I’d picked up a Quadro P2200 as a candidate for using in my dual Xeon home server for transcoding media, but if you watched my previous video I found much less practical solutions for that. Still, it seems daft not to test it out before finding another home for it, so let’s give it a bit of a poke and see if it’s still alive. It’ll also let me trial run a format for testing older graphics cards that’s hopefully a bit more interesting than watching them produce slideshows in Unreal Engine 5 games.
Welcome to Esk Computers, I’m Scott and today on the desk of Esk, let’s be wildly unprofessional with professional graphics cards.
The P2200 released back in 2019. It uses the GP106 die used in the GTX1060, and has the same 1280 shading units, 40 texture mapping units, and a slightly cut down 40 ROPs compared to the 1060 6GB’s 48. This also loses 1GB of GDDR5 RAM, for a slightly odd total of 5GB. The main differences are clock speed and power draw, where the P2200 should be running at 1000MHz, boosting to a smidgen under 1500 MHz, compared to the 1500Mhz base and 2000MHz boost a 1060 normally clocks at.
Also, as this takes no supplemental power from the PSU it’s limited to 75W, where the 1060 can pull 120W. That’s probably why this card sits comfortably at 60-odd degrees under load, despite the compact single slot cooler.
Now, the Quadro cards are supposed to be used for boring adult activities like Computer Aided Design and the like, but to be honest I’m not quite sure how to test that, and there would be little point in doing so. If you’re a proper professional then I can’t imagine seven year old parts are still on your shopping list. I mean, I could run SPECviewperf, but I would have no frame of reference, like a child who wanders into the middle of a movie.
However I think media transcoding might be a reasonable use case for homelabbers, so let’s at least give that a poke before moving on to something more fun. Unlike the consumer branded cards, there’s no locked limit on the number of simultaneous hardware-accelerated transcode streams, and from poking around the internet it seems like it can handle at least 4 simultaneous 4K transcoded streams, or 20-odd 1080p streams.
I can’t think of a way to test that with the kit I have to hand, but we can look at the raw performance. I’ve got my 9800X3D / 7800XT system here with Handbrake, and let’s feed it this 4k webp version of Big Buck Bunny what I found floating around the internet and that. It’s a ten minute, thirty five second short.
I’m going to transcode this to 1080p h264 using the CPU at the Fast and Very Fast presets, the 7800XT with AMD’s hardware VCE and the P2200 with the NVENC.
I nearly aborted the CPU based encodes as despite it saying it was encoding at hundreds of fps it was estimating over a day to complete. Thankfully just a maths error, the fast preset completed in 310 seconds, the very fast in 183 seconds.
The AMD VCE encode took 109 seconds, and the NVENC encode took 118 seconds. So, the old NVIDIA card is two to three times faster than one of the faster CPUs currently available, and as fast as a AMD card many years newer than it. Not too shabby.
Doing the same for h265 gives a CPU fast result of 347 seconds, a CPU very fast result in 248 seconds, and on the hardware side AMDs VCE took 101 seconds, and the P2200’s NVENC took 110 seconds, so the same story there.
I should add that both the AMD VCE files were also about double the size of the CPU encodes, so I suppose that’s trading off faster encodes for more network bandwidth in a streaming context. In terms of quality, I cannot tell the difference between all bar one of these files when casually viewing. I’m sure pausing, pixel peeping and rewinding will find some differences and preferences, but not at a level my tired, old, ruined eyes can distinguish. The exception was the h265 NVENC encode, where the preset for whatever reason decided to crush it to half the file size of the CPU encodes, impressive but quite the hit to the quality. That’s nothing some tweaking wouldn’t fix.
The P2200 doesn’t have any support for the AV1 codec, so you’d need something newer for that, but that really seems to be the practical limiting factor for transcoding in all but the biggest households, this seems to be a very good card for h264 and h265 media distribution.
Moving on, and while this isn’t realistic, for the sake of argument, and testing the format, let’s just say you’d picked this up back in 2019 with the intent of gaming on it. How would it handle the most demanding games of the year, and at what point did it stop giving an acceptable result?
I suppose to answer that we need to define acceptable. Archives of the Steam hardware survey says you’d most likely be on a 1080p screen, and if we think of this as a slightly cut down, slower GTX1060, it’s a lower to mid end card so expecting all the bells and whistles is probably unreasonable. Let’s say you are a realist, and expect to run games at medium to low settings.
I’ve made a list of some of the most graphically demanding games for each year this millennium, so let’s run through that and see at what year it can no longer handle the top of the line games, assuming that your CPU is up to the task. Of course, there’s ways to extend the life of cards by dropping settings, resolutions, or playing the thousands of less demanding games that are released each year, but let’s see how the high end stuff goes.
I’ll pick three games from every year, and if in my mostly subjective opinion it’s struggling too much with the majority of them, either in framerate or coping with the artistic direction, we’ll declare it cooked. I’m open to feedback on alternative suggestions on the games, please sound off in the comments if you have some.
Given that it came out in the middle of 2019, it seems reasonable to give 2018’s games a butchers first. Shadow of the Tomb Raider at Low settings is getting 76 average fps, with 0.1% Lows of 61 fps, so a very playable result in this opening area. The in-built benchmark run at these settings gives 71 average fps, so that tallies up well.
Red Dead Redemption II at the higher end of the Favour Performance settings gets 44 average fps in the benchmark run, and much the same during the prologue gameplay with it getting 41 average fps with 0.1% lows of 35 fps. Ideally we’d want a bit more, but I was happy enough with this performance to give it a pass, and there’s still room to nudge the settings down a smidge.
Final Fantasy XV using the Average settings gets 72 average fps in this thrilling “pushing a car” sequence which reminds me why I stopped playing this type of game. When you finally get to, y’know, do something, it’s getting 61 average fps with 1% lows of 49, so no complaints with this performance. The 0.1% lows of 12 fps are on paper a cause for concern, but I didn’t notice it when playing.
Moving into 2019 with Control, which gives 61 average fps at Low settings. 1% lows were 48 fps, and although it says 0.1% lows of 5 fps, I didn’t feel that stutter in game, so passes muster. Knowing what it looks like at higher settings makes low a bit of an unpleasant visual shock, but it’s working just fine.
I doubt the initial section of Metro Exodus is the most demanding, but it’s 89 average fps at low settings gives it a bit of headroom. A slightly later section outside gets 82 average fps, with 1% lows were 59 fps, and 0.1% lows at 55 fps, but seems fine so far.
Star Wars: Fallen Order has Medium settings as its lowest setting, so why it’s not called low is beyond me. It’s getting 80 average fps, with 1% lows of 63, and 0.1% lows of 56. Plays great, at least in these opening sections. So, that’s another year passed, on to Futuristic Space Year 2020
Cyberpunk 2077 at Low settings gets an average of 56 fps in this run around Kubuki, with 1% lows of 42, and 0.1% lows of 26. This plays okay, but for me this is a bit closer to a twitch shooter than other open world games, so it’s maybe a little bit too uneven to fully enjoy. It’s a borderline pass.
Horizon Zero Dawn should probably be run at Favour Performance settings, where the inbuilt benchmark gets 72 fps. But it’s so ugly I simply can’t tolerate it, so Original settings it is, where the benchmark gets 63 fps. In gameplay it’s more like 51 average fps, with 1% lows of 40 fps and 0.1% lows of 29 fps. This plays okay in my estimation, and there’s always the lower setting if you prefer frames over beauty.
Watch Dogs: Legion at Low settings also takes a bit of a hit in the prettiness department, but can’t argue with an average framerate of 56 fps, 1% lows of 45 fps, and 0.1% lows of 42. That’s playable in my book.
Let’s see if the Quadro can survive 2021. Forza Horizon 5 is maybe a bit of a gimme for any halfway modern card. It’s a well optimised game, and here at Low settings it holds a constant 60 fps on the benchmark when VSync is enabled, or with an unlocked framerate, it gets an average framerate of 105 fps, with 1% lows of 76 fps and 0.1% lows of 69. Nice. I might replace this with Metro Exodus Enhanced or Battlefield 2042, let me know in the comments if that’s more sensible.
Icarus is a game I somehow own despite having never heard of it, to my recollection, until researching this list. Apparently I bought in last year’s Winter Steam sale for three quid. Must stop doing that. At Low settings, with FSR disabled, while this gets a halfway decent average of 48 fps, and 1% lows of 30, the 0.1% lows of 6 fps make this somehow feel much worse than, say Control, which posted similar numbers. Combined with low settings making this game quite aggressively ugly, I think this is the first outright no from me. I wouldn’t want to play this on this card.
I let Resident Evil Village auto select what it reckoned it’s best settings were, and it landed somewhere between low and medium, so let’s run with that. It’s getting averages of 68fps, 1% lows of 52 fps and 0.1% lows of 48 in the opening areas. This feels perfectly fine to play, so overall it’s survived another year. Onwards to 2022.
A Plague Tale: Requiem’s swarms of rodents are both gross and demanding, but even before we get to them we’re struggling to get above 30 fps at low settings. It’s returning an average of 32 fps, but at the moments you really need a bit of smoothness it feels closer to the 1% lows of 21 fps. At least the 0.1% lows of 17 aren’t too much of a further dip. Mmm. Borderline for me.
At the risk of triggering the internet, I just don’t get the appeal of the assorted Souls-alikes. As such this was my first Elden Ring experience, so don’t judge the ineptitude on display too harshly. To it’s credit, It’s getting 60 average fps at Low settings, with really rather good 1% lows of 57 and 0.1% lows of 50 fps. These are certainly very playable results, but the visual cost is pretty extreme. Look at what it’s doing to the shadows on the trees here, for example. Maybe a bit of snobbery here, but while I’d play it like this if I had no other way, I’d be starting to think about an upgrade. I don’t know how well it will come across after YouTube’s compressed the video, but in person it’s a night and day difference here at high settings. Unfortunately the framerate can’t keep up, dropping to 40 odd fps with lows at 24 fps during action, which isn’t good enough for this sort of timing dependent game. Your standards may differ, but I don’t think the P2200 has quite what it takes here.
Spider-Man Remastered was, and arguably still remains, one of those Sony Playstation ports that really ought to run a bit better than it does. Here during the opening swinging and punching and what not, we’re seeing 49 average fps, but with 1% lows of 23 and 0.1% lows of 17. This is at low settings but with no upscaling, so there are routes available for improving this, and again, if I had no other option to play this it’s an okay experience. I’d put this at borderline, leaning towards no. Again, you can make your own mind up about your standard of playability but for me it’s just the wrong side of it.
With one no and two borderline results, by my entirely subjective judgement I’d say this is the end of the road for the P2200 in this silly little test. Even if I did give it a pass, I’m certain 2023’s selection of Avatar: Frontiers of Pandora, Starfield, and Immortals of Aveum would be beyond it.
There’s also a few of the usual test suite we’ve not hit upon that I want to have for my records, even if they’re not hugely relevant to this card so I might as well quickly rattle through them:
In the synthetic benches Fire Strike gets 10403 points, Time Spy gets 3914 points, and Superposition at 1080p high gets 5263 points.
Revisiting Shadow of the Tomb Raider’s benchmark at 1080p High settings gets 56 average fps.
Counter-Strike 2 at 1080p very high settings gets a pretty solid 94 average fps.
Back to Night City, where the Cyberpunk 2077 benchmark at 1080p ultra gets 30 average fps.
And as expected the Unreal Engine 5 contingent struggles, with Black Myth Wukong at 1080p high with framegen and upscaling off gets 14 average fps, and Marvel Rivals at 1080p medium, no framegen or upscaling gets 28 average fps.
So, there we go, if for unexplained reasons you’d picked up a P2200 for lower end gaming back in the day, it’d still be okay for playing the more demanding games three to four years later. Given that it was very much not the design case for the card, that’s a pretty acceptable showing in my book.
I’ve not done anything regards overclocking this, but nonetheless it’s routinely running a few hundred megahertz faster than the specs sheet says it should, so that’s helping out.
However, if you’re looking at a card purely for gaming, well, this ain’t it, chief. For the £110 that it cost me, there’s obviously far better options for the money – a standard 1060 will perform better and cost maybe £50-60 quid, albeit with more power draw and a bigger cooler. On the positive side, if you’re looking at it for its intended purpose it’s efficient, giving a decent performance per watt, and in the compact form factor that servers or workstations demand.
I suppose if you’re doing the classic “game on an old office PC” trick and you’re constrained on power or space then there’s perhaps a case to be made for it. There’s a single slot RTX3050 variant that’s a fair bit more powerful but will cost you about twice as much, and this is about as powerful as a RX6400 but works better in older PCIe gen 3 motherboards, and has the hardware video encoding the RX6400 lacks, so if that’s a niche you’re looking to fill the P2200 still has some fight left in it.
If you disagree with my choice of games and can suggest beefier alternatives please do let me know in the comments down below, and also if you think there’s any value in doing this with graphics cards. Personally I’m keen to try this with some Nvidia 500 or 600 series cards. Until then, if you enjoyed this videotronic missive then consider subscribing, and until next time, take care of yourself, and each other.