Intel needs a MIRACLE to stay relevant now...

so remember a while ago when uh we did that funny little clip of nvidia just pushing amd down before the fight ever got started [Applause] i feel like in this fight we've got a little bit of a revision to that that we have to make so for those of you that i know are probably not going to watch the entire review and are just going to skip ahead let me just go ahead and give you the too long didn't watch intel wasn't ready [Music] normally i would tell you all about how world of warships is a multiplayer online game that's absolutely free to play with millions of players that you get to take control of some of the most iconic warships in world history and how the graphics are beautiful and has a completely revamped graphics engine and all sorts of different play styles but i got something more important to tell you about today new players that sign up using my link in the description below are going to get a pretty awesome starter pack you'll get 700 blooms 1 million credits seven days of premium time the japanese ship ishizuchi and the uss charleston now with that pack you'll get off to a pretty good start but not good enough to take me on so now that i've done this ad i'm gonna go back to playing world of warships while you go ahead and watch this video [Music] all right so amd's 5000 series cpus built on zen's or amd's brand new zen3 architecture uh we got a few things to talk about here we've got some gaming benchmarks we got some cpu benchmarks it's interesting cpu testing is always really difficult and the reason for that is a lot of things have been offloaded to gpus these days a lot of people have leveraged gpu performance and don't do a whole lot of things on cpu anymore so a lot of the tests that you're seeing right here are tests that could also be done on gpu but we need a way to compare them so obviously we're talking about uh amd versus intel because that's the only two players in the game right now and it quite honestly seems like one of these players didn't really show up to the game if you want to know the truth this is so much fun actually i love when a big upset like this happens 2017 ryzen came and intel made a statement yeah but the cpu's just glued you just glued together you glue i wish intel would glue something together at this point maybe something lower than 14 nanometer plus cubed i mean i'm telling you right now it's a little embarrassing that intel is sending out emails and blog posts and all these things quite literally doing damage control trying to get people to remember that they still are around and they still exist because the whole glue that they talked about is how we got to where we are today with amd and i'm not going to do a deep dive into the architecture but what i am going to talk about is some of the changes that they made which is why we're seeing such massive ipc improvements as well as massive just overall improvements the glue they're referring to is the infinity fabric and the infinity fabric is just the interconnect between the chiplets now amd has really invested into this whole chiplet design where you have basically a core structure and then you have a cache that's shared between them and then you can scale up or down the size of the cpu based on how many of those uh components that you basically stack up inside the cpu so or inside the entire substrate i should say and so that is the glue that intel is referring to so you glued the irony is that they've been gluing xeon cpus in that very same way for a while actually so the glue that intel is referring to is the infinity fabric and that's the way that the interconnect works between the chiplet designs so what amd has effectively done has allowed them to have almost like an infinite scaling infinity fabric if it okay not related but almost an infinite scaling in that you could have your your basic four core four thread cpus like you saw back at zen one which is just four cores and a single chiplet or like threadripper where they just have a ton of them going down the entire substrate all having their own interconnects and l3 cache and all that sort of stuff the problem with that was the way the design initially was with zen one and zen 1.5 although 1.5 they really increased the bandwidth and the bus between the chiplets was the fact that l3 cache was divided so you would have the different cores accessing their l3 cache but they didn't have access to the entire amount of the l3 cache and that unfortunately created a lot of latency and a lot of trade-off in the way that that cache gets refreshed especially when you hand off from one ccx to another now ccx is just basically those chiplets that are all created on this one little city of cpus if you will or coors that's referred to as a ccx the l3 cache was the limiting factor in terms of the bandwidth and the overall speed of the cpu so what has happened here with zen3 is we've seen now eight cores per ccx and the entire amount of l3 cache per ccx so now you're getting full access to the entire cache and that's one of the ways that we're seeing the improvements that we're seeing with especially ipc because now you have a much fatter pipeline happening between them the infinity fabric design continues to improve in terms of reducing latency between the ccx's and then just well i guess the manufacturing process just getting better as a whole because remember they are on tsmc's seven nanometer let's just go ahead and show you guys the benchmarks now that you kind of understand what sort of changed over time what we're comparing right here is the intel 10900k to the amd 5900x and 5950x and the reason why i'm not showing a 10 700k here is because of the fact that basically the 5900x and the 100k are almost dollar for dollar comparable and then the 5950x being really the best offering on mainstream cpus right now so let's go ahead and show you guys those benchmarks which realistically just looks like a digital version of a prison yard beatdown [Music] [Applause] [Music] [Music] [Music] [Applause] so there's something funny to talk about with this chart you got i mean okay it's no it's just sad honestly it it shows that intel's aging 14 nanometer process is just that old and tired and even though it has a clock speed advantage it is obvious with the ipc improvements in or almost at intel god things i'm used to saying with the ipc improvements that amd has pulled out with this manufacturing process with the bumping clock speed because that's the thing i said during all the previous zen launches they need clock speed to go with this ipc well they've got it now because here's the thing the 5900 x goes 4.3 all core turbo clock the intel goes 4.7 to 4.8 all core turbo clock the 5950x also is 4.3 at least that's what ours was doing and it has up to a 4.9 gigahertz single core performance as well as the 5900x having a 4.8 so now you take the and and let's talk about the way these clock behaviors behave first of all up two is a very important statement to see when you when you read about these clock speeds on the box and says up to the 5950x says up to 4.9 gigahertz but i observed more than 4.9 many times even it hitting 4.97 the other thing to mention here is even though the 10900k says up to 5.3 gigahertz it rarely stays there it spends a lot more time around 5152 on single threaded performance now the single threaded performance doesn't have to deal with the whole turbo timer that you get on intel's 10th gen however anything all core that takes longer than 56 seconds does then immediately start to turbo down now yes these are functions that you can go in and mess around with an intel to keep the clock speeds up however it's obvious that we didn't do that with any of these systems because this is the out of the box testing that we do and if we were to do that on intel there are tweaks and stuff we can do on amd to give it the same type of benefit so what you saw here is 100 out of the box performance plopped into an asus motherboard and an msi motherboard both of these being top tier class motherboards in their respective genres here in terms of you know amd and intel no sort of shenanigans happening with the asus or the msi trying to do any sort of turbo clocking or overclocking or turbo timers everything was set to their factory defaults for the cpu remember there's a difference between asus factory default and intel factory default so we made sure that all those things were properly set so not only did we see a stupid amount of ipc improvement when we took the clock speeds locked them at four gigahertz and then did a single core test when you take a look at the ipc test it's just sad i mean we're talking 100 points difference between the 5950x and the 10900k core for core at 4 gigahertz and then it was only one point lower on the 5900x that's because they're the same core that's just margin of error that one point we were seeing the four gigahertz locked ipc test barely being behind the intel unlocked test what do i mean by that well the intel single core cinebench r20 run was a score of 526 and that's with it being allowed to boost up to the up to 5.3 but when we locked amd at 4 gigahertz it was a 5 18 and a 5 19. i mean that's how far zen has come but when it comes to gaming though we tested 1080p and 1440 because we need to see what happens with the bottlenecks so what we did to basically make this a level playing field is we took the same msi gamix trio rtx 3090 and put them on each on each bench we did not use two identical cards because those cards can have variants between themselves so that wouldn't be a very fair test we maxed the boost clock or not the boost but we max the fan curve 100 fans we max the power limit and we max the temperature limit to reduce any sort of p-state changes that way it'll hit the maximum boost table and then we did our gaming test set a mixture of high and medium settings artificially bumping up the fps because at 1080p more fps any sort of cpu bottleneck is going to show we also did 1440p that way we could see what happens when you step up the resolution because although 1080p is the dominant resolution in the steam survey i have a hard time believing anyone that would have an rtx 3090 and a 5950x 5900x or even a 10900k is going to have a basic 1080p panel so you might be running 1080p high refresh rate but 1440p now is so much more affordable i have to believe people with this level of hardware would probably be running something higher than 1080p the butt that i mentioned earlier but you give up some fps and gaming is now applied to intel because intel across the board at 1080p was lower fps now far cry 5 is a title worth mentioning here though they're very similar but that's a very cpu-bound title and that's one reason why we like it because it does a lot of cpu physics happening alongside the test so the cpu being loaded by having to do something besides just throw ping pong balls at a wall you know as a world thread it actually has to do stuff 166 to 177 fps of the 5900x 180 with the 5950. and the reason why we saw such a big bump with 59.50 there is because of the fact that again cpu test now it's worth mentioning though certain titles even though we're only showing three right here do have a slightly higher minimum fps on intel but a higher average and a higher max to be found on amd sometimes so they're very close in terms of that performance min max but the average is higher on amd that's not something i was really expecting going into this now there's one other thing worth talking about here that i think surprised me when i first set this up i'm going through my tests and i'm seeing the blender score and i'm seeing the cinebench score and i'm seeing you know the fishy cat and all that sort of stuff i was like man i was expecting a little bit more performance out of 59.50 it's not bad it's beating the intel for sure i was like but at 250 dollars more 200 250 more i was expecting a bigger gap and phil and i were talking about this and then i looked at the boxes and then i realized the 5950x was still in the box and i went wait was this the 5900x the entire time and it was so the huge gaps we saw were the 5900x versus intel at the same price yes i'm aware it has two more cores and four more threads however the most directly comparable spec you can possibly ever test is price and that's exactly how these two stack up i have to imagine though that a 10 core 20 thread with the same clock speeds in the same ipc is still going to beat this intel in all of those tests i have no reason to believe it wouldn't so once we put the 5950x on there then we saw holy cow this is like a pod boss beating down a newcomer prison talk okay i'm hardened i'm a hardened cpu tester shut up stop laughing at me phil i've been watching 60 days in okay intel just got rolled out what did you get in there is an anomaly worth talking about here though uh premiere uh and and some games just we've known this though for a fact especially for a while now even with x299 and threadripper not a lot of games like a lot of cores so we think what's happening here with the 16 core 32 thread 5950x yeah 16 cores 32 threads on that substrate remember how big those had to be in the past now you're getting that at way less than red river cost when it first came out and don't put that against the 1950 because the threadripper owner will just cry themselves to sleep for a week straight if you compare the price and the performance four or five years later premiere for some reason was slower the 1550 x was slower at premiere we're not too sure why and we think that this has that's adobe being adobe anyone that uses adobe products on a regular basis just knows what adobe doping means okay and it is what it is um in terms of games same thing not all games like a lot of cores the 5900x is quite honestly the best cpu we feel you could possibly buy right now at this time and day for gaming content creation productivity value it's all there everything anyone in my opinion has ever asked for and wanted in a cpu is there because before you had to make that trade-off gaming or multi-threading there's no point there's no reason i built that thing back there and i'm telling you right now i'm conflicted it's got a 10-100 k and i know for a fact if i throw the 5900 x in there and redo that system it will go to waste because i don't use it i want to build myself a new system for the office with something out of that cpu that's got a cool theme or something going on to it because i've got a threadripper system in there now that's also going to waste but i can't in good conscience take this cpu and throw it in my personal gaming rig which i know won't get the use that it deserves so i'm okay with leaving intel in there i made a prediction that these were probably going to be faster and it would go to waste if i put it in there and that's absolutely true i now know i'm leaving a little bit of performance on the table with the 3090 being in there but at the end of the day it is what it is amd is top dog intel i don't believe has an answer you know why because the most previous or the most recent statement said their 11th gen is 10 nanometer ported back to 14 plus what the frame what does that even mean

Channel: JayzTwoCents

Views: 948,788

Keywords: zen 3, amd, amd zen 3, amd 5950x, 5900x, amd 5900x, ryzen, ryzen vs intel, i9, intel, intel vs amd, amd vs intel, 10900k vs 5900x, is amd worth it, is amd good, is ryzen 3 good, amd 5800x, amd 5600x, best gaming cpu, best gaming processor, best processor for gaming, best gaming cpu 2020, best cpu in 2020, 5900x benchmark, 5950x benchmark

Id: hRfumXx-Plo

Length: 17min 25sec (1045 seconds)

Published: Thu Nov 05 2020