbalanced, but in this specific system it’s the difference between playable 1080p results or having to drop more often than not to 720p to get playable results out of the P1000.
The value propositions of both cards are a bit weird, as they’re skewed by having to meet this form factor. If you have the space, a full sized equivalent graphics card will be faster and deal with heat better, for example the GTX 1660 6GB I’d used in a previous build for £73 is a much better gaming proposition than either of these.
For outright gaming, this build isn’t what I’d recommend. Either get a full sized, better performing PC rig for the same price if you have the space for it, or if space is really that tight and you don’t need Windows, pick up a Playstation 4 for £120 and put the rest of the budget towards a basic laptop rather than spend around £300 on this.
For the moment I’m leaving the T1000 in here, but probably just until I find a better use for it or improve the cooling situation. I don’t really want this running at 80°C continually. Not the most successful experiment, I suppose, but we learn as much from the failures as the successes. This case is probably best left without cramming a graphics card in here, so I’ll keep an eye out for more powerful APUs to test.
If you have any questions or want further details please leave a comment down below, and if you enjoyed this videotronic missive then consider subscribing. Until next time, take care of yourself, and each other.