Intel has launched their new 10 core i9-10900K CPU, but how does it compare against AMD’s 16 core Ryzen 9 3950X? Let’s see what the differences are in games and applications both at stock and while overclocked.
Starting out with the specs, the new Intel 10th gen i9-10900K is a 10 core 20 thread part. It’s got a 5.3GHz single core turbo boost, but that’s dependent on thermal velocity boost. Likewise the all core boost is 4.9GHz, but that drops back to 4.8GHz if thermals rise above 70 degrees Celsius. The 3950X has more cache, more cores, lower clock speeds, but also costs a fair bit more money.
I’m comparing these two processors because they’re currently the best options on offer from Intel and AMD’s mainstream lineup and I want to see the difference. In terms of price though, the 3950X is around a couple of hundred dollars more expensive, so in that regard a comparison to the 3900X may make more sense, but alas I do not currently have one.
Both processors were tested in the same system, but I’ve had to change motherboards. For the Intel i9-10900K I’m testing with the MSI Z490 ACE motherboard, and for the AMD Ryzen 9 3950X I’m using the ASRock X570 Taichi. The rest of the components were otherwise the same, I’ve tested with 16gb of DDR4-3200 memory running in dual channel at CL14 and with an Nvidia RTX 2080 Ti.
I’ve used the same Fractal S36 AIO with Noctua NT-H2 paste for both CPUs so we can get an apples to apples temperature comparison.
Testing was completed with the latest version of Windows and Nvidia drivers along with all BIOS updates available installed.
I’ve tested both CPUs at stock, and with the following all core overclocks applied, 5.2GHz on the 10900K and 4.3GHz on the 3950X with the same voltage. I tried 5.3GHz on the 10900K and 4.4 on the 3950X up to 1.375v but both failed in blender, otherwise I didn’t try tune these much further, lower voltage is probably possible. When testing stock I’m also not capping the 10900K to its 125w limit, as many motherboards aren’t going to limit this by default anyway, and in my opinion it’s just leaving performance on the table if you have a good cooler.
That said I have disabled enhanced turbo in the BIOS, which is MSI’s version of MCE, as that would boost the all core speed from 4.9 to 5.1GHz. 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 at 1080p and 1440p resolutions afterwards, then finish up by comparing some performance per dollar metrics.
Starting with Cinebench R20, I’ve got the overclocked results towards the top of the graph, and the stock results towards the bottom, with the 3950X above the 10900K. In this test the 3950X had a large 50% lead over the 10900K in the multicore test once both were overclocked, or a 42% lead at stock, which is expected owing to it having 6 additional cores, however the 10900K was scoring 6% faster when it came to single core performance.
I’ve also tested the older Cinebench R15 as a lot of people still use it. The single core performance was similar with the i9 in front, though the 3950X had larger gains in multicore where it was scoring 64% higher with both chips overclocked. The 3950X typically sees lower single core performance when overclocked, as 4.3GHz applies across all 16 cores which prevents it boosting up to 4.7GHz when just one or two cores are active, which is the default stock behaviour.
I’ve tested the Blender BMW and Classroom benchmarks, and as a test that works better with more threads it’s another win for the Ryzen 9 3950X. The 10900K was 28 to 32% slower depending on the specific test.
Handbrake was used to convert 4K video to 1080p with the HQ 1080p30 preset, and there was less of a difference between the two processors this time around. The 3950X was still completing the task faster, but it was only 8% faster at stock and 6% faster with both overclocked. Adobe Premiere was used to export video at 4K, and I’ve used VBR 2 pass so both were running for over half an hour. Like handbrake previously, the 3950X was faster in this test but there wasn’t that big of a difference, it was 8% faster at stock and only 3% ahead when overclocked. The 3950X was actually worse off with the all core overclock. Premiere was also tested using the Puget systems benchmark tool, and at stock the 3950X was ahead in this test. Unfortunately it kept crashing on the 3950X while overclocked, so the overclock may not have been stable, I’m not too sure as I did also try boosting the voltage further and this was the only test that failed. The Puget systems Premiere test also gives us a score for live playback in particular, and at stock the 3950X just had a subtle 4% lead. I’ve also tested the warp stabilizer effect in Adobe Premiere which is a less threaded workload and is used to smooth out a video clip. The 3950X was also completing this task faster than the 10900K, however the 3950X score gets worse with the overclock applied, as we’ve seen in other tests, the 4.3GHz all core overclock prevents higher boost speeds that are typical during stock operation. Adobe Photoshop was also tested with the benchmark from Puget systems, this time the 10900K was outperforming the 3950X. From my understanding this program sees benefit from higher clock speeds and single core performance, which would also explain why the 3950X score lowered again once overclocked. 7-Zip was used to test compression and decompression speeds and is typically a workload where Ryzen gets an edge. The overclocks only really seemed to benefit decompression on the 3950X, otherwise the results aren’t too much above stock in all other instances. VeraCrypt was used to test AES encryption and decryption speeds, and in general I find overclocks to give worse performance in this test, which was also observed here, however the 10900K was still performing the task faster in all cases. The V-Ray benchmark uses the CPU to render out a scene, and as another test that sees a benefit from more threads it’s an easy win for the 3950X which was scoring 41% higher at stock, and 51% higher with both overclocked. In many of the multicore tests the i9 doesn’t see much change, at least when compared to the 3950X. This is noted by the overclocked 3950X scoring more than 2000 points above its stock score, while the i9 was just 113 points higher. The Corona benchmark also uses the processor to render out a scene, again as another test that scales with additional threads the 3950X was completing the task 30% faster at stock and 33% faster when both are overclocked. In the Hardware Unboxed Microsoft Excel test the 3950X was also coming out ahead, completing the large number crunch test 42% faster at stock, or 44% faster while overclocked. Could you seriously imagine an overclocked 3950X office PC just for excel, you’d be smashing those spreadsheets. Like most other single core tests, the 10900K was ahead in the Geekbench 5 single core test, scoring 9% higher at stock and then 14% higher with both overclocked as that 3950X misses out on single core performance by boosting the all core speed. In the multicore test though, the 10900K was around 18% slower than the 3950X, granted this was one of the smaller differences when compared to the other multicore workloads tested.
Here’s how the Intel i9-10900K stacks up against the Ryzen 9 3950X in these tests at stock.In most cases, the i9 is slower, but that’s simply because the bulk of these tests are core heavy workloads, and the Ryzen chip has 6 additional cores. The i9 has the win in the single core tests though, something Intel are really trying to push, which also seem to include Photoshop and AES encryption/decryption.
When we look at the differences with both processors overclocked, the 3950X is further ahead in more cases than not, it appears that it is seeing a larger benefit in the overclocked state, though this would in part be due to the overclock being more beneficial as it applies to a higher core count. Some of the less core heavy workloads like Photoshop do improve with the i9 overclocked though.
When we look at the total system power draw from the wall with the blender test running, the 3950x is using less power. The gap widens a bit with both chips overclocked, though it may be possible to improve this if you can get away with using less voltage, both of mine were using 1.325v when overclocked here.
Higher power draw typically results in more heat, and that was the case here where the 10900K was running hotter too Like the power draw, once overclocked the i9 gets worse off.
Although not exactly directly comparable, the 10900K was running at higher clock speeds during this same test though as we just saw it was drawing more power and creating more heat to sustain this, all while the 3950X was completing this blender test faster. If you recall, my 10900K was sitting on 70 degrees in the stock test, so thermal velocity boost was swapping the cores between 4.8 and 4.9GHz, which is why the average speed is between this range at stock.
Let’s get into the gaming results next, I’ve tested 16 games at both 1080p and 1440p resolutions, we’ll start with stock results and look at overclocked, average differences and cost per frame afterwards. Battlefield V was tested running through the same section of the game in campaign mode. I’ve got the 1080p results down the bottom, and the 1440p results above that. In this game the 10900K was able to hit 6% higher average FPS at 1080p, lowering slightly to a 5% gain at 1440p. Assassin’s Creed Odyssey was tested with the games benchmark tool, and at 1080p the results were similar with the 10900K reaching 6% higher average FPS at 1080p, though at 1440p this increases to a 7% lead over the 3950X. Call of Duty modern warfare was tested in campaign mode, and the differences were far more minimal here with the i9 just 1% ahead of the 3950X at both 1080p and 1440p resolutions, so no real noteworthy difference between them. Borderlands 3 was tested using the games built in benchmark, and there was a bigger difference this time. At 1080p the i9 was 11% higher than the 3950X with a similar change to 1% low, then it was just 3% faster at 1440p. Control was tested by performing the same test pass through the game on both machines, and this seemingly GPU heavy game, or at least with these settings, saw one of the smallest differences out of all 16 games tested in average FPS, though there was a 5% boost to 1% low at 1080p with the 10900K. Red Dead Redemption 2 was tested using the games benchmark tool, and this game saw one of the biggest improvements with the 10900K out of all titles tested with an 18% higher average FPS at 1080p, and 8% at 1440p. Shadow of the Tomb Raider was also tested with the games benchmark tool, so no 1% low data here unfortunately. The average FPS difference was extremely minor at 1440p, however at 1080p where presumably the CPU matters more the 10900K was reaching 15% higher FPS. Fortnite was tested using the replay feature with the exact same replay file on both processors. The difference to average FPS between the two was one of the smallest out of all games we’re looking at, with slightly larger differences observed in the 1% lows. Rainbow Six Siege was tested using the built in benchmark with Vulkan. This was another game where the difference to 1% low was larger than the average frame rate, which makes sense given that’s typically more sensitive to processor performance. The 1% lows with the 10900K were 13 to 14% ahead of the 3950X for instance, while average FPS was just 4.5% ahead at 1080p and 1% at 1440p. CS:GO was tested using the Ulletical FPS benchmark, and as a game that typically favours CPU clock speed with core count not being a major factor, the 10900K saw above average gains to average FPS with a 10% boost over the 3950X. Likewise I believe Dota 2 was seeing higher performance from the 10900K for a similar reason, though strangely the 1440p results were actually higher on the i9 at 1440p, something I also saw in my 9400F vs 3300X comparison. The Division 2 was tested using the games benchmark tool, and the results weren’t too different at 1440p. At 1080p the i9 was just 5% ahead for average FPS, however there was a much larger 19% boost to the 1% low here. Overwatch was tested in the practice range, and while this runs better than actual gameplay, it more easily allows me to perform the exact same test run, which is ideal for a comparison like this. Both processors were close to the 300 FPS frame cap at 1080p, then at 1440p the differences are still quite minor. The Witcher 3 was another game that was hitting higher 1% lows compared to the differences seen to average FPS, especially at 1440p in this already GPU heavy game at ultra settings. At 1080p the 10900K was hitting a 23% higher 1% low, and 14% higher at 1440p. Ghost Recon Breakpoint was tested with the games built in benchmark, there was less of a difference to 1% low performance here compared to average FPS now. The i9 was just below 8% ahead at 1080p, and closer to 7% at 1440p. Far Cry New Dawn was also tested with the games built in benchmark, and this game saw the biggest improvement with the i9 processor out of all 16 titles tested. Even the 1% low performance from the 10900K wasn’t too far behind the averages from the 3950X.
On average over these 16 games tested, the Intel i9-10900K was around 9% faster at 1080p on average when compared to the Ryzen 9 3950X. As you can see, the results really vary by game, Far Cry New Dawn at the top for instance saw big gains with the Intel processor, while others like Overwatch and Control saw basically no change, though the i9 was always ahead in average FPS.
When we step up to 1440p the 10900K now has an average 6% lead over the 3950X in average frame rate. We’re more GPU bound here, so the CPU starts to matter less, and the difference would be even less at 4K.
I’ve also got the differences in 1% low performance too, as these results are typically more sensitive to CPU changes when compared to the average frame rates we just looked at. This time at 1080p the 10900K was 12% faster than the 3950X. Again at 1440p the results drop back a bit, but the i9 is still 9% faster than the 3950X when it comes to 1% low performance at this resolution. This would translate into a smoother experience with less dips, granted based on the high numbers we saw earlier from either chip there’s probably an argument to be made about how much you’re likely to notice this.
These results aren’t too surprising when we consider that Intel’s last gen i9-9900K was also beating the 3950X when it came to gaming. If your focus is gaming, clearly the 10900K is the better option out of the two, especially once we take price into consideration. The 3950X is at least a couple hundred dollars extra but performs worse in games. Given the 3900X performs very closely to the 3950X in games but for around $440 USD, it would make more sense to look there instead if gaming is the priority, however if you do also need the high core counts for other tasks, as we saw earlier the 3950X was able to outperform the 10900K there, granted at a higher cost.
We’ll just briefly look at some overclocked gaming results, in Far Cry New Dawn it was possible to get a few extra FPS in all instances but nothing close to enough to change the conclusion.
In Call of Duty Modern Warfare the overclocks were negative in all instances and resulted in a few FPS less, but again it’s not really much of a difference.
In Assassin’s Creed Odyssey performance was down for the 10900K in all cases once overclocked while the 3950X instead received subtle gains, end of the day, at least in these titles, just boosting the all core speed doesn’t seem worthwhile in games when you factor in the extra power and heat for often times worse performance.
All things considered, based on these results and the current prices, the 10900K is the way to go for gaming as it’s cheaper and performs better, however I suspect the 10700K will perform closely in many games for less money.
That’s not to say the 3950X is at all bad for gaming, the results it offers are still realistically quite excellent, but to get the value out of it you’re also going to want to have some heavy multicore applications to take advantage of it, as the 10 core 10900K can still do quite well in plenty of multicore tasks, so it would depend how much your time is worth.
In terms of future upgradability, AMD recently announced that it will be possible for existing B450 and X470 motherboards to support upcoming Zen3 processors, though support will depend on which boards companies pick to update, and that’s not currently clear. If I was buying a 3950X system today I’d probably be leaning towards X570 personally due to the potential issues there, but if you’ve got an older Ryzen system going to something like the 3950X could in theory just be a drop in replacement after BIOS update.
Intel 10th gen processors on the other hand launched with the new Z490 chipset and new motherboards, so if you’ve got anything older you definitely need a new motherboard anyway. It’s likely the motherboards will also support 11th gen in future, but in typical Intel fashion nothing has been stated there, perhaps for good reason after the mess AMD recently got itself into with AM4 support. It would seem that either platform should have some legs, the main difference is that AMD have confirmed next gen support while with Intel we’re just kind of assuming based on past behaviour.
At the end of the day it only seems worth paying for the extra cores with the Ryzen option if you need them, as was the case last generation, Intel is ahead when it comes to gaming. You could definitely make a case for the 3900X in terms of gaming as it performs similarly to the 3950X for a fair bit less money, the 10900K will still outperform it, but the value would change significantly if you can get by with a measly 12 cores. Let me know which CPU you’d pick and why down in the comments, Intel’s new i9-10900K or AMD’s Ryzen 9 3950X?