I’ve compared AMD’s two best Zen 3 processors in games and applications to help you decide which to pick. The main differences in specs are seen in the core and thread counts, the 5950X has 4 additional cores. It’s also got a little more cache and a slightly higher max boost clock speed, so really just minor changes outside of the core differences. I’m comparing these two CPUs because the 5950X is the next step up after the 5900X, but the 5950X costs $250 USD extra, or 45% more money, so let’s find out if it’s worth it.
Both processors were tested in the same system. I’m using the Asrock X570 Taichi motherboard with 32gb of DDR4-3200 CL14 memory in dual channel and MSI’s GeForce RTX 3090 Gaming X Trio to minimize GPU bottlenecks. Both processors were cooled with my Fractal S36 AIO using Noctua NT-H2 thermal paste, as neither come with a stock cooler.
I’ve tested both CPUs at stock and manually overclocked. With both at 1.375 volts I was able to run the 5950X stable at 4.7GHz while the 5900X was just a little behind. It’s worth noting last generation I couldn’t get my 3950X above 4.3GHz, so Zen3 appears to be offering more overclocking headroom. With that in mind we’ll first check out the differences in various applications, as well as power draw and thermals, followed by gaming tests afterwards, then finish up by comparing some performance per dollar metrics.
Let’s start out with Cinebench R20. I’ve got the stock results from both processors towards the bottom, then the results with both of them overclocked above. At stock the single core performance is almost the same, and single core performance actually lowers with the static overclocks in place, as these all core overclocks prevent the higher less threaded boost speeds from being reached. The multicore scores are seeing nice improvements from the overclocks though, particularly for the 5950X, which is beating its stock score by 22% once overclocked.
Although Cinebench R15 has been replaced by the newer R20 just covered, I wanted to also include the results of this one too as many others still use it, that way I’ve got more numbers for you to compare with, similar results here in any case.I’ve tested the Blender Opendata BMW and Classroom benchmarks, and as another workload that loves higher core count it’s a win for the 5950X, which was completing the test around 20% faster than the 5900X at stock. Again as a multicore workload, the all core overclocks are beneficial here. Once the 5950X is overclocked for instance, it’s able to complete the task about 19% faster than stock.
The V-Ray benchmark is another workload that heavily favours higher thread counts, which is why at stock the 5950X was scoring 21% higher than the 5900X, the second largest improvement out of all workloads tested. Once both are overclocked though the 5950X is now 33% ahead, as that all core overclock matters more in tests where all cores are utilized. The Corona benchmark also uses the processor to render out a scene, and again it was one of the bigger differences out of the apps tested. At stock the 5950X was completing the task 19% faster, but then 34% faster once overclocked, so again nice gains from overclocking. Handbrake was used to convert one of my 4K laptop review videos to 1080p. I was honestly expecting a larger difference here due to the core count differences, but the 5950X was only able to export the video file 6.6% faster than the 5900X at stock. Overclocking helped, but didn’t make as much of an improvement when compared to those previous rendering workloads. The 5950X was 8% faster with the overclock applied compared to running stock, while the 5900X had a 5.5% boost when overclocked. Adobe Premiere was used to export one of my laptop review videos at 4K. I’ve tested this with both VBR 1 and 2 pass settings. 2 pass isn’t able to make use of hardware acceleration, so it takes longer. The 5950X is 15% faster with 2 pass compared to 9% faster in the 1 pass test. Either way though, like Handbrake the differences here aren’t as much compared to those rendering workloads. I’ve also tested Adobe Premiere but with the Puget Systems benchmark tool, as this tests for more things like live playback and more rather than just raw video export times. The differences were even closer together here though, with the 5950X scoring just 4% higher than the 5900X at stock, and this margin doesn’t change once overclocked, as both chips only improve a little there. Adobe Photoshop was also tested with the Puget Systems benchmark tool, and the scores are even closer together here. The 5950X was just 1% ahead of the 5900X at stock, so margin of error stuff and probably not something you’re likely to notice in practice. I’ve used 7-Zip to test compression and decompression speeds, and the difference in decompression was the largest out of all workloads tested at stock with the 5950X 25% faster here, otherwise it was 15% faster when it came to compression. VeraCrypt was used to test AES encryption and decryption. I might stop testing this soon as the results can be pretty different, even after taking the averages of 10 tests. Generally I’ve found overclocking to hurt this test, perhaps due to weaker single core performance. Microsoft Excel was tested using the Hardware Unboxed large number crunch test, and while I’m sure either of these processors would make a nice spreadsheet machine, the 5950X was completing the task 15% faster at stock, then 32% faster once both are overclocked. Like Cinebench, Geekbench wasn’t seeing any big differences in single core performance at stock, and as expected lower single core results are present with both chips overclocked. In multicore score though, the 5950X was 10% faster than the 5900X, increasing to a 17% higher score when both chips are overclocked.
These are the differences between the 5950X and 5900X CPUs in all of these applications, as we can see it really depends on the specific workload. The single core results are all closer to the bottom of the graph, as there really isn’t much difference between them there. The bigger differences depend on whether or not we’re actually able to run tasks that benefit from more than 12 cores, and in general this is rendering type workloads like V-Ray, Blender, Corona, and even Cinebench.
These are the differences once both processors are overclocked. There’s a nice improvement observed in the tasks that do better with more cores, again mostly those rendering workloads just mentioned, which are all now 30% plus better compared to 20% or lower at stock.
Power draw, thermals, clock speeds
I was able to overclock the 5950X 50MHz higher on all cores, and although that will affect results I think this graph puts things into perspective. These are the clockspeeds both chips were running at with the blender test running. At stock the 5900X was able to reach higher clock speeds than the 5950X, so once both are overclocked the 5950X sees a larger change, and this is why the gap between the two widens with the overclocks in place. It goes from being behind to being in front in terms of clock speed.
These results also map to the total system power draw measured at the wall.At stock the 5950X is actually using less power than the 5900X, which is kind of impressive when you remember that it’s also completing the blender render almost 20% faster too. Once overclocked though the 5950X is sucking down 22% more power than the overclocked 5900X.
More power generally equals more heat, so with both overclocked it’s no surprise that the 5950X is running warmer than the 5900X. At stock though the 5950X was actually cooler than the 5900X, and this seems to be because it was drawing less power and running at lower clock speeds, despite it actually having more cores. This exact same behaviour was also noted in my 3900X vs 3950X comparison too.
Next let’s get into the gaming results, I’ve tested games at 1080p, 1440p and 4K resolutions. Assassin’s Creed Odyssey was tested using the games built in benchmark tool, I’ve got the 1080p results down the bottom, 1440p in the middle, and 4K up the top. The 5950X was ahead of the 5900X in all cases, though not by a large margin. At 1080p the 5950X was 5% faster in average frame rate, then only around 1% faster at 1440p and 4K. It’s not all gains though, I’ve tested Microsoft Flight Simulator in the Sydney landing challenge and there was basically no difference at all. The 5950X was slightly ahead, but it’s margin of error stuff and definitely not a difference you’re going to notice when playing. Red Dead Redemption 2 was tested using the game’s benchmark. The 5950X was around 3% faster at 1080p, then the differences are more minor at higher resolutions, so again not a huge difference, but still a win over the 5900X, but for 45% more money probably not one that’s worthwhile. Shadow of the Tomb Raider was also tested with the games benchmark, this isn’t a great CPU test but if you caught my 5950X vs 10900K comparison these AMD chips are definitely offering a nice improvement compared to Intel, there’s just far less of a difference between these two. Battlefield 5 was tested by running through the same campaign. Both were able to hit the 200 FPS frame cap at 1080p, then results are very close together at the higher resolutions as processor differences start to matter less and less. Call of Duty Modern Warfare was tested in the same manner, and again no real differences between the two chips so let’s move on. CS:GO had a little more difference at 1080p, where the 5950X was 4% faster in average frame rate, then even less of a difference at higher resolutions once more. Rainbow Six Siege was tested with Vulkan and the game’s benchmark tool. Again no real differences here, however at all resolutions the 1% low from the 5950X was slightly lower. Each bar on the graph is an average of 5 test runs, and the 5950X 1% lows were much more inconsistent for some reason, perhaps 16 cores is just too much for it. Far Cry New Dawn also had some inconsistent results despite averaging 5 test runs, but either way we’re looking at a couple of FPS difference best case which isn’t anything amazing. Death stranding was tested by walking through the same part of the game, because that’s what the game is all about right? The 5950X was down a little in 1% low performance, however the average FPS was always ahead of the 5900X regardless of resolution.
Over all 10 games tested we’re only looking at a 1.8% higher average frame rate with the 5950X at 1080p, so there aren’t any huge differences with these titles. Best case, Assassin’s Creed Odyssey was 5% better than the 5900X. When we step up to 1440p they’re basically the same once we average things out. This is expected as higher resolutions will better utilize the 3090 graphics, making the processor differences matter less. Again at 4K there aren’t any major differences between these two chips, basically most games aren’t going to care too much about the jump from 12 to 16 cores, 12 is already plenty if we’re being realistic. At stock the 5950X has a 100MHz higher boost clock and a bit more cache, and I think it’s these differences that are responsible for the small differences in games that we’ve seen rather than the core count.
Cost per frame value
When we look at cost per frame, the 5900X is clearly better value for gaming. In most cases the frame rates aren’t all that different in the games I’ve tested, and the 5950X costs 45% more money than the 5900X. Basically for gaming, the 5950X isn’t really worth considering over the 5900X, consider it instead if you’re running core heavy workloads that can utilize the extra cores, as that’s where the gains are. Gaming as a secondary task will only be slightly better than the 5900X.
Although the 5950X does perform better in core heavy workloads, if we consider how much more money we’ve got to pay to achieve the Cinebench R20 scores I got, the 5900X is offering better value. The values of this graph are basically Cinebench scores divided by processor price, I chose Cinebench as the performance difference was closer to the middle of the pack out of the applications I’ve tested.
Sure the 5950X performs better in core heavy workloads like this, but you’ve got to pay more to get those results. If time is money then the 5950X is probably worth considering, otherwise you’ll need to decide if you want to pay the extra cost for the performance differences covered here. It’s also worth considering that these price differences are for the CPU only, the percentage differences become less in the context of a new $2000 system for instance.