Zen 2 - Even More EPYC!
Vložit
- čas přidán 31. 03. 2019
- ♥ Subscribe To AdoredTV - bit.ly/1J7020P
► Support AdoredTV through Patreon / adoredtv ◄
Bitcoin Address - 1HuL9vN6Sgk4LqAS1AS6GexJoKNgoXFLEX
Ethereum Address - 0xB3535135b69EeE166fEc5021De725502911D9fd2
♥ Buy PC Parts from Amazon below.
♥ NEW USA Store! - www.amazon.com/shop/adoredtv
♥ Canada - amzn.to/2ppgYsX
♥ UK - amzn.to/2fUdvU7
♥ Germany - amzn.to/2p1lX6r
♥ France - amzn.to/2oUAK2Z
♥ Italy - amzn.to/2p37Uui
♥ Spain - amzn.to/2p3oIBm
♥ Australia - amzn.to/2uRTYb7
♥ India - amzn.to/2RgoWmj
♥ Want to help with Video Titles and Subtitles?
czcams.com/users/timedtext_cs_p...
-- Video Links Below --
HPC Event at Rice - • High Performance Compu... - Věda a technologie
No it's not an April Fools joke, if it were I'd have said something really outrageous, this is just mostly confirmation of previous leaks.
I had intended to do an April Fools video but had no time as usual! Also had I done that one I would have made it in a faux British accent or something...
Thank god this comment was already here, I was worried.
*sigh of relief*
"I had intended to do an April Fools video but had no time as usual!" Thank God. There are far too many of those already.
Not the best timing, eh? Maybe you should've waited a day. Good video, though :p
Didn't realise that you were putting on a British accent in all of your videos!
Yeah.. risky day to release a video, but at least you have an excuse now if you're wrong on your claims.
8 months from now: "Haha... Got ya!... April fools!"
LOL, Jim pulling out the option select
The real April Fools joke is giving people correct information when they expect you to lie.
Uh ? @hairypatchboy
AMD's law: The Number Of CPU Chiplets Doubles Every Two Years.
LMAO yes! :P
lmao !
lol add vertical stacking as well.
I want a CPU CUBE! that would be like thermal splosion.
@@yulusleonard985
"add vertical stacking as well."
That's a future law that will be in effect when adding chiplets to a single layer reaches its limits. Maybe that law will be stated this way: "The number of chiplet layers double every 2 years" . You afre going ahead of the yourself and everybody else. Lets do one law at a time.
Intel: Moore's law is dead, we can't cope with it
AMD: we are gonna crush Moore's law and make it cry
...and make it our bitch ;P
MOORE POWAH! :D
stop.
Intel isn't wrong. Zen 2 is not a huge breakthrough in transistor size, which is the subject of Moore's law. AMD has just been a lot more creative in dealing with finding other avenues of improvement
😂 😂 😂 Rip intel
Zen 2 rumor vids by Adored are my drug; I don't care if it's just confirmation of previous leaks. You could confirm leaks every day and I would still watch. I am beginning to think I am not a normal human being.
tech is finally exciting again....
now if only GPUs could be this revolutionary instead of making physx 2.0....
@Demicore : We're both abnormal, then.
That's one hell of a date to release a serious video
That number plate. Goddam genius. How did the whole of tech press miss that. Just like with CES, Lisa Su is playing subtle but fun games. Seek and and you shall find.
Pretty sure after the Borderlands 3 teaser video, more and more of these videos are going to be over-analyzed looking for Easter Eggs. ;)
i honestly think it's bait. amd was clearly not showing their best there, but they didn't want to make it obvious so they pretend it's an r7, ya see
@@ikbendusan Thing is, we already know from the leaks that there will be an R9, and that R7 will have an 8-core.
@@ikbendusan if you use common sense and look at amds marketing schemes.....
intel i3/i5/i7/i9
amd r3/r5/r7 but we have no r9, which means new chips can be r9, so 12/16 core will be r9. its common sense.
intel z370 vs amd x370
its all marketing. i just hope navi wont be labeled with Nvidia gimmick numbers like they did with intel. amd r3060/r3070/r3080 would be fucking stupid imho.
With hindsight seeing the slides with wrong core counts the number plate deduction was icing on the top.
Basic marketing means i9 needs a Ryzen 9 for the xtreme tier, more cores than typical consumer chips.
"I have no sources and make this all up" would've been a funny april fools video
Thank you for existing, Jim. These videos are just amazing.
Monday mornings are so blah. Except with Jim saves the week! Coffee poured, headphones on for 25 minutes amd fifty seconds of Adoredness.
Yeah buddy💪🏽💪🏽😎
I like how you had a freudian slip and said "amd"
That's a good analysis, I think you are right on the base clock numbers. Most people don't understand that super-computers and servers care far more about performance efficiency than they care about pure performance. It is very clear that at these lower frequencies Rome will be able to double performance at the same power consumption.
The consumer chips operate hotter... closer to the tail end of the sweet spot and even outside the sweet spot, meaning higher frequencies but lower power efficiencies. At these higher frequencies (presumably 4.5-5.1 GHz), performance CANNOT double at the same power consumption just by adding cores. For the same number of cores we should expect these higher frequencies, the better FP units, and a modest (10-15%) IPC gain. In total my guess is that we will get a 25% improvement over Zen+ at the same power consumption for the same number of cores due to these factors.
When we get into the many-cores Zen 2 consumer and threadripper chips (such as, say, a 16-core Zen 2 Ryzen) we will be able to achieve a doubling of performance at the same power consumption at the base clock, but I'm guessing only a +50-60% improvement in total performance at whatever turbo it can achieve with all cores running (due to power limitations).
-Matt
64 core 3.0 GHz..... Dank..... That's a fast boi
Anyone with 2 functioning brain cells wouldn't buy intel even if it were offered at cost. Which is what intel is doing to give away vaporware supercomputers in 2 yrs. They'll be slow and power hungry. LOL at any idiot stuck with that garbage. They'll have to get rid of it just to avoid the power costs of the shit performance.
yfw the 1700 only boosts to 3.2GHz on 8 cores
Rome is out of control
can't wait for the twitter "critics" to say that this video isn't credible because of its release date.
Moore´s LAW = Im absolute!!!
Murphy´s LAW = Hold my beer...
I heard you left Discord running while recording this episode :P 18:44
Yeah I was listening to this on my phone and it really confused me cause I didn't think discord made that kind of notification sound on my phone.
Wow thats probably the most thought through April fool i watched today!
Chapeau, Jim chapeau.
Wait what thats real!?
Thats even better! ;D
Watching your channel made me an expert on Scottish accent! And I love your tech stuffs as well.
Just one accent of many though. :D This week a western islander complained to me about how people in Edinburgh say their 'A's kinda funny. People west of Dumfries sound very different to people in Glasgow, or Aberdeen. There are all sorts of Scottish accents.
@@mapesdhs597 Yeah I know. In my opinion Scottish accents are the most difficult of all English accents. Listening to AdoredTV on a regular basis, now I can say that at least I understand this one. :P
@@tarekmahmudbhuiyan7368 Actually most Scottish accents are quite mild. Try living in Lancashire, Merseyside or Birmingham in England - way worse. :D
That discord pm sound at 18:45 confused the hell out of me when i saw i had no messages but heard the sound.. :D
same, jumped on my other screen :))
Very interesting my dear Watson. I love your EPIC stories.
Awesome catch on the licence plate and its intended meaning. You nailed it again!!!!
AdoredTV, always a great video, I never miss one. Over the years you have educated me and I've never regretted a PC investment since. I admire your humility as you don't debate with emotion, logic and fact are you're tools. Keep up the great work. On Payday I'll send what I can via Patreon, not much but I know sometimes every bit helps. Thanks for the insite.
Admit it Jim. You're psychic. Great information as usual.
And again, another spectacular video from Jim. Keep it up!
Love your videos, nice to hear a excited voice again. Keep it up great job!
I demand greater details! More please! Thanks Jim, another interesting watch😀
Oh man so many EPIC informations that i'm overloaded with hope, this acctually made my day much better. thanks Sherlock!
Your puns regarding epyc arent really getting better... ^^
Still love them! :D
What are you talking about!? His puns are soooo epic.
first comment that made me laugh, congrats Don Guru de Bro :D
there is no April fools day for the leak industry, great vid as usual Jim!
Great video as always :) Just got back from my honeymoon and the last video was right before it so thanks for all the wedding gifts!
Great Video. Looking forward to Epyc in the DC. Already pushing for it for my new ESX clusters for VDI.
I always stop everything i'm doing to watch your videos. Thanks for putting out a great video as always.
I don't know why but your videos just make me happy
Another impeccable analysis Jim!
Please AMD, gib 12 core R7 and 16 core R9. I have no problem buying a new board.
R9 would be a real middle finger to Intel.
Ryzen 3000 could possibly fit two chiplets with an I/O chiplet. So two 8-core chiplets with Ryzen 9 designation is possible and 12-core would be binned down from the 16 core.
Didn't watch the video yet, hope its not an April fool's joke
Forgot about that. lol
25 minutes April's fools joke? I don't think so
@@medimeb That would show quality!;)))
Doesn't April fools end at 12 O'clock?
pc hardware is no joke
Just wanna say that I'm honored to be a patreon as of today. Thanks for everything you share Jim!
Cheers bud!
18:45 :^)
how unprofessional
unsubbed
canceled patreon
disliked all your videos
contacted the reddit police
your days are numbered, Jim
Great stuff as usual man, Zen2 is looking so good. I can't fathom what they'll do after Zen3, the innovation is real.
We don't need any epyc April fool's jokes from you Jim - what AMD is pulling off here is no joke. It's very helpful you're explaining how you are putting this picture together, which raises at least my confidence in what you're saying. Tremendous work once again.
As always Jim Good work and thank you!
Absolutely impressive research and explanation!
Much more impressed with that molecular dynamics simulation demo than the C-ray demo used in the earlier show, and I say that as he who admins the original C-ray page. :D I hope the PCIe lane provision improvements filter down to mainstream desktops with Ryzen 3K. It's annoying the way atm there's only one Gen3/x4 link for M.2 to the CPU. Has any vendor done a storage focused board, with just one slot for a GPU at x8 but using the other 8 lanes to provide two more M.2 links to the CPU? I did look but couldn't find anything.
Isn't the problem that the M.2 interface doesn't have enough connectors to allow for more than 4 lanes? With PCIe rev. 4, we'll see something like doubled performance though, so it's not all that bad. More than 4 GB/s isn't really *that* important even for most prosumers, unless you're running a lot of VMs under heavy IO load all stored on the same SSD.
@@stale2665 Hehe, tell that to studios already working with 8K in Flame. :D It's not btw that I was suggesting there should be links with more than 4 lanes at Gen3 (though I believe there are Enterprise PCIe card SSDs that use 8); rather, that atm manstream chipsets only have enough CPU lanes to support one such link, though in theory a board could provide three (in addition to lesser chipset links) if it only bothered with 8 lanes for the GPU. Hmm, maybe vendors figure anyone meddling with that sort of thing would prefer TR anyway (the average ANSYS user would beg to differ).
But anyway, re I/O, well remember there was a time when a certain someone said something similar about having enough RAM. ;) There are already pro users struggling with big data; that the rest of us are little familiar with their plight doesn't detract from the problems they face. From 8K to defense imaging, 4GB/sec isn't enough. A guy I know at one major movie company who was tasked with sorting out a solution for 8K came up with a RAID using several HP Z drives to give just under 10GB/sec, which nicely matches their SAN as it happens. Still a bit messy though. Thing is, the Big Boys league of pro/industrial users were working with relevant quantities of data to require this level of tech 20 years ago, except back then it needed a zillion parallel FC links to make it possible and it cost a fortune, but it did work:
www.sgidepot.co.uk/onyx2/groupstation.pdf
Nowadays the expectation is to find solutions that do the job without needing the cash reserves of Zuckerburg's wallet.
GIS, heavy imaging, studio editing and other fields already need far more bw than one M.2 port can provide. These days the average satellite space mission can kick the living snot out of even a decent sounding M.2 RAID with how much data gets generated.
I just get bored too easily. I want to be able to backup my 1TB archive in less than a minute. :D Then I want to post mental AS-SSD numbers on a forum and point snooty fingers at all the poor slow people. I still can't buy a spaceship, like I was promised 30 years ago, so instead I'll settle for being able to copy God's dodgy movie collection in six seconds onto something smaller than my thumb. 8)
I've been watching too much Zero Punctuation. Sorry. :}
Love the videos, this always feeds my inner need to know more. Shame Zen desktop release is going to take a bit longer, but I would rather they got it right this time, remembering zen initial release fiasco.
Good video, Jim, as always. I finally moved to Zen+ system, so glad to hear that Zen2 to is pushed back a bit :D
18:45 I thought that was Discord on my computer at first XD
Excellent video as always!
Thank you again Jim, love these vids.
Awesome video, or should I say Epic video as always :) keep up the great journalism work, I don't mind that a large part of it is an educated guess, it's all fun.
keep up the work. Enjoy your videos.
Keep going Jim! Super as always! 10+
Very interesting analysis as always! :)
Lots of really big things here, if true. My kind of video! :)
AVX2 at base clock speed and faster than Intel is huge. That would propel TR/Epyc ahead a great deal for certain workloads. Also, the faster memory system sounds most interesting.
I was excited before, but now I'm really, really looking forwards to the Threadripper 3000 unveil.
Props for the link to the HPC event, too. Somehow I had missed that one.
can not wait for this chiplets, getting nearer and nearer replacing my good old sandy bridge system :)
Ryzen5 3600x, 500 series Am4 mobo and 3200mhz memory. My wallet is ready 😎
😎
seeing a new video upload from you and it is immediately a better day ;)
My Source at AMD from early 2016 has had initial design planned Chiplet GPUs for Navi, but was told that was pushed back in the lineup. I knew about GPU's having the design back then, as I was in with a Source back then on the Radeon team, but had no idea about the CPU's. It truly caught me by surprise that CPU's came first.
Damn, right on time for Lunch, thank you!
Have a nice day reader
Thanks mate
Wonderful video, thank's Jim! 🙂👍
I can appreciate a serious video when everyone and their dog is out to make the most lame, contrived jokes and trying to shovel them down your throat
I'm having flashbacks to the old Voodoo multi-chip days except now they're on the same package. Say hello to a future Radeon Voodoo 3000...
It's a good day when Jim uploads :)
The chipset motherboard delays are being caused by Intel bribing board manufacturers to purposefully cause delays of the AMD Ryzen CPUs. Calling it now, I don't trust shintel at all!
This is genuine infogasm
Very exciting. Can't wait for the R7/R9's personally.
Small correction, the 7601 is NOT the highest frequency Epyc, the 7371 is AFAIK (Base Clock
3.1GHz
, Max Boost Clock 3.8GHz, All Core Boost Speed 3.6GHz).
Edit: My apologies, that's comparing a 32 core vs a 16 core chip... I need coffee.
Another thing, there's NO JEDEC standard for ECC RAM faster than 2666, and I doubt we'll see it unless Intel gets on board since they hold 90+% of the market. I doubt AMD got the clout to do that by themselves, and no RAM manufacturer would make the sticks just for AMD. Look up how long it takes JEDEC to ratify faster chips...
Enjoyed the video. As usual you have more than earned my Patreon Contribution.
Great video as well!
I am now 4 weeks overdue on my new pc build. This is killing me too Jim. I would love a dual socket 64 Core EPYC render machine with 15 x RTX 2080 Ti's in it but I fear my wall socket would run out of breath and die.
Fascinating, thank you.
🙏 Thanks for the analysis. Can't wait for Ryzen 3XXX series.
Hi Jim great video, especially on the number plate, I know its guess work, but still it does seem plausible.
However I do think you are missing something since the CPU in that system with the Forza Demo could most likely be an 12 core Ryzen or even the 16 core. (Just saying)
Regarding everything else I'm quite excited.
God Bless.
When you start the video in the wrong way and rewind because you haven't heard the "a'rait guys, how's it goin' " ...cause it wouldn't be the same...
Yeah it's just not the same without the first two seconds.
I'm looking for a good upgrade to my i7 7700k is it worth waiting for AMD or just get the i9 9900k?
Been holding my new pc build off untill the new Ryzen is released. Seeing the memory price drop lately I was considering buying some in advance. My eye fell on a nice Corsair ddr4 3200 kit. Seeing that your source confirms the support of that I am tempted to order soon :-)
Great job Jim. I see that you are comparing notes with RegGamingTech. Paul seems to be hearing the same thing. What really caught my ear was when you said that AMD has over 100 OEMs lined up. I see that AMD is expected to gain 10% Server Market share. I am expecting about 15%. I bought 400 shares of AMD at $3.60 so I am stoked by your report. This will be EPYC...
you should make a decent bit return this year. the stock is doing really well already.
About that TB3, it was hazy... Do you mean that they are delaying the launch to introduce TB3 support to boards? Also does anyone know if Huawei and Lenovo will update Matebook D and E series respectively to Picasso APUs?
I check my subs feed and I see a new upload from your channel... I CLICK IMMEDIATELY!!! THIS IS THE ONLY CHANNEL I CONSISTENTLY DO THIS FOR!!! That says something, huh?
cheers again Jim
ooft I do love some Adored videos
Damn Jim! I have to go to work now! Oh wait! I work for myself, so I don't think I'll get fired if I'm late!
Great work as always Jim.
o7
We have been officially told that 7nm will bring either 2X efficiency at the same clocks or 1.35X the clocks at the same power.
From that alone we can basically calculate that Rome at the same clock speeds as Naples will run twice the number of cores at the same power. But of course that also assumes that Rome is the same configuration as Naples and we know it isn't.
Other efficiency gains in the architecture will be able to be translated to clockspeed, so if Rome gets a 10% efficiency gain in architecture, that'll translate to roughly 3-5+% (higher clocks aren't linear) more clocks for the same power. 5% of 2.2GHz is 110MHz so with a 10% architectural efficiency gain we could assume a 2.3GHz base for Rome over the 2.2Ghz of Naples and Boost would be 2.7Ghz vs 2.85GHz. With Rome being a radically different design due to the i/o die, there may well be 20-30% architectural efficiency gains because you aren't running 8 sets of I/O like you would if the design followed Naples more closely.
Going from 2.7 to 3.0GHz all core represents an 11% increase in clocks which is in the lower third of the clock speed gain multiplier for 7nm at the same power and could easily be accounted for in the 7nm mix of 2X efficiency or 1.35X clocks with the clock speeds translating purely from the architectural gains as the cores and I/O should be treated separately and the I/O on Rome will draw no more than Naples as the I/O is the same and not 2X as much (same number of memory controllers and the same number of PCIe lanes on the same 14nm process = same power or less with updated design) 10W saved on I/O will be roughly 10% more Core power available.
So most of this higher clocked Rome is old hat and easily determined from the next horizon event last year when we all saw the layout for the first time. It didn't need leaks to figure it out.
Not the end of the world but a late introduction may make it hard for some clients to wait for Zen 2 and Navi. Found it strange no mention of the latter's time table. Would be nice to know. Thank you for all your hard work. It is appreciated.
Nice work, keep it up!
I watched the whole video, amazing work, and dedication. I was really hoping for more info on Ryzen 3000, thats what im interested in not Rome.
160 PCI-E lanes on top of all the other known stuff, Intel REKT :D
That plate is such a cool "easter egg".
160 PCI-e lanes is just absurd. Could you imagine 10-way Crossfire? That would be fucking cool. It's kind of sad that it's limited to 4-way on a uarch level.
Any ideas about the design of the 7nm APUs?
yeah i dont think its gonna clock lower than current zen at least. i would think minimum 4.5 all core at least overclocked with no tdp. probably with the tdp limit in place
I was watching something else and I had to drop it immediately as soon as I saw the notification. How rude of you, Jim!
Love the discord notification at 18:47😂
So I am understanding ryzen 3000 after computex. So the believed launch dates of June or July given once they get the chipset they have to have time on the bios and production are still correct. What will make me mad is if they delay that to 4th quarter. or say nothing at computex and stick to rome.
Nice to see Jim's lighter side w/this April fools joke
So do you mean that the x470 motherboards wont be able to get the increased memory speeds?
Pizza all hot'n'ready and video is fired up in 1440p 60FPS, my body and brain is ready!
What type of pizaa, you fat cirious gamer
18:46 nice one
Jim uploading at a normal time of day? Something has to be wrong...
I think he did that as an April fools joke.
11 pm here...
Jim's been kidnapped by Intel and they are going to slowly start poo pooing Zen 2. Or at least thats what I think.
Great video thanks.My guess is that AMD will officially launch EPYC ROME on the 1st of MAY (50th anniversary) on a special event to be announced within 2 weeks for availability in June as you said and Ryzen 3000 at Computex for release in July or maybe August if late. We will see soon.Lisa always said that Rome will be available before Ryzen on 7Nm.
I need to know what it going to cost, the new Am4 motherboard?
What do you (or other people) consider to be the main reason for the chipset hangup? Specific technical issues OR taking their time for a much better day 1 launch showing (as opposed to the various issues 300 chipsets and Ryzen 1000 had 2 years ago)? I'm fairly sure they want as close to a flawless launch this time around, but to what extent bigger technical (hardware?) issues are a factor I don't know. Of course, the 2 are connected, but well, we'll see.
This is just the tech improving on the level we are stuck in. If you can't make it faster, make it cheaper or consume less power. All these little improvements are stacking up to making a very efficient cpu.
Just liked it but I will watch it this evening.
I know it will be good.
Thanks for the video :)
I hope AMD keeps on succeeding.