Honestly impressive how well it performed as an 8 chip 8 core CPU. Obviously communications between cores aren't going to be as good as a single die but I'm amazed at how good they were still. AMD's killing it with their multi chip approach.
He basically made a "frequency optimized" part like the Epyc "F" parts, which run 1 core per CCD but have all the cache still available and run at higher frequencies.
Dishonestly not very impressed how well it performed as a 8 chip 8 core CPU. Not obvious the communications between cores wont be as good as a single pie but also Im not very amazed at how good they were.
That's the problem with AMD's choice of having few cores per CCX. Now at least they increased them to 8 vs 4 in the past, but you should still see that as 8 machines. The inter-core latency makes it a pain for applications that need to share low-latency data. But it's awesome for clouds and virtualization where its sold as portions of a physical CPU.
@@levieux1137 sure for people that do not understand engineering ...it has nothing to do about a problem....AMD has no problem....that is the wy the design works
200k points is amazing to reach especially since previous record was 50k lower! But for me, even more amazing to see that you can reach 150k~ points (previous record levels) on 7995wx on custom water loop and use it for daily renders etc. That's just crazy. If I were someone who makes money on those kind of workloads I would have rushed to build one immediately. 5000$ would be a bargain when the time is money.
I find it really amazing that Adobe continues to not optimize for the AMD chips. I guess at this point they probably won't, since most of the work is being done elsewhere.
Der8auer always asks the stupid questions nobody else does, but that I desperately want to know the answers to. I was literarily looking just last night for benchmarks of someone disabling all but one core per CCD on past threadrippers and found nothing, just to know what the cross core latency penalty was in gaming scenarios. Thank you for testing this :)
@@altgm11 I meant stupid in a playful way, not a demeaning one ;). Just because gaming, while supported, is so far out of this things intended use case that it isnt even listed or discussed officially
Another great informative video :) BTW, I'm pretty sure that time you crashed the CPU into the socket of a very expensive motherboard is the only time you've ever cursed on TH-cam...lol!
@@mrfarts5176 I think this is doubtful unless intel have a masterplan that involves appearing to be unhinged. In fact if it is an AMD false flag operation then that is marketing genius because I always build AMD PCs for people on principle now.
I really like the first person view style of this video, it's cool to see the motherboard and CPUs up close like that! And also the CPUs are a beast too :)
Would have been intresting to see QTGMC de-interlacing and or Davinchi Resolve proformance with raw footage rather then Adobe at this point! Also vhs-decode if you want a truly painful single core test hehe.
I just realized the CPU is pulling over 270A, which is more then my MIG welder or spot welder I have at home can do. And the CPU still survives! That is wild.
Reminds me of when McLaren went after Bugatti Veyron's speed record and Bugatti was like oh,,, guess we'll take the limiter off. Then destroyed them..lol ( for a car analogy)
AMD just never stop impressing me with those computing monsters, especially this one. What's mind-blowing is that Ryzen Threadripper and Epyc CPUs are as powerful as a decent video card now, in term of floating point performance. And I totally agree about the socket, I'd prefer AMD server socket over Intel's - seen Intel's server processor installation, it's a little bit sketchy, leaving you with weird problems like a few RAM memory sticks not detected correctly, or even loss of PCIe interfaces to a few hardware like SSD and video card. AMD's socket is much more secure and I like it best as it made installation of Ryzen Threadripper 1900x in my workstation much easier. I also should check out E-PMD wattmeter as I kinda wanted an easy way to measure the amperage consumption at 12 Volts DC power rail leading to the CPU voltage regulator, just to see how much juice the CPU consumes.
The 7080X will beat 129K with a single click by enabling PBO no need to do a manual OC to get that score. Also, with Zen4 it's EXPO for memory now with full complete official support from AMD. DOCP was originally done by Asus
$5000 price is insane too, so it's not for average user. Your game tests didn't quite show, how many cores was actually used. Was that guy at the end pouring LN2 without any protective gear? Quite insane video indeed.
what protective gear? LN2 is inert, and while cold, the Lidenfrost effect will protect you against it spraying around. Really, handling boiling water is more dangerous.
Would be interesting to see what overclock results are reachable with pbo to raise the power limits and positive or negative voltage offsets, in order to remain in a power budget that can be handle by watercooling for daily use.
problem is that most of the software can't handle so many threads/cores. how many cores were under load with adobe 4k rendering? score should be higher. I think disabling hyperthreading should help to boost performances on many of the used software.
This is in many cases a Windows problem for all the apps that use the default Windows thread scheduler, which still maxes out at just 64x threads per process (why disabling SMT on AMD's >=64-cores often helps in numerous workloads). 🤷 The only way to push past that is to instead write & use your own custom thread scheduler like the one in Blender to actually have a single app process be able to utilize >64x threads, which is a lot of work (and Blender only did this a few years ago, ala when the TR 3990X launched in early 2020 it would only use half the CPU if SMT was on). If you want to use a >≈64-core CPU for SERIOUS work, you should SERIOUSLY consider switching to a Linux based operating system. 🤷 Linux is just simply SOOOOO much better equipped to natively handle super high core/thread count CPU's than Windows! Many workloads are ≈+10-40% faster on Linux vs Windows w/ said 96-core TR 7995WX!
Very much so. This is done by core isolation. So when you have this many cores, if you running windows, you can isolate the ccd through task manager, processes tab. Right click, select set affinity. Turn off cores you want to dedicate to your music production programs on all processes save the music applications. Leaving just those cores on for the music applications. Then turn off automatic windows updates.
@der8auer Would love to see you continuing to include PUBG Esports benchmarks. No one else is benchmarking CPUs for PUBG, especially competitive Esports settings (1728 x 1080).
GN Steve: *complains about how the last Threadripper review people jumped straight to the gaming results Roman: This is not meant for gaming, I know it’s not meant for gaming, but let’s see how it goes anyways
For me it's hard to imagine what it's like to use a powerful CPU like this. I don't do any hard work on my computer, nor do I game. I still use my Ryzen 7 4750G and are somewhat satisfied, but still would like to buy one of the earlier versions of the Threadripper. I've seen a few good prices for only the CPU or a package with the motherboard, CPU and ram. I'm surprised they can be so cheap as a used system.
I was able to afford the bottom level first gen Threadripper with a 1950x, and it's still working great. This generation is completely out of touch for me. The entry level CPU here is nearly 2x the cost of my 1950x. Way WAY too expensive. Give me the 16 core version for $700ish just like my 1950x.
Then just buy a R9 7950X then. 🤷♂ X670E's I/O capabilities are by FAAAAAR the best ever seen on a "mainstream" platform. Do you REALLY need 88x CPU PCIe lanes? O_o
If you're happy with 16 cores (32 threads), why not go with 7950X3D? It has that kind of core count and is faster than Threadripper in single core applications. 14900K is also a good option which has even faster single core performance and very respectable multicore performance.
Other than the cost of entry, it would be tough for me to justify upgrading from my 5800x3d b550x system. I think it would be unreal if someone sent me a board and cpu zen 4 threadripper, but I don't really see that happening. Considering you have the choice, I would run that in my home system (my everyday machine) for a while, just to see the comparisons. I bet if you edit videos it'll go a bit faster, and if you save a good bios for gaming, it's not a big deal to change it up. That's what I do.. One thing I dislike about there almost always (not so much now but before yes) getting BIOS updates.. I would have to re-create all my saves for different things I do with my system. Yes it makes a HUGE difference if you're going to be doing a specific workload or game or whatever, SAVE the bios for that task.. it's a good thing
Despite the instant love benchmarks and temperatures, what amazes me most are those VRMs without cooling that work thousands of watts of power: If I had just heard it I don't think I would have believed it. Rendering: 😍🤩😲🤪
Hallo @der8euer, ich hab vor Jahren meine CPU Übertaktet und das mithilfe deines Videos. In dem Video, hattest du nahe zu die gleichen Komponenten genutzt, jedoch finde ich dieses jetzt nicht mehr und nach einem kompletten Reset würde ich gerne meine CPU wieder übertakten. Jedoch finde ich dieses jetzt nicht mehr. Kannst du mir hier evtl einen Link zu senden? In dem Video müsstest du ein GIGABYTE Z390 GAMING X Mainboard mit einem Intel i5-9600K verwendet haben. Danke für deine Hilfe und deine Videos.
You need to try to disable 6 cores per ccd instead of 7. 16 cores. Leaves hyperthreading on the table with them able to communicate internally and externally. Not just strictly just externally like with 8.
Glad the HEDT market is getting some attention again, now if the XOC market could get some love from some motherboard manufacturers for AM5 I'll be a happy chappy.
If only sapphire rapids wasn't godly slow out the box they might have had a slight chance but even overclocking them hard massively increases the power consumption
It's almost like Sapphire Rapids is an entire process node behind Zen 4, huh? Either way, consumer HEDT matters to far too little people for Intel to really care that much about this. A lot going forward rests on how successful Intel's rapid node advancements go in the coming two and three years. AMD will always be a little bit behind leading edge, but Intel will not have such a disadvantage if their own process tech catches up to TSMC. This has the potential to reverse the whole current situation regarding efficiency. Just potential, mind you. Intel still has a fair bit to prove it can actually deliver.
Competition is good for consumers anytime to bring prices under control. AMD is dominating currently that's why they can jack up the price of the 64 cores SKU from $3990 for a TR 3990X to $4999 for a TR 7980X .
@@benjaminlynch9958 That was the plan before 3nm node setbacks. If you were paying close attention to the matter, it's now considered AMD won't get on it at least until 2025, possibly even 2026.
Other reviewers are saying that threadripper mb have flaws which cause problems for the performance. The sales of threadrippers is very low and why wuuld someone spend $4000 on the cpu plus thousands more on memory, mb and its overkill. When you can Ryzen 7 or 9 and a 1/3 of the cost and get similiar results unless your doing some crazy modelling or complexcalculation it more than enough power.
Intel just has to hope that Zen doesn't scale well (performance & efficiency)as a big.little else they are forever going to be fighting an uphill battle.
Hmm...lets just hope Intel continues to compete, which is the best outcome for all consumers. Intel is still 2 nodes behind Amd and need to stay in business beyond Granite Rapids!!! I was maybe looking to get a xeon w-2595x next year but now definitely looking to get a threadripper 7970x instead.
While I'm not doubting the efficiency let's not pretend this is a gaming beast, it's a monster in a very distinct type of work. Let's see it up against the relevant Xeons
It's literally up to twice as fast in heavily multi-threaded workloads vs the flagship 56-core Sapphire Rapids workstation part... (Usually around ≈+60-80%, but not always). Shit, that same exact Xeon ALREADY lost by ≈-20-50% to the PREVIOUS generation 64-core TR 5995WX!!! 😳🤯 And Intel's chip also is DRAMATICALLY less power efficient, because let's be real here, of COURSE it fucking is... 🤦😑 ..... 🤷 I mean fuck... the 96-core + 8-channel memory TR 7995WX might straight up TRIPLE Sapphire Rapids' performance in some highly multi-threaded workloads!!! 😳🤯
That’s crazy. If I can justify the price tag, we’d have a kick ass server at home, where this HEDT system will have more cores than some of the servers we have at work.
Specific gaming application, dB, I think you showed no game needs more than an 8C monolithic processor but the 1C x 8 ccd multiprocessing example is very interesting. Also in the Intel i9 example the e-cores did not seem to matter in the games u relied for your example? mb
@@Six_Gorillion I agree for many applications e cores do nothing but support multi tasking. Some productivity apps, well compiled thread directed apps will utilize e cores as a SIMD single instruction multiple data and or a parallel array. Not all do but those that do like Adobe suite appear to do so but apps must be compiled to take advantage of e cores as a parallel processing array and most do not. mb
@@mikebruzzone9570 most devs have been too lazy and incompetent to program for multiple cores for years now. Where do in this landscape intels memecores fit in exactly? How are they benefiting the scene in any way? It's marketing wank, pure and simple.
@@Six_Gorillion Adobe all I am confident have been optimized for e core SIMD and or some form of parallel compute. Not the first time Adobe among the few. Most devs too lazy [?] well what about Ryzen 2P never has been explained and well known Epyc multiples of octa multiprocessing and or partitioning. The issue here is AMD does not want to say it 8x1, x2, x4, x6, x4, x12 etc. Why NOT because the tools exist? Most PhD contract coders and hyperscaler don't want that known on their competitive advantage of they know and no one else does. Albeit this eventually gets out from the CS/DS graduate schools. On Intel side where are the SR development cards for W 34/27xx_ mb
Intel has been in trouble for a LONG time... pretty much ever since they failed to execute and advance on their 10nm (now Intel 7) node. They're trying to recover and have a few years to go... we'll see if Pat's vision pans out.
Congrats on the record run, 200k+ is really something.
still remembering hitting 3k and being super happy xD.....
A rare case that the title claims something is "insane", and I'm inclined to believe it before watching.
Yup, TH-cam is full of hyperbolic hyperbole.
Cause debauer makes quality content. I feel the same.
Do you believe that after watching?
I had to watch as derbauer called this insane .. and he's right
So, the CPU killed its father, and then walked around the stage, soaked in his blood, singing "Il dolce voce "?
Honestly impressive how well it performed as an 8 chip 8 core CPU. Obviously communications between cores aren't going to be as good as a single die but I'm amazed at how good they were still. AMD's killing it with their multi chip approach.
He basically made a "frequency optimized" part like the Epyc "F" parts, which run 1 core per CCD but have all the cache still available and run at higher frequencies.
Intel said they don't like HEDT they like Low end desktops.
Dishonestly not very impressed how well it performed as a 8 chip 8 core CPU. Not obvious the communications between cores wont be as good as a single pie but also Im not very amazed at how good they were.
That's the problem with AMD's choice of having few cores per CCX. Now at least they increased them to 8 vs 4 in the past, but you should still see that as 8 machines. The inter-core latency makes it a pain for applications that need to share low-latency data. But it's awesome for clouds and virtualization where its sold as portions of a physical CPU.
@@levieux1137 sure for people that do not understand engineering ...it has nothing to do about a problem....AMD has no problem....that is the wy the design works
That spike to 1.6 Kw, than to 100w was quite funny to watch
I remember feeling pleased at being able to get my Pentium 166 MMX CPU to run at 233MHz. How times have changed!
Crazy how efficient ryzen continues to be.
yoda.....?
That's TSMC for you, just like nvidia's 4000 series GPUs
@@keonxd8918yup..if they change to samsung☠️
I was about to tell everyone exactly why, but I don't give ~10x stock tips to furries.
It's not Ryxen. The thread ripper is based off their server Rome architecture.
200k points is amazing to reach especially since previous record was 50k lower!
But for me, even more amazing to see that you can reach 150k~ points (previous record levels) on 7995wx on custom water loop and use it for daily renders etc.
That's just crazy. If I were someone who makes money on those kind of workloads I would have rushed to build one immediately. 5000$ would be a bargain when the time is money.
I find it really amazing that Adobe continues to not optimize for the AMD chips. I guess at this point they probably won't, since most of the work is being done elsewhere.
Adobe is like M$ (Dollar). Up to no good deeds.
Because GPU is optimized for raster performance over more generic cpus
Der8auer always asks the stupid questions nobody else does, but that I desperately want to know the answers to.
I was literarily looking just last night for benchmarks of someone disabling all but one core per CCD on past threadrippers and found nothing, just to know what the cross core latency penalty was in gaming scenarios.
Thank you for testing this :)
@@altgm11 I meant stupid in a playful way, not a demeaning one ;). Just because gaming, while supported, is so far out of this things intended use case that it isnt even listed or discussed officially
@@altgm11 It is stupid.
@@aidenkhaal3295 any question, born out of curiousity, in my book isnt stupid.. so same here ;)
why? you can do it yourself ....n run the programs....very easy
@@atta1798 yea just spend thousands of dollars to test a stupid idea on this cpu, very easy
"Then I was having some very stupid thoughts again" I can relate man.
Another great informative video :) BTW, I'm pretty sure that time you crashed the CPU into the socket of a very expensive motherboard is the only time you've ever cursed on TH-cam...lol!
well who would have not.
Would yt let him curse in bosch on his english channel 🤔
Userbenchmark HQ explodes in a rage at this tagline
They have to be so pissed. I don't even understand why, either. They must be getting paid by intel or be intel employees.
@@mrfarts5176 I think this is doubtful unless intel have a masterplan that involves appearing to be unhinged.
In fact if it is an AMD false flag operation then that is marketing genius because I always build AMD PCs for people on principle now.
@@reggiedixon2 hehe if amd was behind this it would be incredible.
I really like the first person view style of this video, it's cool to see the motherboard and CPUs up close like that! And also the CPUs are a beast too :)
Would have been intresting to see QTGMC de-interlacing and or Davinchi Resolve proformance with raw footage rather then Adobe at this point!
Also vhs-decode if you want a truly painful single core test hehe.
I really liked that you did gaming benchmarks, not only productivity bullshit and also the 1 core per CCD idea was funny too.
I just realized the CPU is pulling over 270A, which is more then my MIG welder or spot welder I have at home can do. And the CPU still survives! That is wild.
It doesnt pull 270A at all. 270W.
Keep in mind Power is Volts times Amps. Your welder doesn't run at
Wow, that overclock performance is crazy!
What? A 1.6 kilowatt CPU? That is crazy and I want one.
Reminds me of when McLaren went after Bugatti Veyron's speed record and Bugatti was like oh,,, guess we'll take the limiter off. Then destroyed them..lol ( for a car analogy)
I think all nvme's should have the retaining latch . Those screws are just too tiny - LoL !!
OMG!!!! Over 200,000 points!!!! Absolutely insane!!!!
And the CPU only costs 5000 kidneys!!!! Absolute bargain!!!
Excited to see you do direct die or something else crazy to really push it hard
AMD just never stop impressing me with those computing monsters, especially this one. What's mind-blowing is that Ryzen Threadripper and Epyc CPUs are as powerful as a decent video card now, in term of floating point performance.
And I totally agree about the socket, I'd prefer AMD server socket over Intel's - seen Intel's server processor installation, it's a little bit sketchy, leaving you with weird problems like a few RAM memory sticks not detected correctly, or even loss of PCIe interfaces to a few hardware like SSD and video card. AMD's socket is much more secure and I like it best as it made installation of Ryzen Threadripper 1900x in my workstation much easier. I also should check out E-PMD wattmeter as I kinda wanted an easy way to measure the amperage consumption at 12 Volts DC power rail leading to the CPU voltage regulator, just to see how much juice the CPU consumes.
This dual voltage input screams to me that they'll have half 3d chips in the future - or perhaps half dense.
🎉
Congrats on the new record 👍
200k+
The 7080X will beat 129K with a single click by enabling PBO no need to do a manual OC to get that score.
Also, with Zen4 it's EXPO for memory now with full complete official support from AMD. DOCP was originally done by Asus
Why the AMD has always issues with text print orientation on their Zen CPUs? Its real OCD trigger.
wild cpu. amazing to see what it can acheive
$5000 price is insane too, so it's not for average user. Your game tests didn't quite show, how many cores was actually used.
Was that guy at the end pouring LN2 without any protective gear?
Quite insane video indeed.
what protective gear? LN2 is inert, and while cold, the Lidenfrost effect will protect you against it spraying around. Really, handling boiling water is more dangerous.
Would be interesting to see what overclock results are reachable with pbo to raise the power limits and positive or negative voltage offsets, in order to remain in a power budget that can be handle by watercooling for daily use.
Thats 200k is for massive respect 👏👏👏👍👊
Sweet, you and GN have videos up at the same time. Thanks! Have been excited for this launch.
NDA has fallen, thats the reason
Nice! Love your lab, too. I can live what I want through you! Thank you, love your tech videos.
It would be interesting to see otherwise fab failure few core CCDs in such a package, easily cooled by basically anything.
that's basically what some of epyc does. There are even server chips with 1 enabled die per CCD
At 15:48 it looks like the connector on the bottom is on fire.
It really does. Must be some leds down there that were lighting up the vapor.
it just and LED
@EGK20 How can you write so few words and get only half of them right?
problem is that most of the software can't handle so many threads/cores. how many cores were under load with adobe 4k rendering?
score should be higher. I think disabling hyperthreading should help to boost performances on many of the used software.
This is in many cases a Windows problem for all the apps that use the default Windows thread scheduler, which still maxes out at just 64x threads per process (why disabling SMT on AMD's >=64-cores often helps in numerous workloads). 🤷
The only way to push past that is to instead write & use your own custom thread scheduler like the one in Blender to actually have a single app process be able to utilize >64x threads, which is a lot of work (and Blender only did this a few years ago, ala when the TR 3990X launched in early 2020 it would only use half the CPU if SMT was on).
If you want to use a >≈64-core CPU for SERIOUS work, you should SERIOUSLY consider switching to a Linux based operating system. 🤷 Linux is just simply SOOOOO much better equipped to natively handle super high core/thread count CPU's than Windows! Many workloads are ≈+10-40% faster on Linux vs Windows w/ said 96-core TR 7995WX!
This is mind-blowing how further you could push these CPUs 🤯🤯
$5000 CPU, should outperform 2x 14900k's.
When doing extreme overclocking and record hunting, you go with real-time for the CB23 process, right?
NICE!!! 👍 you took a big risk pouring nitrogen to the hardware!
does this have any practical use for music-makers? meaning low latency. Cheers
Very much so. This is done by core isolation. So when you have this many cores, if you running windows, you can isolate the ccd through task manager, processes tab. Right click, select set affinity. Turn off cores you want to dedicate to your music production programs on all processes save the music applications. Leaving just those cores on for the music applications. Then turn off automatic windows updates.
You had the Intel W790 HEDT, why not compare with it in the review?
I thought Intel W790 is a chipset, not a CPU. Which Intel CPU do you think is anywhere close to this AMD monster?
JESU CHRISTO!!!! 201k? Now THAT is some serious AMD E.E. SMACKDOWN !!!!!
Wild!
Nice video, well done, thanks for sharing it :)
@der8auer Would love to see you continuing to include PUBG Esports benchmarks. No one else is benchmarking CPUs for PUBG, especially competitive Esports settings (1728 x 1080).
GN Steve: *complains about how the last Threadripper review people jumped straight to the gaming results
Roman: This is not meant for gaming, I know it’s not meant for gaming, but let’s see how it goes anyways
Grats on getting the record. Can't wait to see the performance for the next gen cpu's for AM5.
Holy shit it absolutely MURDERS that CBr23 render pass.
The insane part to me is it completed it in like 4 seconds.
For me it's hard to imagine what it's like to use a powerful CPU like this. I don't do any hard work on my computer, nor do I game. I still use my Ryzen 7 4750G and are somewhat satisfied, but still would like to buy one of the earlier versions of the Threadripper. I've seen a few good prices for only the CPU or a package with the motherboard, CPU and ram. I'm surprised they can be so cheap as a used system.
That is indeed some powerful performance. I can see that selling!
I was able to afford the bottom level first gen Threadripper with a 1950x, and it's still working great.
This generation is completely out of touch for me. The entry level CPU here is nearly 2x the cost of my 1950x.
Way WAY too expensive. Give me the 16 core version for $700ish just like my 1950x.
But the performance leap is quite astounding, you have to wait for two more generations to be able to afford this current cpu, that's the tech way.
Then just buy a R9 7950X then. 🤷♂ X670E's I/O capabilities are by FAAAAAR the best ever seen on a "mainstream" platform. Do you REALLY need 88x CPU PCIe lanes? O_o
If you're happy with 16 cores (32 threads), why not go with 7950X3D? It has that kind of core count and is faster than Threadripper in single core applications. 14900K is also a good option which has even faster single core performance and very respectable multicore performance.
I'm liking the efficiency improvements. It's been a while since I've seen progress in that department.
Games dont use all cores even on 5900x, can you run xmrig on it and post results?
That thing rips
Amazing stuff!
De-lid the CPU and see if that helps for O/C. That would be an interesting video.
My IBM PC 4.7mhz machine with 640K took a full 15 minutes to boot.
Other than the cost of entry, it would be tough for me to justify upgrading from my 5800x3d b550x system. I think it would be unreal if someone sent me a board and cpu zen 4 threadripper, but I don't really see that happening. Considering you have the choice, I would run that in my home system (my everyday machine) for a while, just to see the comparisons. I bet if you edit videos it'll go a bit faster, and if you save a good bios for gaming, it's not a big deal to change it up. That's what I do.. One thing I dislike about there almost always (not so much now but before yes) getting BIOS updates.. I would have to re-create all my saves for different things I do with my system. Yes it makes a HUGE difference if you're going to be doing a specific workload or game or whatever, SAVE the bios for that task.. it's a good thing
Why is that SSD sticking out... My OCD is off the charts!
It would be cool if they had a separate 78003d built in the thread ripper! So when you want to game you switch to it!
What about the new 128 core RISC-V? be worth a comparison.
Puzzle of a workstation board only having 4 DIMM slots ... next to useless.
Was this compared with Xeon? Probably you mentioned it already and sorry if i missed it.
what's the pcie1x slot above the wifi?
Mr. Roman, I think you need top-down camera pretty soon 🙃
Despite the instant love benchmarks and temperatures, what amazes me most are those VRMs without cooling that work thousands of watts of power: If I had just heard it I don't think I would have believed it.
Rendering: 😍🤩😲🤪
Oh my God! Hope the old Threadripper will be more cheaper now!🤑
Hallo @der8euer,
ich hab vor Jahren meine CPU Übertaktet und das mithilfe deines Videos.
In dem Video, hattest du nahe zu die gleichen Komponenten genutzt, jedoch finde ich dieses jetzt nicht mehr und nach einem kompletten Reset würde ich gerne meine CPU wieder übertakten.
Jedoch finde ich dieses jetzt nicht mehr. Kannst du mir hier evtl einen Link zu senden?
In dem Video müsstest du ein GIGABYTE Z390 GAMING X Mainboard mit einem Intel i5-9600K verwendet haben.
Danke für deine Hilfe und deine Videos.
You need to try to disable 6 cores per ccd instead of 7. 16 cores. Leaves hyperthreading on the table with them able to communicate internally and externally. Not just strictly just externally like with 8.
And use ryzen master to figure out which on is the fastest cores. To get even better.
@der8auer EN Try Kerbal Space Program 2 and build a really really big ship. Bet them cores come in handy?
Seems like R23 needs an update!
Is there a de-lidding / direct die solution for this 7000 series of threadripper
Glad the HEDT market is getting some attention again, now if the XOC market could get some love from some motherboard manufacturers for AM5 I'll be a happy chappy.
Geil......love your "stupid" thoughts, mate.
< 90k R23 @ 350W, oh my LORD
>
So, are we still sure no one will ever be fired for buying Intel?
Steve live streamed with the AMD guys and got close, but couldn't crack 200K. Regardless, CPU speeds are approaching light speed!
If only sapphire rapids wasn't godly slow out the box they might have had a slight chance but even overclocking them hard massively increases the power consumption
It's almost like Sapphire Rapids is an entire process node behind Zen 4, huh? Either way, consumer HEDT matters to far too little people for Intel to really care that much about this. A lot going forward rests on how successful Intel's rapid node advancements go in the coming two and three years. AMD will always be a little bit behind leading edge, but Intel will not have such a disadvantage if their own process tech catches up to TSMC. This has the potential to reverse the whole current situation regarding efficiency. Just potential, mind you. Intel still has a fair bit to prove it can actually deliver.
@@maynardburger Considering AMD will now be stuck on 5nm for quite a while, Intel's chances to retake the node advantage are quite good.
Competition is good for consumers anytime to bring prices under control. AMD is dominating currently that's why they can jack up the price of the 64 cores SKU from $3990 for a TR 3990X to $4999 for a TR 7980X .
@@HenrySomeoneZen 5 parts on N3 node in 2024…
@@benjaminlynch9958 That was the plan before 3nm node setbacks. If you were paying close attention to the matter, it's now considered AMD won't get on it at least until 2025, possibly even 2026.
If Roman says it’s insane in his non-hyped Germanic way, I knew it would be amazing.
Other reviewers are saying that threadripper mb have flaws which cause problems for the performance. The sales of threadrippers is very low and why wuuld someone spend $4000 on the cpu plus thousands more on memory, mb and its overkill. When you can Ryzen 7 or 9 and a 1/3 of the cost and get similiar results unless your doing some crazy modelling or complexcalculation it more than enough power.
Intel just has to hope that Zen doesn't scale well (performance & efficiency)as a big.little else they are forever going to be fighting an uphill battle.
Hmm...lets just hope Intel continues to compete, which is the best outcome for all consumers.
Intel is still 2 nodes behind Amd and need to stay in business beyond Granite Rapids!!!
I was maybe looking to get a xeon w-2595x next year but now definitely looking to get a threadripper 7970x instead.
cant wait to see what Intel glues together to try and compete
Imagine Intel making a small batch of one die 128 core cpu.
Very impressive.
Size matters!
Enermax AIO. You are a brave soul :P
You're my AMD Hero!
this ok for doom?
While I'm not doubting the efficiency let's not pretend this is a gaming beast, it's a monster in a very distinct type of work. Let's see it up against the relevant Xeons
It's literally up to twice as fast in heavily multi-threaded workloads vs the flagship 56-core Sapphire Rapids workstation part... (Usually around ≈+60-80%, but not always). Shit, that same exact Xeon ALREADY lost by ≈-20-50% to the PREVIOUS generation 64-core TR 5995WX!!! 😳🤯
And Intel's chip also is DRAMATICALLY less power efficient, because let's be real here, of COURSE it fucking is... 🤦😑 ..... 🤷
I mean fuck... the 96-core + 8-channel memory TR 7995WX might straight up TRIPLE Sapphire Rapids' performance in some highly multi-threaded workloads!!! 😳🤯
@@Cooe. wow, I'm looking forward to seeing those tests.
The latest Threadripper is a great chip but when you look at the price it's a niche product.
That’s crazy. If I can justify the price tag, we’d have a kick ass server at home, where this HEDT system will have more cores than some of the servers we have at work.
can you try disable all the chiplets except 1 with all cores on it enabled and test in game? I wonder if it would be better.
How many times faster is this compared to 7980XE
5:38 that was the excuse for the AMD eng to get a coffee time, " look Boss the board is still booting" , hehehhehehehehehheheheheheh
I wonder how my first ever PC would perform in comparison.... probably get a cinebench score of 0.001
Specific gaming application, dB, I think you showed no game needs more than an 8C monolithic processor but the 1C x 8 ccd multiprocessing example is very interesting. Also in the Intel i9 example the e-cores did not seem to matter in the games u relied for your example? mb
e cores are meme cores. marketing wank and utterly useless for a gamer. waste of sand.
@@Six_Gorillion I agree for many applications e cores do nothing but support multi tasking. Some productivity apps, well compiled thread directed apps will utilize e cores as a SIMD single instruction multiple data and or a parallel array. Not all do but those that do like Adobe suite appear to do so but apps must be compiled to take advantage of e cores as a parallel processing array and most do not. mb
@@mikebruzzone9570 most devs have been too lazy and incompetent to program for multiple cores for years now. Where do in this landscape intels memecores fit in exactly? How are they benefiting the scene in any way?
It's marketing wank, pure and simple.
@@Six_Gorillion Adobe all I am confident have been optimized for e core SIMD and or some form of parallel compute. Not the first time Adobe among the few. Most devs too lazy [?] well what about Ryzen 2P never has been explained and well known Epyc multiples of octa multiprocessing and or partitioning. The issue here is AMD does not want to say it 8x1, x2, x4, x6, x4, x12 etc. Why NOT because the tools exist? Most PhD contract coders and hyperscaler don't want that known on their competitive advantage of they know and no one else does. Albeit this eventually gets out from the CS/DS graduate schools. On Intel side where are the SR development cards for W 34/27xx_ mb
Why does your charts for performance say ram was at 6800 c34 but the bios screen you show say ram is set to 6400
Intel has been in trouble for a LONG time... pretty much ever since they failed to execute and advance on their 10nm (now Intel 7) node. They're trying to recover and have a few years to go... we'll see if Pat's vision pans out.
Damn thats a ton of VRAM on that motherboard
This must require some kind of monster power supply to run it! I see two mainboard power connecters?
What are these faces? What do they do? Also, what is CCD? Normally, that is a charge-coupled device (a camera)?