Gadgetory

welcome back to our run box today we are finally able to check out the 12 and 24 course 2nd gen thread robots known as the 29 20 X and the 2970 WX now I say finally because it was about two two-and-a-half months ago now that we first checked out these 16 and 32 core models notice the 29 50 X and the 2990 WX we have had to wait quite a while for the 12 and 24 core parts but AMD did warn us when we reviewed the initial second gen thread refer parts that the rest of the lineup wouldn't arrive until October and I suppose we are still October even if we are at the tail end of October but here we are finally able to review them but before we get too far into the review today's video has been sponsored by LastPass with LastPass you don't have to write or remember any passwords they can generate a unique password for every website store it securely in the cloud and autofill into websites to make your life much easier and so much more secure both Tim and myself have been using LastPass for years now so it's a service that we trust and can highly recommend for more information please check the link in the video description ok so as is often the case with these day 1 reviews we weren't given too much time to put this thing together so that been the case rather than massively rush all the testing i'm going to provide two videos there'll be a follow-up video tomorrow which will include more gaming benchmarks today's video is focused on productivity benchmarks and with a very brief look at gaming before we get into the benchmarks here's a quick refresher covering the design and specs there's two models in the WX series and the W signifies that this is a workstation series the workstation 2990 WX and 2970 WX models are configured very differently to the 2950 X and 29:20 X processors whereas the 12 and 16 core models pack two zeppelin dies the 24 core and 32 core processors feature for typically such a configuration would have four dual channel memory controllers for eight channels this however isn't possible on the X $3.99 platform limiting these super core heavy parts to quad-channel memory although there are two more Zeppelin dies the additional two dies our compute dies in AMD's words this means they have no local PCIe or DRAM access for that they must travel via the Infinity fabric to the i/o dies as there are twice as many dies the Infinity fabric bandwidth is also half so now the throughput between dies is just 25 gigabits per second assuming using ddr4 3,200 memory because of this design that sees two of the dies without direct access to the DRAM it means unlike the 29 20 X and 29 50 X the 29 is 70 WX and 2990 WX use Numa exclusively aimed a claim at this quad Numa configuration allowed them to create the world's first 32 core consumer processor and just as important as allowed them to do it while maintaining backwards compatibility with existing tr4 products this did cause some compatibility issues or rather I should say performance issues with games and applications and I guess also the Windows 10 operating system the Windows 10 scheduler has been proven to be very inefficient at managing these core heavy CPUs we found significantly better performance when testing the same applications with a Linux based operating system unfortunately this store remains a huge problem for the 2990 W X and therefore we suspect it will also be true for the 2970 W X and while Microsoft hasn't done anything to improve the situation at least yet AMD has the latest installment of the Rison master software which will be available for download today introduces a dynamic local mode and crucially to enable this mode you don't need to reset the system in AMD's words dynamic local mode is a new piece of software that automatically migrates the system's most demanding application threads onto the thread Ripper 2990 WX or 2970 WX CPU cores with logical memory access in other words applications and games that prefer local DRAM access will automatically receive it but applications that scale to many cores will still be free to do so it's great to see AMD working hard to improve the user experience with these high-end desktop processors but ultimately the biggest improvements will come when Microsoft updates the window scheduler for 2990 WX and 2970 WX will be safer purchases for those using a Linux based operating system but for certain tasks such as rendering they are still beasts in the Windows environment specs wise the 12 core part is virtually identical to the 16 core part obviously - a few cores and the same is true in comparing the 24 core and 32 core parts that being the case I won't go over all the clock speeds cache capacity and all that stuff again that will just delay us getting to the good stuff the benchmarks for testing all systems work and figure with ddr4 3,200 cell 14 memory the quad-channel platform has received 32 gigabytes and the dual channel system 16 gigabytes the third river cpus were benchmarked with the NM axe lick tech 360 tr4 while the sky like X CPUs used a 360 millimetre open-loop setup this won't skew these stock performance in Intel's favor though the coffee lake and second gen rise and parts were both tested with the corsair h 115 i pro and finally the graphics card of choice is gigabytes RT x 28 ET ID gaming OC alright time for the benchmarks as expected the 29:20 x roughly matches the 1920 x and 29 50x when it comes to sustained memory bandwidth performance the 2970 WX though its surprised with a throughput of 67 gigabytes per second which is a few gigabytes per second fast in the 2990 WX and 6% more memory bandwidth than a typical thread Ripper CPU Cinebench single thread performances right where we'd expect it to be the 29:20 x scored 178 points placing it on par with the 2150 X while the 2970 WX match the 2990 WX and the older 1920 X of course it's the multi-threading performance that's of most importance for these core heavy CPUs and here we see the 2970 WX cleaning up coming second only to it's 32 core version a score of roughly 4,300 points made it just over 30% faster in the Intel Core i 979 ATX e meanwhile the 29:20 X just edged ahead of the 1920 X making it just 25% slower than the Intel 16 core 79 60 X and in advance I'm going to apologize for all the X's in this video again the 2920 ex offers a small performance improvement over the 1920 X this time in our blender workload the 2970 WX is also a big step up from the 79 80 XE as it was again a little over 30% faster it was also 36% faster the thread rep 429 50x taking just nine point five seconds and that meant it was just 13% slower than the 32 core 2990 WX meanwhile the 2970 WX was still faster than the 79 80 XE in Corona though unlike the previous tests where it won by a 30% margin or greater here it's just 8% faster the 2920 X was again a few percent faster in the 1920 X and 9 percent faster than the 9900 K and that is of course running the 99 Rekha without a TDP limit in place the last rendering benchmark that we're going to look at is v-ray and again the 2970 WX makes short work of the more expensive 18 core processor from Intel he repeat the 79 80 XC by a 22 percent margin there was also just 13% slower than the 32 core 2990 WX so a solid result for the new 24 core part the 29:20 X also does well despite offering just 5% more performance than the 1920 X and here it was 13% faster than the core I $9.99 hundred K the PC mark 10 gaming benchmarks shows the 2970 WX doing much better than the 2990 WX though the 2950 X is still by far the best thread Ripper CPU in this test and while the best CPU overall 7-zip compression performance is an issue for the 32 core process or at least when testing on windows and thus far the issue has not been addressed and therefore we see the 12 core 29:20 X just being the 32 core part while it's mashed the 24 core model still when it comes to file compression performance Intel does appear to have a reasonably significant advantage however when it comes to decompression the AMD processors certainly do come into their own the 2990 WX is a weapon here and the 2970 WX is also mighty impressive beating the 79 80 XC by a 29% margin the Microsoft Excel benchmark might be a bit redundant now as most of these high-end desktop CPUs take less than two seconds to complete the workload but we have the results so why not include them Intel does enjoy a slight performance advantage in this test but with the Monte Carlo simulation being such an extreme workload I doubt anyone is ever going to notice Intel's performance advantage in this application where you might notice Intel's performance advantage is when working with h.265 content as their superior AVX implementation can be seen here AMD's really cool heavy CPUs also have an issue with the Windows schedule in this test which certainly doesn't help that's the reason why the 12 core thread Ripper CPUs are seen matching the 24 core and 32 core models that said AMD 16 core 2950 X did put up a decent fight but even then the 79 60 X is still 14% faster though in terms of value the Intel CPU is still much worse since reviewing the 2990 WX adobe premiere has seen a major update and with that performance of the 32 core processor has been much improved though it's still slightly slower than the 16 core model still it was considerably slower previously but 2970 WX basically match the 2990 WX and therefore was slightly slower than the 2950 X still while the 2950 X offers the most bang for your buck of the higher-end thread Ripper CPUs in premiere the 2970 WX still stacks up well against the Intel competition as it is just 10% slower than these 79 80 XE then we have the new 29 20 X which also does very well though it is only 7% faster than the 1920 X and at launch does cost considerably more still the 29 20 X was on par with the core I $9.99 ok which ran much hotter without the TDP limit here we have our mainstream warp stabiliser test running just a single iteration I haven't had time to run our dozen simultaneous warp stabilizer tests just yet but I will include that data in future content again the 2950 X's seem to be the best value thread Ripper CPU it's certainly at the high end when testing and premiere while the higher 24 core and 32 core parts do seem to lag behind ok guys don't freak out but for now I've managed to squeeze in just a single game test like I said the focus was primarily on productivity but tomorrow I will focus on gaming performance for the new thread Ripper CPUs and I will have over 20 graphs for you guys to go over here we have shadow of the Tomb Raider providing a small sample of the work I've done so far now here we'll start with the 1080p data and again I'm using the r-tx 20 atti if you want to know the exact quality settings used please check out Tim's optimization video for shadow of the Tomb Raider with the dynamic local mode enabled the 2990 WX and 2970 WX are still providing sub optimal frame time performance though the average frame rate isn't that bad however we do see the 29 20 X and 29 50 X matching the frame time performance of the 2,700 X while beating it by a reasonable margin for the average frame rate I actually went back and retested the 2700 X just to make sure there wasn't any kind of update that improve performance but I did get the same 93 FPS on average in our particular pass in the game so it seems that the 12 and 16 core thread Ripper CPUs are offering a slight performance increase here over the 2,700 X which I have to say is most unexpected however once we moved to 1440p the 2,700 X is now able to roughly match the average frame rate of the thread Ripper CPUs while offering slightly better frame time performance though overall performance was much the same between the 29:20 x 2950 X and 2,700 X they're all comparable to the 9900 K as we are mostly GPU bound in this title again even with dynamic local mode enabled we are seeing some serious frame time issues with the 2970 WX and 2990 WX processors this persisted even at the extreme 4k resolution as far as I could tell the dynamic local mode was working but aim D has reported a few bugs with enabling it hopefully they're sorted out for the public release which should be available now and perhaps I'll retest with that version anyway as expected with the exception of the 24 core and 32 core models we're very much GPU bound at 4k and here the 29 20 X was able to get the most at the RT X 28 ETI moving on to power consumption in here in Hamburg the results might look a bit odd given the 12 core and 24 core parts are seen pushing total system consumption to similar levels there are however a few reasons for this firstly handbrake doesn't utilize the 24 and 32 core parts very well so they are being pushed as hard as they could be the other reasons include clock speed and voltage the 29 20 X maintained an all core clock speed of three point nine five gigahertz at one point to twelve volts while the twenty nine seventy WX ran at just three point five five yards using one point zero seven five volts that's a ten percent decrease in frequency and eleven percent decrease in voltage blender is a better test to show maximum system draw for these higher core count processes and here we see the twenty nine ninety WX pushing total system consumption of 345 watts while the 2970 WX consume 10 percent less power this time the twelve core model is seen reducing system consumption by 15 percent which is reasonable given it runs at higher clock speeds with more voltage in the end when fully utilized the twenty nine seventy WX only push total system consumption eight percent higher than that of this 79 80 XE and remember it was 30 percent faster in this workload meanwhile the twenty nine twenty ex pushed total system consumption of 264 watts which is basically what we saw from the 29 50 X and seventy nine 60 X when it came to temperatures both CPUs run extremely cool with the nm AK slick Tec 360 tr4 in place we were looking at peak temperatures of 40 to 45 degrees after an hour long blender stress test which is just insane therefore I decided to remove the liquid cooler for this test and try it a big air cooler I went with the Wraith Ripper the 29:20 x maxed out at just 61 degrees after an hour and again this allowed it to run at the same frequency and voltage that was seen previously with the liquid cooler on the MSI X 399 creation unbelievably the 2970 WX ran just two degrees hotter hitting 63 degrees again using the wraith Ripper air cooler on the MSI X 399 creation using the default bios settings just XMP was loaded so when it comes to overclocking you likely won't be limited by thermals speaking of which I could get both CPUs into Windows at 4.2 gigahertz but only the 29 70 WX was able to pass our stress tests at that frequency the 29 20 X would pass quite a few tests such as Cinebench and quite a few gaming tests for example but it would lock up after just 10 minutes of our blender workloads and in the time we had available wasn't able to stabilize the overclock that set at 4.1 gigahertz with 1.3 volts it was rock-solid and spat out a score of 2653 points a 6% boost over the stock are the Box performance okay before wrapping things up let's go over a few performance versus price scatter plots starting with blend up so on this graph you want to be positioned as low as possible and as far to the right as possible lower of course means cheaper and further right means faster as you can see the blue dots which represent the Intel CPUs well they're all higher and more to the left term the competition for example the thread route for 2920 X destroyed the core i7 78 20x it also beat out the core Oh 999 hundred K and in terms of value the 1920 X just destroys them all meanwhile the 2970 W X is much cheaper and faster than the 70 960 X so an easy win there therefore for those of you taking on 3d rendering tasks the thread Ripper series really is a must when it comes to video encoding it's a bit more mixed again the ultimate value award here must go to the 1920 X but the new 29 20 X while not as good as the 1920 X in terms of value it does beat out the Intel Core i7 78 20x by a rather convincing margin and even edged out the ninety nine hundred K the 29 50 X also does very well but unfortunately the 29 70 WX and 2990 W X are underutilized and have issues with encoding workloads particularly on the Windows 10 operating system so at least at the very high end of things Intel does come out on top for this type of workload that said the margins when encoding with premiere are far less brutal for AMD and while the 2990 WX is a still worse value than these 79 60 X you could argue that the 29 70 X actually isn't offering reasonable performance for a much lower price that said the 2150 X again makes so much more sense as it's not only faster than the 24 and 32 core parts but it's also cheaper the new 29-28 roughly matched the 900k and 78 20x the Intel CPUs were slightly better but it is a close call of course those price vs. performance graphs don't tell us everything about this comparison in terms of value now normally AMD does tend to catch Intel out with a lower motherboard prices along with the inclusion of a decent box cooler but that's not actually the case with the thread Ripper series so the cheapest x3 99 boards they typically cost around $300 u.s. where is the cheaper X 299 boards they start just under $200 u.s. and this really is a factor for those buying the lower end CPUs on these platforms of course if you're buying a 16 called Plus CPU you're probably looking at spending over $400 u.s. on the motherboard and that is for either platform as for the cooler well neither the thread Ripper or sky like X CPUs come with the cooler so you're looking at a similar expense there where AMD currently has a value advantage is in the platform itself they offer loads of PCIe lanes our ECC memory support and the ability to expand beyond just 18 cause the last ones actually quite a big one especially when third gen thread Ripper will work on the same motherboards so that MSI X 399 creation that you spent $500 on won't need to be tossed for a considerable amount of time yet ok so let's first talk about the thread Ripper 29 20 X this chip has an MSRP of $650 u.s. which is much cheaper than 800 dollars us that the 1920 X came in at basically AMD's offering 12 cores for the price that Intel will give you 8 and you basically get three times the l3 cache and just over twice as many pcie 3.0 lanes and then there were just two tests where the 78 20x actually managed to beat the 29 20 X and those two tests while one of them was Cinebench the single threaded test this stomped it in the multi-threaded and then the other test was PC mark 10 that was very close for all the other rendering and encoding benchmarks the new 12 core every of a part really did wipe the floor with the 78 20x it was basically an on contest unfortunately I didn't have time to retest the core I 970 900 X in time for this video I already had to retest a lot of CPUs so ya couldn't retest them all but based on the data that we do have on hand the 2920 acts will dominate that part as well that's mostly because the technical Intel CPU costs over 40 percent more so that's obviously a big markup for not really much of a performance uplift if any that just leaves the 9900 cane honestly if productivity tasks what you'll be focusing on then I believe the 29 20 X is the smarter buy it will end up costing a little more overall but for applications that utilize the 12 core thread Ripper CPU well heavily overclocked $2.99 or ok will melt trying to keep up of course if you mostly wanted game then the 99er ok is a better choice there I'd recommend you forget the note ok exists at all let's just be honest right now pretty much doesn't instead I'd recommend getting an 87 ok or perhaps the 2700 X the choice is yours so whereas I found myself with nothing bad to say about the 29 50 X when I reviewed it back in August it's pretty much the same story with the 29 20 X in late October the only issue for the 29 20 X right now is the 1920 X selling for a little over $400 u.s. that's the obvious choice right now and if I were in the market for a high-end desktop system that's what I'd be snapping up before the stock runs out then we have the 24 core thread Ripper 2970 WX which is a 3d rendering Beast just like the 2990 WX unfortunately there right now windows won't elating at the most out of this processor in most workloads so unless you're using or are prepared to use Linux it's probably not a CPU I'd recommend unless you absolutely know it can blast its way through whatever workload you'll be throwing at it overall I see even less reason to invest in Intel's X 299 platform and if AMD didn't already have the high-end desktop market a segment all stitched up last year they certainly do now Tim basically summed up the situation well in his video titled Intel surrenders to thread Ripper with new skylake X refresh it's very disappointing that Intel's remain stubborn on pricing and just refused to become more competitive even with the skylark X refresh but I guess no one will be really surprised by that anyway I'm glad we now have the 12 cord 29 20 X on the menu and I'm hoping that we won't have to wait too long to Windows gets an update to better utilize the 24 and 32 core models and well that is going to do it for this one if you didn't race off to pick up a 1920 X then I hope that means you enjoyed the video feel free to subscribe for more content and if you appreciate the way we're doing our box then consider supporting us on patreon to get access to our little community of enthusiasts on discord and you can also catch our monthly live streams I'm your host Steve and I'll see you again next time you