Multikill: AMD Ryzen 9 5900X CPU Review & Benchmarks - Gaming, Workstation, Overclocking

now for the second of four cpus that we're reviewing all in one day we started with the amd ryzen 9 5950x that's an 800 16 core 32 thread cpu now we're stepping down one tier to the 5900x which is a 12 core 24 thread cpu priced at 550 this is a follow-up to the initial msrp 3900x which was 500 at launch later became 400 and it's the direct competitor to the 10 900k the 10900k was for the longest time about 500 to 520 30 or so and so now the 5900x comes in and competes not only in the production benchmarks where the 3900x always did fine but also potentially takes the gaming crown from intel before that this video is brought to you by linux linux is a linux server provider that gn has used for years now for its own servers alongside dedicated hosting linode also makes it easy to cut out third-party vpn servers and to build your own vpn that you fully control doable through a few guided clicks in its interface linux has hundreds of guides for even more advanced server builds and with linux you can roll out your own services like plex and next cloud for your own media and file sharing services gamers nexus viewers get a 100 credit good for 60 days on new lynnode accounts visit lino.com gamersnexus or click the link in the description below so as we said in the 5950x review this review cycle is a little bit weird just because like in gpu land suddenly amd is out here somewhat alone in the charts without too much of a a recent competitor to compare to the 10 900k is the most appropriate comparison in terms of price and it's still a desktop part not an http part but it's starting to lose rank in the charts so amd has improved in ipc it's improved significantly in frequency as we saw and we'll look at that again for the 5900x and intel is stock and it doesn't have anything coming out until 2021 with rocket lake s so amd is going to be relatively uncontested at this point we'll see how the inventory is for this one as tsmc does have a limited allocation for amd that it's splitting between multiple components but either way the parts we're looking at to just again recap very briefly 5950x 800 16 cores 3900x 12 cores 550 5800 x 450 that's getting steep and eight cores and then 5600 x also steep 300 and kind of replacing parts that no one wanted to replace because they wanted instead of 3600 replacement which was a cheaper part but there's reason for amd's price creep and the 5900x review more so than the 5950x review will show you where amd is able to potentially try to justify the price increases because the gains are disproportionate to the price increase that is to say most of them are a steeper incline in increase than the price is so that's the thing andy's got going for it let's get into the review we'll start with frequency validation production benchmarks because it is still a 12 core part and gaming to see if intel can hang on to the last couple of strongholds that it has frequency testing is up first we're validating the boosting behavior of the r9 5900x for a few reasons the main one is just to make sure the cpu is working as intended and advertised this also shows the frequency response to thermals plotting the 5950x first from our initial review the average all-core frequency decays about 20 megahertz after 600 seconds of testing this is expected behavior from precision boost too the max per interval frequency on any given core was between 4020 and 4151 megahertz plotting the 5900x next average all core frequency was 44.08 megahertz decaying to about 4350 by the 600 second mark once again the maximum frequency on any given core across a single interval was just over 4 500 megahertz for one blip in time but commonly was a little bit below that to again demonstrate the frequency drop against temperature rise we see a couple of megahertz increase for every few degrees dropped in temperature and this plot helps illustrate the behavior here the 5900x gained about four to five degrees celsius over the course of this test which alongside the power ramp costed about 50 megahertz we normally see smaller changes in frequency for four or five degrees like in our live stream testing we've seen anywhere from 15 to 20 megahertz but in this instance we see that the initial burst has a harder impact as a result of the cpu starting from near the ambient and then experiencing a very slightly latent reduction in frequency while the cpu begins to heat we'll next look at the cinebench r20 single core benchmarks for the maximum frequency at a given interval and on the highest boosting core per interval this is not an all-core load we've intentionally zoomed this chart as well to see the frequencies plotted a little bit easier the 5950x held about 50 50 megahertz on any given single core while the 5900x did about 49.50 megahertz on its fastest single core per interval the dip in the middle is just a change in benchmark software and can be ignored as background noise amd claims a 4.8 gigahertz frequency on the 5900x and so it's outperforming that by 150 megahertz out of box on our model as we said in the 5950x review this is because of silicon fitness attributes within the cpus where the higher quality cpus will boost higher this is particularly true under ideal thermal conditions as can be afforded by stock behavior cooled by an arctic liquid freezer to 280 and especially when it's under just a single core workload v-ray is up first as noted in the 5950x review which is already on the channel we had a couple data points from a few months ago on these charts but not too many those are indicated with asterisks and are within two percent of the numbers if we were to fully rerun them but they're really just here for an easy reference all the other data is brand new the v-ray tile-based rendering benchmark had the amd r9 5900x at 26 000 points stock with an overclock bumping at 4.4 percent higher the 5900x entries actually flanked the previous 3950x despite the latter's threat advantage and 200 higher original msrp this might be why amd was so comfortable raising its price floor this time around the 5900x stock cpu outdid the 3900x stock cpu by 25 which is a massive generational improvement the intel i9 10900k competitor runs 18 000 points with the 5900x stock about 42 percent ahead of that the overclock didn't really help intel very much here adobe photoshop is next as discussed in our 5950x review that's again already live photoshop has historically been a benchmark that intel has maintained full chart topping control over it's been uncontested here for years it wasn't long ago that even a 9600k made more sense than some of the top-end ryzen cpus now that's changed amd holds four out of the top five entries currently on this chart and this test benchmarks scale transform warps filter application like blurs and colors and other functions of photoshop so it's fairly conclusive the 5900 x scores 13 30 points in aggregate with an overclocking slightly improving that but not in a way that's worth it if all you do is photoshop the 5950x stock cpu leads the 5900x stock cpu by just 3.2 percent showing that the extra threads don't really matter in this application at this point the 10 900k stock result is bested by the 5900x by 13 this is thus far the most important launch amd has yet had for its zen architecture for retaking the last couple of strongholds that intel has had intel is at this point losing nearly every single stronghold that it had as someone who said the same of amd about a decade ago and for the last decade it's insane to watch it unfold in reverse onto the compression and decompression tests measured in millions of instructions per second or mips and these threadripper class cpus are constrained here but we still see good scaling up to the 5950x the 5900x cpu plotted 115 000 mips with the overclock barely changing that to 117k mips the 10900k is the direct competitor so against that the stock to stock amd advantage is about 33 percent as for the 10980xe from our last round that one's at 115k mips which places it between the 5900x stock and oc results and behind the 5950x stock results for reference for owners of the 2000 series the 2700x does 54k mips here and was previously amd's high-end desktop cpu decompression shifts the stack around a little bit the amd r9 5900x runs 175k mips the overclock at 182k mips and the 3950x stock cpu at 188k mips the 5950x stock cpu out does the 5900x stock cpu by 25 percent here and the 5900x out does the 10980xe which was a thousand dollars if you were ever able to find one by 17 the 5900x also out does the direct 10900k competitor by about 60 percent the 10900k oc was within variance of the 10900k stock here as a result of the slightly more variable nature of 7-zip decompression the 5900x improved 20 percent over the amd r9 3900x at the top end we see the amd threadripper cpus have become bound by memory just like the compression chart but they still rocket ahead of the desktop class parts the blender cycle's tile-based rendering tests are next using gn's in-house made benchmarks we have a lot of experience with blender and we've built cpu benchmarks based on what we've needed for our own use most recently we used blender to 3d model the molt for our gn bar runner bar mat which is available on store.camerasnexus.net the gn logo rendered test required 10.6 minutes to complete on the amd r9 5900x when overclocked slightly better than the 11 minute result of the stock 5900x the amd 5900x only leads the 3900x by 11 percent here so the generational gains are clearly contingent based upon workload blender is an all-core 100 workload and there's no room for boosting so this makes sense the 10900k required 17 minutes for this render allowing the 5900x and 3900x to both push the core count advantage this workload draws one tile for each thread present so intel was never expected to win in this test anyway but the gap is growing now at 34 reduced time on the 5900x versus the 10900k as for the 5950x if you're debating whether you should buy the 5900 or 5950 that would give you an additional time reduction of about 17 versus the 5900x in the gn monkey had seen render the r9 5950x cpu required 7.9 minutes ahead of the 5900x stock cpu by about 16 percent the 5900x was ahead of the 3900x by 16 as well and the 10 900 k's stock 14 minute result was beaten by 33 percent by the 5900x the 2700 axis stock result was beaten by 56 percent by the 5900x our main takeaway is that for blender or tile-based rendering applications in general where you hope to leverage the cpu the 5900x might be the best option for someone who wants to free up an extra 250 for a higher-end gpu rather than spending it on the 5950 you could spawn two blender instances for example and render frames front to back on the cpu and then back to front on the gpu with cuda or optics which would get you the best of everything especially if the price difference is enough to move up to a 3080 from a 3070 or similar it all depends on how you use the application ev rendering would mostly benefit from a gpu upgrade here while tile based rendering could use both efficiently ffmpeg is next we dropped handbrake and instead went with a custom ffmpeg benchmark to better balance thread loading on the 16 core cpus we first do a single file benchmark but the scaling does still slow down at 16 cores so to better occupy all of the threads we also transcode three files simultaneously as a second benchmark this is where the lower core count cpus start to break down under the load in the single file test the amd r9 5900x required 18 minutes to transcode or 17 when overclocked while the 3900x required 22 minutes stocks to stock that's an 18 reduction generationally the 5900x stock outperforms even the 16 core 3950x by six and a half percent here while the 5950x out does the stock 5900x by 11 finally versus the 10 900k we're looking at an advantage favoring the 5900x by about 27 in the triple file test the 3950x out does the 5900x as we'd expect and flips the rank from the single file transcode the 5900x now outranks the 10 900k by 35 rather than 27 in the previous single file test and the cpus at and below eight cores disproportionately lose their scale versus the higher core count cpus for example the 2700x is taking about three times longer which is linear but it's a non-linear increase in time required for those additional two files on some of the other cpus like the 5950x finally for this round here's the chromium code compile benchmark the 5900x required 57 minutes to compile the code base so it's eight percent faster than the 3900x in this one that's not worth a swap from the 3900x to the 5900x in our opinions but it's worthy of consideration versus say a 2700x or a 3700x if you're still on the 3900x this alone wouldn't really be enough to move over to the 59 unless you're making money off of doing this type of work and you've been held up by it the 10980xe technically outranks the 5900x here but at the original price it's not a great comparison the 5950x wins in both price and performance and for perspective outperforms the 5900x stock by 18 the 5900 x is still a good balance at its lower price than the 5950 but either of them would make more sense than spending a thousand dollars on the 10 980. threader per cpus still offer them on the best performance overall but you need to be careful of a few things with this testing one of them is the motherboard cost does balloon the total system cost and another one is to be careful to buy enough memory for cpus like the 3970x or 90x or even the 5950x to some extent to ensure that you don't end up pinning the i o like an ssd during the compile that brain threaded for down below even at 10 600 k if it were happening and it's something we demonstrated in our cpu methodology update for this year time to move on to games we're starting with civilization six specifically using the ai turntime benchmark now that we've refactored this test the turntime benchmark is valuable as a different unit of measurement rather than fps because it helps to illustrate the real time required to sit and wait for your next turn this is tested against five ai players a total of four times so 20 turns of data the 5900x technically was faster than the 5950xoc but only by decimal points the 5900x stock cpu finished turns on average in 26.9 seconds right alongside the 5950x stock result of 26.6 seconds this is actually distant enough that given how this bench behaves it's probably an accurate representation of the real differences and not just error that said it's not a difference anyone would really notice the 10900k did this in 31 seconds per turn so the 5900x ran 13 faster than the 10 900k stock benchmarks compared to the 3900x the generational improvement is a reduction in turn time of 17 by moving to the 5900x intel isn't completely out of the fight though in red dead redemption 2 the 10 900k is still the leading force the 5900x stock cpu ran at 147 fps average when stock and overclocked with lows disproportionately behind the average as compared to intel they're still fine but the positioning isn't leading intel's 10900k leads the 5900x stock to stock by 5 the 10-700 k also maintains a lead and is technically supposed to be cheaper running 153 fps average versus 148 fps average versus the last generation the 5900x stock result outperforms the 3900x stock result by about 20 fps average or 14 tested at high settings instead which is useful for imposing a gpu bottleneck the cpus are almost entirely constrained by the 380fcw3 the 10900k maintained its lead at 177 fps average which is a 12 improvement over the stock and the r9 5900x and it's 159 fps average in the past we've seen similar behavior where as it becomes gpu bound amd can fall behind intel and we see that in the past couple years of reviews where there's typically a wall that intel hits and then a different wall that amd hits a little bit lower despite both being gpu constrained in this chart the lows are also running tighter with intel cpus including the 10700k the 5900x oc proved troublesome in this one with some dips in frame time consistency but it managed to improve technically to 164 fps average compared to the 3900x even with the gpu limit imposed the 5900x improved 12 over the 3900x stock cpu total war 3 kingdoms and the battle benchmark comes up next at 1080p the 5950x stock cpu ran at 206 fps average 146 fps one percent and 127 fps 0.1 percent compared to the 5900x the difference is nothing they're functionally equivalent in this test and so buying up the chain wouldn't thus far benefit anyone if you're mostly gaming the 5900x does everything the 5950x does in gaming for the most part and the 5950x should mostly be reserved for production tasks or use cases otherwise you're just wasting money we'll see soon how the 5600x does in these tests at the lower priced end of course as the 600 series of amd's cpus often does well enough that you could also disregard the 900 series but we'll see in the next review which will be on this channel today the 5900x oc is below the 5900x stock result here because the single core boosting frequency is still higher than the all core oc sometimes it's better to have the all core oc and sacrifice the higher boosting frequency but in this game losing the peak frequency can hurt the performance and we're seeing that here we'd see the same in the 5950x and it's oc except it's already bottlenecked elsewhere and so it ends up just looking the same the 5900x 205 fps average keeps it at 10 fps ahead of the 10 900 koc result and 7.2 percent ahead of the 10 900k stock result the 10700k seeds ranked to the 1500x by 11.5 and then compared to the 3700x and 2700x we're looking at around a 40 to 50 fps advancement in this title in the 1440p version of this test the 5900x operated at 139 fps average but we want to emphasize here that we're so gpu bound as shown by comparing to the previous chart that these cpus are bouncing off the limit and therefore can't produce reliable data the ceiling appears to be about 132 to 139 fps average with intel stuck technically a little bit lower than amd in the division ii at 1080p the amd r9 and 5950x set the ceiling at 241 fps average for the stock cpus with the 5900x at 234 fps average the 5900x stock cpu outperforms the intel i9 10900k both stock and overclocked improving by about 3.6 percent against the stock 10900k tomb raider is next in this one tested at 1080p first the 5900x stock cpu held an average fps of 201 which has it right alongside the 5950x stock cpu the 3900x cpu ran at about 146 fps average so we're seeing a generational uplift of 37 stock to stock if you look back at the 2700x result of 130fps average and the 3700x result of 142 you can see the schism that forms between zen 3 and zen 2 as opposed to the one between zen 2 and zen plus despite the maturing architecture amd still pulled out noteworthy gains in gaming applications for this generation the improvements have clearly slowed in production workloads where we're looking at maybe 11 in many of those cases but games can greatly benefit from ipc uplift and frequency uplift especially in those scenarios where limited thread boost in frequencies can be fully leveraged f1 2020 is another game that's highly bound by the cpu evidenced by the lack of gpu limitations even at 355 fps average here the 5950x plotted a 355 fps average when overclocked and outdid the 5900 axis 346 fps average stock the stock by 1.7 as expected the 5950x should be reserved primarily for production systems versus the 10 900k stock results of 322 fps average the 5900x maintains a 7.4 lead compared stock to stock the 5900 xoc proved slightly problematic for stability in this one and underperformed versus stock so we removed it from the charts we'd need to increase the voltage or decrease the frequency by a little bit to get it to work well in this particular game compared to the 3900 x's 255 fps average the 5900x improved by nearly 100 fps or 35.8 percent that's possibly the biggest single year improvement we've ever seen in a cpu release assassin's creed is next this one has always been interesting for its mix of core and frequency requirements all of these are within a few fps of each other due to gpu limitations so the 10900k and 5900x which are within 2 fps in this test are the same the 5950x is the same as well the overclock results don't mean anything the only place that an actually meaningful rift forms is at the 3900x but above that all these cpus give about the same experience and that's because they're hitting the gpu limit gta 5 is actually pretty interesting and it's because of the age of this test we brought it back for this one to show this point even in this game where intel has mostly kept distant control versus say the 3900x at 112 fps average amd has finally managed to wrestle back elite the 5900x runs at about 140 fps average here with the oc doing nothing the 5950x benefited from about a 4fps uplift versus the 5900x with its oc also doing really nothing outside of variants the 10900k got stuck at 130 fps average allowing amd an 8 percent lead price for price that's a big change from the 3900 x's max of 112 fps average and has restructured this chart as a result power testing is last for this one we test power at the eps 12 volt cables rather than the wall which gives us a better look at the per component power draw without all the inaccurate noisy data of the rest of the system the 5900x stock cpu measured about 133 watts in blender when under an all-core workload higher than the 120 watts of the 5950x this is the behavior we saw between the 3900x and 3950x as well where the 3950x tended to plot lower in power consumption than the 3900x part of this is due to the higher silicon quality of our 5950x for this one which allows a lower sustained voltage and less lower power the overclocked 5900x did about 193 watts in this workload when it was set to a 1.28 volts get v core and 4.7 gigahertz all core frequency the 10900k for reference pulls about 130 watts stock after tau and 317 watts with our overclock next up is the cinebench all-core benchmark except we take this measurement within the tau period and thus see higher power consumption on some intel cpus the 5900x remains around 134 watts then 198 when overclocked and the 10900k rockets to 200 watts for its 52 second window of tau so then the 5900 x at 550 it's not too far off of the original 3900x pricing the 3900x still a very good cpu at 400 it basically became an unbeatable choice and if it's still that price then it might be worth considering that to save the money if you're doing production workloads because yes this new one is better but if you're still talking 400 bucks for the 3900x that was a good price and depending on availability it may still be a good deal but the 5900x at 550 it is actually defeating the 10900k in a lot of the games now the 10900k does hold a victory in a couple of places primarily red dead redemption 2 but it's lost in photoshop where it previously was one of the leaders all of which were pretty much intel cpus and it's lost in a lot of the other games the 5900x primarily shines as being just what it exactly is a cheaper alternative to the 5950x as opposed to being a more expensive alternative to a 5800x in other words if you're buying just a gaming cpu and these lower end stuff as it stands now might be the ones that make more sense to look into where you could then re-allocate the additional budget into a gpu where you'll get more direct benefit because as we've all seen there's diminishing returns at some point with cpu scaling as you increase in price class the 5950x for example doesn't make sense to buy for a gaming only system it's still a workstation part and that's where the 58 and 56 might step in the 59 would be a good compromise if you do something like say blender work where you can leverage tile based rendering for cpu from front to back and then gpu from back to front of your animation and split the workload between them with two different tile sizes to optimize for each component that's where 5900x makes sense if you only have so much money to spend you can take the extra money the extra 250 or so dollars that you would have spent on the 5950x and put that into a 38 instead of 3070 or whatever the case may be and that could potentially get you more uplift than the additional cores would be if you actually are able to use cuda or optics in your workstation tasks so that's the way to dice this one up if you're pretty cpu bound and you don't necessarily need the memory bandwidth then the 5950x does start to make sense if you need memory bandwidth well wait around for threadripper and if you can otherwise split a lot of your work into gpus and you need some budget for it the 5900x steps in at that point so that's going to be it for this one check back for this 5600x review and the 5800x review coming up next on the channel subscribe for those as always you can go to store.gamersaccess.net if you'd like to help us out directly like by buying one of our brand new gm bar runner bar mats you can back order the uh mod mats that are on the store as well mouse mats will be back in stock soon otherwise patreon.com gamersnexus thanks for watching we'll see you all next time [Music]

Channel: Gamers Nexus

Views: 802,062

Keywords: gamersnexus, gamers nexus, computer hardware, amd r9 5900x benchmarks, amd ryzen 9 5900x cpu review, amd 5900x review, amd r9 5900x review, amd r9 5900x vs r9 5950x, amd r9 5900x vs r9 3900x, amd r9 5900x vs intel i9-10900k, intel i9-10900k worth it, intel i9-10900k overclock vs amd, amd r9 5900x overclocking, amd r9 5900x power consumption, best cpus for video editing, best photoshop cpus, best gaming cpus 2020

Id: utWSSlyabjc

Length: 27min 8sec (1628 seconds)

Published: Thu Nov 05 2020