hi and welcome back to a new video in October I was visiting Asus in Taiwan to take a closer look at the new AMD thread rer CPUs with the recent AMD thread rer CPUs AMD now also has CPUs for the workstation area with a sen 4 based cores and to be more specific today we are going to look especially at the 79 adx so the 64 core AMD thread Ripper for launch there will be multiple CPUs starting with 24 cores 32 cores 64 96 cores and also different versions when it comes to memory channels the highest will be the 96 core AMD thread rer 7995 WX which also features eight memory channels that's what you can also identify by the WX in the name because the CPUs without the WX for example the 7980x will only feature for memory channels but with all the experience I could gain during this week it doesn't really really matter as much so also the big CPUs work totally fine only running four memory channels only maybe if you have a very memory intensive workload very similar already to Intel's Sapphire Rapids you may only use RDM ddr5 modules with ECC which will make the entrance cost into the platform a bit more expensive especially if you're looking at high memory speeds but let's just start with a quick look at the mainboard and we are using a very early version of the Asus Pro WS tier x50 Sage WiFi fire okay and as you can see the board is pretty much empty when it comes to any kind of cooling or covers we can only see a small aluminum block on the bottom right where the chipset is underneath and also on the bottom left where we have a 10g network chip but I have to say that the chipset is very very efficient so even without the cooling block without the aluminum block it was working perfectly fine so when I used it without the heat sink and I touched the chipset it was still very cold the voltage supply of this mod board is extremely interesting and I'm pretty sure you straight noticed how the socket is fully surrounded by power stages and inductor we're counting a total of 43 faces around the socket on the left side we can count 23 and on the right side we can count 20 Faces and also there you can read that it's signed with thestral moss W and thestral is the code name of the motherboard and Moss W is the orientation so just west of the socket and out of those 43 faces if we look at the upper region so in the center of the socket we can count four for S so and three faces for the memory voltage of the CPU or the memory controller and underneath the socket we have four faces that are sitting a little bit more separate and those are for vdd iio unlike am5 which is only using a single core voltage this CPU is supplied by two core voltages which is pretty interesting which is mainly caused by the high amount of ccds for example if we look at am5 and the 7 950x it only consists out of two ccds whereas if we look at the big thread rer the 96 core it contains out of 12 ccds and this CPU right here the 64 core with eight active ccds which is just a huge amount of chiplets and also covers a huge surface so it's just more precise to split up the voltage Supply into two core voltages and also allows pretty interesting configurations generally speaking though the CPUs are not that much different from previous thread rer generations and also if you look at the sp6 socket mechanically speaking it is I would say almost identical to the first AMD thread rer which I have very good yeah experiences with say it like that at this point though I have to point out that I'm more a friend of this AMD socket and solution compared to the Intel solution simply because I know that the CPU is sitting firmly and safe in the socket and it's not sticking to the heat sink which just happens way too too often on the sapphire Rapids uh platform and yeah it just happened to me too many times that I'm accidentally bending some pins so I prefer this solution as long as the carrier frame doesn't break so I started mounting the CPU in the socket added some memory and also mounted the AO to start with my testing in addition we also added an Elm more PMD to the motherboard which allows to check the power consumption during some bench marks or tests so we can also take a look at that and as usual with AMD or current AMD Generations the first boot just takes forever and on this platform even longer that what you might be used to with am5 if you populate all memory slots I also checked the time for fun on first Boot and I had to wait 4 and a half minutes to be able to access the BIOS the first time now obviously if you're working in the mboard development department you have to undergo this procedure many many times which also led to Asus to come up with a special trick especially for extreme overclockers so if you enable the ln2 mode there is also an additional reserved switch on the bottom of the board and if you flip this three out of the four memory channels will be disabled and this drastically decreases the boot time and now think about it I'm only using the quad Channel main board there's also the OCTA chennel board and if you disable seven out of the eight channels it will definitely decrease your boot time with all this initial memory drama in mind I was honestly surprised how well this platform can handle High memory speeds I just loaded this XMP profile or in this case called dop with 6,400 MHz and it only took one and a half minutes to boot and that's the second boot obviously but checking in Windows you can see 6400 was instantly applied and even 2,000 MHz Infinity fabric clock so AMD definitely did some good work here when it comes to memory clock so what's the first thing you would do with such a heavy multi-threading CPU obviously just run cinebench R23 now 98,000 points stock that is pretty impressive and that's more than 24900 K combined if you could add this up and then again if you compare the power consumption the 7980x only consumed 350 watt for 98,000 points now think about it a single 4900 K gets 41,000 points but consumes 330 watt so almost the same power consumption but it's delivering more than twice the performance when it comes to mul threading that's just extremely impressive how efficient this CPU is running due to the very low core voltage of about 800 Mt under load you can also see why the CPU stays that cold with about 50 to 70° cus depending on CCD and load but also looking at cinebench we see a clock of about 3.8 to 4 GHz underload in cinebench but also if you check Hardware info you can see that there is much more potential in the CPU because on six of the cords we can see a boost of 5. 6 GHz and on all other cores we saw a boost of up to 5.3 GHz with those boost clocks in mind I of course asked myself the question how will this CPU perform in gaming and even though this CPU might primarily not be made for gaming I thought it would just perform pretty well that's why we're looking at Remnant 2 and in this game I figured out that the CPU also performs pretty well here and we can see low temperatures about 45° C under load while the CPU boosts to about 4.8 GHz across the loaded course the rest is idling at a low 2.7 GHz but I also want to point out there is not even a way for me to display all of the cores here at the same time and that's just a typical problem with having 64 cores AKA 128 threats in this benchmarking scene we can achieve 110 FPS on average and 65 FPS in the 1% low this was tested in 1080P resolution and ultra settings with this setup we are about 10 10 to 15 FPS behind a 7 950x 3D and about 25% behind a 7,800 x 3D just looking at the minimum FPS but you should keep in mind that we are testing in pretty much the worst case scenario that means 1080p with 4090 which results in CPU bottlenecking and you have to keep in mind that most of the time if you would have this setup with a 4090 you would probably play in 1440p resolution at least or probably even 4K and in the higher resolutions you will not be able to tell a difference between this CPU or any recent AMD desktop CPU that's why I also tested Assassin's Creed Valhalla in 4k we can see 74 FPS in minimum and 114 FPS on average and that is absolutely playable and only slightly behind the recent AMD desktop CPUs also I want to point out that the power consumption during Remnant 2 was on average 144 watt while this is about about twice as much as the recent AMD desktop CPUs for example a 7,800 x 3D it is still extremely efficient if you take into account how many cores this CPU has because this is still less than for example 14 900k out of my own personal interest I also decided to run some Adobe Premier rendering benchmarks because at this point I was so surprised by how well this platform worked that I'm still considering to buy this for my personal PC maybe project irrationality 2.0 not quite sure about that so I tested Adobe Premiere and I finished the 4K video rendering with 197 seconds which is the same performance as a 12900 K and as I already stated in the 14900 K review today most of the work is done with the GPU if you render with Adobe Premiere so it doesn't really matter as much but it performs overall pretty nicely at this point I was also thinking maybe there is a way how I can optimize the CPU further and maybe also over clock it for gaming so I went into BIOS and theoretically you can overclock this CPU per CCX or the individual chiplets as I pointed out multiple times the 7980x consists out of eight active ccds and you can clock them individually which is pretty easy to do for example if we go to bios and clock one to 5 GHz but then again if we go to Windows you can see that you are left with the typical AMD problems if you go into the so-called OC mode this means that you lose all kind of boost functions on all the other chips and you will also lose all kind of power saving features without further adjustments you will actually lose performance if you do what I just did you can see we're losing about 12K points in cine bench and also the CPU is now consuming a ton of power we can see over 500 wat in cinebench which is caused by the higher voltage so I went back into the BIOS to adjust my overclock you might remember from earlier when I was talking about the split up voltage Supply that you have for the left and the right region of the CPU two individual voltage supplies this is what we're going to use now and that's still a disadvantage over Intel's Sapphire Rapids platform because there you can clock every core individually and also individually Supply an individual voltage for each core which is quite amazing but here we are left with two voltage supplies so vid Zer and vid1 for half of the CPU each and then I did some further testing and decided to Supply 0.9 volt to one half of the course with 4 GHz and then the other half of the CPU is going to be supplied with 1.22 volt which is able to run 5.2 GHz across the other 32 cores back in Windows we can see a solid performance increase to 115,000 points thus an improvement of about 20% when it comes to Performance and we are still in a very acceptable temperature region as you can see the quick course with a high overclock of 5 5.2 GHz reach about 80 to 90° cius and the slower cords with 4 GHz still sit about 50° C what's still quite insane though is the power consumption because we're drawing about 500 watts during cine bench so now if you're doing this kind of overclock and you want to run some real render Benchmark or maybe some render load that takes longer than a few seconds then I would advise using something else than an AO because surely this will be saturated quickly I guess custom water cooling would be the better option with this overclocking setting in Remnant 2 we can reach 96 FPS in the 1% low and 117 FPS on average and that is a perfect setting for any kind of daily gaming especially if you consider that only 1080p resolution and also if you look at the power consumption it's only 161 watt which is impressive because stock was 144 and this is still with the manual overclocking very efficient absolutely impressive obviously we can also clock all the 64 C of this 79 adx to 5.2 GHz for this I had to slightly increase the vid of rail 1 so we running 1.22 volt on the one half and 1.23 volt on the other half and this results in a cine bench score of 128,000 points that is really impressive it's about 30% Improvement that you can easily do so now keep in mind that an Intel 34.95 x 56 core can also achieve this score but it needs liquid nitrogen and there is no way that you can do it with water cooling and for sure not with aiio then I was having some very stupid thoughts again because the 7980x consists out of eight active ccds and I thought why not just disable seven out of the eight cores per CCD and basically create the biggest eight core in existence and that is possible with a slightly hidden menu in AMD overclocking there is something called bitmap which allows you to disable or enable the course per CCD and now it's getting really interesting so just think about it we now have this massive CPU that runs with one active core per chip so we have a huge area to dissipate the heat so I said 1.3 volt and 5.5 GHz because I thought that should work right now check out the temperatures we can only see 60° C and with this clock and voltage that's pretty cool if we compare the performance it is above a 7,700 X that also runs with eight chords but eight course on a single chip and with additional tuning I was able to run all those ccds between 5.5 and 5.7 GHz depending on the individual chip quality and as expected CCD Z was running the best I was able to reach 20,900 points with this but what about gaming interestingly the performance did not really change at least compared to our manual 5.2 GHz overclock we are sitting at 71 FPS in 1% low and 121 FPS on average you can probably explain this because we are now forcing the CPU to use the chiplets and they have to communicate with each other whereas if we are running stock condition or the manual 5.2 GHz from earlier only the CCD zero is used so the basically the entire game runs on one chiplet whereas now we force the CPU to use all the chiplets to communicate with each other which also introduces additional latency between the chips but considering that it was working working amazingly well of course I was visiting Asus mainly to also do some extreme overclocking on the side which I also wanted to do with the new AMD thread rer CPU but you have to keep in mind we have S four cores here and there were no changes expected over maybe am5 and we know from am5 that around 6 GHz most of the CPUs max out and that's what we also can see here with the new AMD thread rer I was able to clock the CPU to 5.9 GHz at 1.5 volt and this way Achi aeve a score in C bench of 140,000 points at this point also keep in mind that if I'm doing this kind of stuff right here I'm also using OBS screen recording in the background which usually seals a few perc of the performance so keep that in mind if you ever compare my performance numbers by the way in this state the CPU consumed 930 wat power and also to sum this all up if we go back to gaming in Remnant 2 on average with the 5.9 GHz overclock we can see 132 FPS and also 82 FPS in the 1% low at the end of the week Elmore also teamed up with shamino to run the 96 core AMD threat Trooper on an octar Channel motherboard so with eight dims to try to achieve a new cine bench record and especially considering the base performance of this platform with 96 core we quickly knew that we should aim for 200,000 points above 190,000 points was pretty easy to do at about 5.9 GHz but above that and closing into 200 ,000 points not so easy to be honest what I found personally quite impressive though is that cinan R23 at this speed and this amount of course only takes about 3 seconds if we take a look at elmor's PMD while running The Benchmark at these kind of settings we can observe power consumptions around 1,600 watts and this also explains why those boards allow to use two psus and in this kind of state that's definitely recommended because if we observe 1,600 wat on a PM the spikes in power consumption will definitely be higher meanwhile Elmore and shamino had to set up the system multiple times because you know at those temperatures things just don't work out every single time you have a lot of condensation and everything but eventually they were able to break the record over 200,000 points and honestly I'm really sorry for the bad focus in exactly this moment but you can see in benchmate that we have 2,51 points and the CPU was running between 5.9 and 6 GHz for this result I also want to point out at the time shooting this video the current record in cinebench R23 was at about 150,000 points and keeping that in mind it's insane what kind of performance leap AMD was able to provide with this generation honestly flying to Taiwan I was not sure what to expect because typically I have mixed relations with AMD thread rer they were yeah very annoying when it comes to memory tuning I was often not a fan of the socket and so on but this generation really surprised me it worked so extremely well I had a ton of fun with this platform and I might use this even as my new daily that's still something I have to consider all right that's it for this video thanks for tuning in see you next time [Music] bye-bye
