Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

AMD Ryzen Threadripper 1950X & 1920X Review: Core i9 Killer

2017-08-10
so you've watched as a specification slowly leaked out over the past couple of months and lastly you saw it unboxed around the world via countless media outlets now the time has come for us to finally discover if Amy's latest and greatest high-end desktop CPUs are worth the asking price or they just hot power-hungry and overpriced to Lincoln for suckers Before we jump into the benchmarks here's a quick look at the specifications one more time for this comparison we'll be testing a thread rip in nineteen fifty X 1920 x along with the horizon 7 1700 and 1800 X meanwhile we'll also have some competition from Intel with the core i7 700 X 78 20x and core I nine 7900 X when compared to the horizon 7 1800 X consumers are faced with a 75% price premium for the 1920 X and yet they do only get 50% more cause that said though you do get quad channel memory support and significantly more PCI Express Lanes with the thread Ripper CPU what I'm most interested in is how the thread Ripper CPUs compare with Intel skylake X parts and that's what we're here to find out finally a quick note on the test system specs for testing the sky like X CPUs we use the as rock fatality X 299 professional gaming i9 motherboard while the thread Ripper CPUs were tested with the asrock X 399 tyshee both platforms were fitted with G skills tridon z ddr4 3200 CL 14 64 gigabyte quad channel memory kit and a big thanks to g skill for providing all that memory the ex 299 system was installed on the Praxis wet bench and the cooler used was Thermaltake specific RL custom liquid cooling kit meanwhile thread ripper was tested on a sheet of foam on my desk using thermal takes flooring RGB 360 all-in-one liquid cooler all right enough chitchat let's get to it kicking things off let's check out the memory bandwidth performance as you probably know the Rison 7 CPUs feature a dual channel memory controller whereas thread Ripper offers quad channel memory access much like these skylake x cpus as a result of memory bandwidth has now been increased by 50% over horizon 7 and now Red River is on par with the core o 970 900 X memory performance is looking good so let's move on and check out CPU performance with Cinebench well as promised by AMD the 1950 X is good for a multi thread score of over 3,000 points we saw 3028 points which is pretty incredible as a 39 percent increase over the I 970 100x both of which do of course cost $1000 us the core I 9 CPU was however 10 percent faster for the single threaded test though that's not exactly the point of these CPUs the 1920 X was also 13 percent faster than the 79 her X and 40 percent faster than the i7 78 20 X so a clear win here for AMD in terms of price versus performance before moving on to the more serious benchmarks I just wanted to see how thread Ripper stood up in PC mark 10 which looks at typical sort of office type of workloads so they're not very cool heavy to my surprise a thread Ripper scored very well in this test and did considerably better than the core o 970 100 X Excel is another office type benchmark but this one does utilize many threads especially when running the extreme Monte Carlo simulation here the 1920 X just edged out the 17 900 X taking an impressive 1.86 seconds to complete the workload that said it was 12% sold in 1950 X which took just one point 66 seconds making it the fastest desktop CPU we've ever tested here for those of you unaware veracrypt is an open source utility for on-the-fly encryption and features optimized implementations of cryptographic hash functions and ciphers which boosts performance on modern CPUs it also supports paralyzed encryption from multi-core systems as well as hardware accelerated aes to further improve performance in short it takes advantage of the many course thread Ripper has on offer and we certainly see that here when compared to the core I 9 700 X the 1950 X was 48 percent faster for the 50 megabyte test and 35% faster for the one gigabyte test the 1920 X also easily beat the 1700 X 2 living up to 25% more performance next up we have non-encrypted compression and decompression performance using 7-zip whereas hyper-threading sees a similar level of efficiency when compressing and decompressing SMT is significantly more efficient when doing decompression work SMT boost performance of thread Ripper by around 15% for the compression workload whereas its are a massive 60% increase for decompression disabling SMT saw the 1950 X deliver around 50,000 MIPS for both tests however with SMT enabled the 1950 X just beat Bakura 9 some 100 X for the compression test but crushed it by a 53% margin for the decompression test meanwhile the 1920 X matched the compression performance of the 78 20 X but was 48 percent faster when comparing the decompression results overall strong results here for AMD's new thread Ripper CPUs can break is a very popular application for encoding video and we've used it to convert a 4k h.264 video to 1080p using h.265 and recorded the average frame rate the 1950 X was 6% faster than the 70 100x in this test not a huge gain but given both CPUs cost the same amount it's a good result for the thread Ripper CPU that said the 1920 X does cost 3 percent more than the 78 20x and yet it was just 5% faster here so a less impressive result for AMD in that comparison for me personally the Premiere Pro CC results interests me the most as I spend a lot of time rendering these 4k videos on my core i7 69 50x editing machine back when I first reviewed the core I 9 7 100x I said at the time I had planned to probably replace my 69 50 X with for every power I had noted that at the time that the current version of Premiere Pro isn't that good at taking advantage of high coil count CPUs as a result the 700 X was just 4% faster than the 69 50 X so a bit of a disappointing result therefore the Intel CPU I should note that this benchmark is cruder accelerated by using a g6 1080 TI so keep that in mind that said there aren't a huge amount of effects in my videos for the GPU to accelerate give it a higher clock semi 100 X was just 4 percent fire laughter than my 6950 ex and both are technical parts I was hopeful thread Ripper could save me a bit more time well needless to say I was pleasantly surprised when the 1950 ex took just a hundred and twenty eight seconds to render the one minute and thirty second sample video as a twelve percent improvement over the 7900 ex meanwhile the 1920 X was also just seven percent sold in the 700 X so it looks like my next build will indeed be a 1950 x3 wooper CPU moving on to the rendering tests we have blender and first we're running the rising graphic 27 tests here the 1950 X offered a massive 29% performance improvement over the seven 100x in fact the 1920 X was able to best the core on an CPU as well albeit by a slim margin in any case a very strong result here for AMD so let's move on to check out what be more extreme blender workload has for us the gooseberry workload takes some serious firepower to complete and around half an hour and we see this is the 1950 X turns in a time of almost 32 minutes but that's a decent improvement on the 36 minutes it took the core o 9 700 X it's actually around a 13 percent improvement in fact this time the core I known CPU does just beat the 1920 X though the margin was extremely slim Corona comes with a standalone benchmark it renders the fix seen six times and we take the time it takes to complete the entire task this application loves threads the more the merrier and the quicker you'll be done and you can move on to something else here we see yet again that the 1920 X is able to best the core I 9 700 X even if it is only by a slim margin meanwhile the 1950 X races ahead smashing the 700 X by a convincing 31% margin pulverize another ray tracer it's been around for many years and we're using the official benchmark for testing this application here the 1920 X beat to the i9 700 X by a rather convincing 10% margin and that same B's best result yet meanwhile the 1950 X was a whopping 37 percent faster than the core I 9 CPU and a fellow's wondering almost 90 percent faster than the 1800 X this concludes the productivity benchmarks time for a few quick games now before we get into the gaming benchmarks be aware that there are two different memory access modes for thread Ripper and they do impact gaming performance quite a bit by default the CPU is configured to use the distributed mode which uses uma or uniform memory access this method works best for the productivity workloads just seen and since this is the default configuration I decided to test games using this method first but I will also compare uma and numa performance as well so first up we have battlefield one and performance doesn't look great if I'm honest that said it doesn't really look bad either the thread Ripper CPUs deliver similar performance that of the r7 1700 that does however make them quite a bit slower than the sky like X CPUs particularly the 7820 X and 17900 X also please note though for these games I have reduced the quality settings slightly to try and eliminate the GPU bottleneck moving a total war Warhammer we see some pretty ordinary performance particularly in relation to risin 7 it's possible the game will be updated to better support thread Ripper that said though we are pushing over 100 FPS at all times so while comparatively slow when compared to the 1800 X and 700 X thread rippers not exactly slow though the 1950 X performed very well when testing with ashes of the singularity escalation basically matching the korra 970 100x meanwhile the 1920 x wasn't that far behind the 78 20x and both red river cpus were much faster than the rise of 7 models moving on previously we found that risin 7 has proven to be very fast in Civilization 6 and while not as fast thread report also performs very well in this title beating the Intel CPUs the last game we have time for is f1 2016 and here thread ripper slips behind the horizon 7 CPUs and well behind the sky like X CPUs again please note I have reduced the quality settings in game down to high again to try and remove the GPU bottleneck even how we are using a gtx 980ti at 1080p even so with the quality preset reduced to high we are seeing some pretty massive frame rates for the Intel processors ok so as I just mentioned before getting into the game benchmarks that we just looked at there are two different types of memory access modes for every bar the distribute mode which is enabled by default users uniform memory access and again so far all the gaming testing you've seen has used this mode however you can switch to local mode or non-uniform memory access and this allows each of the Zeppelin dice to prioritize which caused access certain parts of the system memory basically prioritizing the nearest cause and this improves overall latency for gaming applications that tend to place a premium on fast memory access so let's see how this impact gaming performance as well as a few productivity workloads switching to Numa has improved the battlefield 1 performance and now thread Ripper is delivering similar framerate to the horizon 7 1800 X in fact the 1950 X creeps they had ever so slightly and now isn't that much slower than the core i7 78 20x and Cora 970 900 X f1 2016 also sees decent gains when switching to Numa though this time it isn't able to match the rise in 7 1800 X and that means it's also well down on the Intel CPUs performance and civilization 6 was already very good and here we see when switching onuma no extra performance was had looking at the application performance we see a slight performance decline with povery nothing serious but the urban method works better here blender also saw a slight decline in performance 1950 X was about 4% slower using Numa lastly we have the corona test and here Numa and Huma enabled the same performance so not much else really needs to be said let's move on to some power consumption figures power consumption is one area where AMD continues to surprise me and I have to say Intel's actually being quite surprising here as well recently though for a totally different reason whereas the 10 core 71 RX push total system consumption to an incredible 281 watts in the corona benchmark the 16 core 1950 X reaches 257 watts a 9% decrease in total system consumption that's a truly impressive result for AMD given the 1950 X was faster and every single one of our productivity workloads some basic knitting and yet it consumes less power meanwhile the 1920 X was extremely fuel-efficient consuming 26 percent less power from the 1950 X and just 18 percent more than the 1800 X it also somehow consumed 15% less power than Intel six-course 70 100x all right let's move on to see how things heat up though once we start overclocking using the thermal tape flowering RGB 360 tt premium edition all in one liquid cooler damn that's a serious mouthful we found some pretty good temperatures with thread Ripper using the factory installed Tim the flow was fitted to the 1950 ex with the CPU overclock to 4 gigahertz for all 16 cores under full load the system peaked at 88 degrees and I actually thought that was pretty good given what we've recently seen from the core I 979 rx which pretty much cooked our all-in-one cooler from Corsair however after removing the cooler and wiping away the Tim we covered the entire copper base in paste and try it again to my surprise temperatures were reduced quite drastically now peaking at 83 and a half degrees but often sitting at just 81 degrees with the custom pace still applied I reverted back to the stock out-of-the-box clock speeds and now the 1950 X was hitting just 71 degrees for those wondering temperatures seem to be much the same with the 1920 X under the same conditions as expected the thread Ripper CPUs were able to hit 4 gigahertz but unfortunately like horizon 7 weren't able to go beyond that point at least not on all cause it's possible with extreme voltages you might hit 4.1 maybe even 4.2 gigahertz like what we've seen with summarize and CPUs but for 24/7 usage 4 gigahertz looks to be the limit anyway at 4 gigahertz the 1950 X spat at an incredible score of three thousand four hundred and eight points in Cinebench though due to the way XFR works this overclock actually slightly reduced the single thread performance because we're not hearing those 4.1 4.2 x FR frequencies on a single core you could probably find chewing the overclock to avoid this but with time against us we didn't play around too much with the overclocking overclock the blender render time for the 1950 X was cut by an impressive 11 percent while the 1920 X was reduced by 6% this means even when overclocked to the core Oh 979 her ex wasn't much faster than the 1920 X and it wasn't able to match the stock 1950 X in this application interestingly this time when testing with Corona the 1920 X saw a massive 17% performance improvement when overclocked while the 1950 X was just 7% faster this meant both overclocked thread ripper parts were much faster than the overclocked 7900 x premier brings us back to the kind of margins we previously saw in blender here we find the 1950 X was 9% faster once overclocked wildly 1920 X was just 5% faster this meant when testing with Premiere Pro CC the 1920 X did trail de couro 970 900 X while the 1950 X was quite a lot faster please note that the overclocking performance in warhammer was tested using uniform memory access and here the overclocks did little to boost performance finally we have the overclocking power consumption figures and here thread Ripper looks quite good I have to say especially in relation to the core I 979 to X whereas the Intel CPU push total system consumption of 388 watts the 1950 X hit 358 watts an 8% reduction for what was often a 20 to 30 percent increase in performance for the price versus performance comparison I've taken the current retail pricing for each CPU from Amazon and you can find those links in the video description first up let's take a look at blender here we see the thread Ripa 1950 X delivering the best performance as it's situated the furthest right are scatterplot as you can see it also sits lower than the core I 9/7 our X making it the cheaper option to give you a better idea of what's going on here is a linear guide from the horizon 7 700 to the thread Ripper 1950 X basically CPUs position to the left of this line represent poorer value while anything to the right is better in terms of price versus performance this gives us a good idea of how things look between AMD's cheaper state core CPU and the new flagship part as you can see the Intel Core i7 7800 X and 7820 X are very competitive while the 1920 X offers the overall value in fact the only CPU to deviate a notable difference from the line is the core own own 7900 X and as it is to the left of the line we find it to be rather poor value next up we have Corona and this benchmark looks to be largely dominated by AMD throwing up our linear line shows that aim D's done a very good job here as all four of their CPUs to live a very similar price versus performance ratio meanwhile Intel's best value offering here is the core i7 78 20x while the 7800 X and 7900 X represent rather poor value the povery price versus performance scatterplot looks remarkably similar to what we saw with Corona let's bring up that line again to show you what I mean again we see that all for AMD process is a quite Titan alone meaning they'll deliver very similar value again only the core i7 7820 X looks half decent from Intel while the 1700 X and 7900 X represent very poor value finally I wanted to check out price versus performance in Premiere Pro CC as you can see the AMD processor again looked to be ahead and bringing up our linear lines certainly shows this to be the case yet again we set the core i7 7820 X does the best of the skylake CPUs while the overpriced 700 X again represents very poor value well now that was a lot of data and pretty much all pointed to one thing total domination by aim DS new thread Ripper CPUs the only weakness was once again gaming but there really is no point in buying thread Ripper for gaming rather if you keen on gaming and you have $1,000 CPU money to play with Pocket the change and get a 7700 K the core i7 77 ok is significantly cheaper and much better suited for high-end gaming in the future end games do use more cause you simply upgrade them to whatever makes sense for gaming the reason I've recommended the horizon 5 1600 over the 7700 K in the past is down of the simple fact that the rise in processor is much cheaper it's also available on a more affordable platform and for 90% of gamers out there it will deliver the exact same experience so in terms of bang for your buck Rises and wins said though as I've also noted previously if you want the ultimate gaming experience with maximum frame rates the 77 okay is the best option having settled that it is worth noting that when it comes to gaming thread Ripper is more than capable and it's likely much better than Intel Xeon Gold 6130 for example a $1,900 u.s. 16 core 32 thread server chip which operates at a base frequency of just 2.1 gigahertz anyway enough about gaming that quite clearly isn't the point of these 12 core and 16 core CPUs productivity is the name of the game here and for serious workloads Verizon has already proven to be a beast thread Ripper though takes things to a new level we now have a $1,000 u.s. processor from AMD but when compared to the competition it might actually be worth that asking price we've had $1000 desktop processors for quite some time now but it's been a very long time since we've seen such an extreme desktop CPU from AMD and yet at this price I do believe they have the best offering in terms of price versus performance in the case of Broadwell II has well the ivy bridge-e and so on Intel did hands down offer the best options for power users for the simple fact that they were the only choice so for the past five plus years Intel is kind of just won by default back when Intel had no competition the pricing of the early extreme edition CPUs did bother me but how can you complain when there's simply no alternative then by the time has well he came around with all kind of just accepted the $1,000 asking price for Intel flagship part and at the time that wasn't 8 core 16 thread beast then Broadwell he came around and well Intel obviously couldn't imagine a world where they would return the competition so they went completely wild with pricing we ended up with a 69 50 X at $1,700 us now those things are very different although skylax is still technically out in front when comparing CPUs of the same core count it doesn't really matter as I said time and time again absolute performance isn't everything and that's not what I focus on with my reviews I'm 100% all about that price versus performance ratio I review things as though I'm going to buy them myself and having recently purchased a core i7 7 100 X and Anna soos x-29 motherboard for testing I'm still feeling the sting let me be clear though the sky like X architecture and the X 299 platform actually isn't that bad yes there are a few head-scratching aspects but overall it does work very well the problem for me is the price or at least the price versus performance in terms of pricing skylake X has certainly improved but as I said in my previous video comparing the 7820 X and r7 1700 we've kind of just gone from stupid pricing to a little less stupid meanwhile AMD's come out of nowhere and decided they actually do want some market share and the strategy to win it back seriously aggressive pricing thread Ripper while certainly a lot more expensive than Verizon 7 kind of continues with the lowball pricing and this is a big problem for Intel in a way the core count trying to become somewhat irrelevant at this point the core o 970 100x and thread ribbon 1950 X both have a lot of course and they both cost $1000 the difference being that the 1950 X was often found delivering 20 to 30 percent more performance while consuming almost 10 percent less power under full load so in a nutshell for the same price you get a heck of a lot more CPU more performance lower power consumption and improved operating temperatures so in conclusion why would you buy a core i 9 700 x over this glued together 1950 X Madden oh but that's Intel's problem Intel did just announce the official specifications for their 12 14 16 and 18 core Skylar hex parts and based on what they announced at a guess I would say consumers will need to spend at least $1,400 us to acquire the i9 79 40 X to match AMD's 1950 X and I seriously doubt the 40% price premium will be worth it the other big issue the X 299 platform faces other than the extreme price is its complete lack of ECC memory support whereas thread Ripper support ECC memory skylake X chips don't and that means anyone serious about their workstation won't even consider Intel high-end desktop platform unless Intel's willing to move on pricing and I'm quite certain they're not I can't see why anyone would invest in the x29 platform and that's my honest opinion who would have thought that in 2017 AMD would come back and completely knock intel off their high-end desktop perch certainly wasn't me but I'm super glad we have some competition back I can't recall the time when it was more exciting to be a PC and through this and on that note I'm going to end this rather long review I'm your host Steve I'll see you guys again real soon because I have put together some simulated thread rippin 1900 X benchmarks you're not going to want to miss those
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.