32 cores. intel fanboys on suicide watch https://www.amd.com

>>60461881
Just right after Intel killed its EPIC lineup. Well played.

>>60461881
Wait, EPYC? Holy jej ayymd is still keking over Itanic.

This is great and all, but fanboys don't care about server parts. From what I can tell, they care about two things. Clockspeed, and FPS. And Coffee Lake S will have less of both.

>>60461881
Fuck, I just got a ryzen 7

>>60461983
Aaand who the fuck cares? With that AMD literally secured WHOLE 1S server market for themselves.

Nyce

>>60461881
OMG what a shitty name
But boy do I want one
No wait 32

File: thread-800x449.jpg (41KB, 800x449px) Image search: [Google]

41KB, 800x449px

>>60461991
>>60462130

Threadripper is AMD's HEDT platform with 16 cores 32 threads. You should be looking at that if you really want something like this.

File: 1489103005792.png (162KB, 633x900px) Image search: [Google]

162KB, 633x900px

>>60461881
NOOOOOOO

>>60462278
I know
I was trying to be funny

Intel has a 32 core too, probably at shit clocks and 205w but its there.

Intel also has 32 cores with AVX512 for doing real work

AYYMD literally irrelevant as they can't run AVX512 code at all

You think cores are important here? Lol, the memory size, vandwidth and IO count are far more impressive.

Fucking underage

>>60462539
>shit clocks and 205w but its there
Literally DOA

File: image.jpg (213KB, 1140x647px) Image search: [Google]

213KB, 1140x647px

>>60462587
Lol, RAMlets, when will they ever learn

>>60462590
this

>>60462548
Genuine question, who uses vertexes on CPU?

>>60462638
this is the most hilarious PR slide
I can just picture the looks on those guys' faces when they were putting this together

>Muh Coars
is it bulldozer all over again?

File: Comb17052017044915.jpg (491KB, 1315x2227px) Image search: [Google]

491KB, 1315x2227px

>>60464192
shut the fuck up IMG_XXXXfag

File: 1491239101547.jpg (81KB, 406x406px) Image search: [Google]

81KB, 406x406px

>>60462856
>looks on those guys' faces when they were putting this together
Something like this

Wonder if this thing will put any downward pressure on existing HEDT and Ryzen 7 CPU pricing.

>>60464252
No you retard.
These things are at best 10x more expensive

>>60464252
Threadripper will put pressure on inter HEDT but not on mainstream market

>>60464252
No, it's a datacenter SKU with absolutely cosmic margins.

Begun the core war has

>>60464397
>begun
Intel's already lost it.

>>60464402
Nah, they're close now.
But when 7nm hits and Intel barely manages 36 cores while AMD is already at 48 with more IPC then it's over.

>31 cores
>probably able to make use of only one core just like the rest of AMD's processors
Absolutely EPYC

>>60464462
>Nah, they're close now.
Their highest 2S offering is 28c housefire. And Intel's 1S offerings are just pathetic. They lost. That's it.

>>60464491
RIPtel

File: bengalcats_1.jpg (52KB, 600x600px) Image search: [Google]

52KB, 600x600px

32 cores my nigga?????

>>60464537
And 112(128) PCI-E lanes. And octachannel memory.

>>60461881
But can it play Far Cry 1 800x600 at 900FPS? Yeah. Thought so.

>>60461991
im still on i5 2500k. maybe i should wait for 64 core cpus.

>>60464296
>at best 10x more expensive than (intel) HEDT
I don't think any businesses will buy $10000 AMD 16c/32t Zen cores don't command that kind of price.

>>60464352
>>60464372
They are marketing Threadripper themselves as their own HEDT platform, I think consumers (and businesses) will read into that as meaning HEDT-like pricing relative to intel's HEDT platform. It'll be more than $1000 for sure but more than $2000? Unlikely.

>>60464876
>They are marketing Threadripper themselves as their own HEDT platform, I think consumers (and businesses) will read into that as meaning HEDT-like pricing relative to intel's HEDT platform. It'll be more than $1000 for sure but more than $2000? Unlikely.
Anon, the thread's about EPYC 32cores rapemachines.

SINGLE. SOCKET. HOLOCAUST.

> https://www.amd.com/system/files/2017-05/TIRIAS-AMD-Single-Socket-Server.pdf

> ~30% of 2S servers are only populated by one CPU, they just need the extra I/O or other features
> Only gotten because of Jewtels pricing and configuration tiers
> AMD's EPYC offers more features, increased I/O, more performance, and more memory per 1S server than Intel, at a cheaper TCOO

Oh dear. This doesn't look good for Intel. EPYC's 2S systems similarly offer more of everything at a presumably similar/cheaper cost than Intel.

>>60465844
2S is a meme with EPYC. 1S is where it is.

>>60461881
EBYN line :DDDDDDDDD
[spoiler] I should stop with this stupid joke [/spoiler]

>>60465885
You should stop using spoiler tags on boards that don't support it first.

And what would any regular user need more than 8 cores for that makes it worth it to pay extra and sacrifice single core clock speed? Come on, I'm listening.

>>60465983
What the fuck are you talking about nigger?

>>60465983
What would any regular normalfag care about server hardware?

>>60465983
to make modern websites with 10000 javascript frameworks work without lag.

I can't wait to get my hands on a dual-socket consumer motherboard for the first time since the Intel Core2 Extreme Hovercraft Edition

How fast do you think encoding 4K H.264 videos to HEVC or VP9 will be on a 16-core Epyc, let alone two of the motherfuckers?

File: who the fuck makes unanimated.gif (301KB, 1024x576px) Image search: [Google]

301KB, 1024x576px

HELLA
F*CKIN
EPYC

File: laughing bitch.gif (165KB, 500x454px) Image search: [Google]

165KB, 500x454px

http://www.bitsandchips.it/52-english-news/8372-rumor-crazy-high-yields-for-ryzen-dies-over-80

File: 1492855909643.png (944KB, 1228x1502px) Image search: [Google]

944KB, 1228x1502px

>>60461881
I'M LITERALLY SHAKING RIGHT NOW

I HATE AMD

>>60462539
28 cores is not 32 anon

File: naplespackage.png (1MB, 1024x775px) Image search: [Google]

1MB, 1024x775px

>>60467533
>80% yields for fully functional dies
Jesus christ, that is an absolutely ridiculous yield for a new design.

Also that's an interesting way of laying out the chips, I thought they'd be vertical relative to the substrate in a row ( | | | | )

>>60462587
>4TB of RAM
>Infinity fabric running at 100GB/s
>128 PCI lanes
what were you saying?

File: the_reg.jpg (117KB, 1920x1080px) Image search: [Google]

117KB, 1920x1080px

32? 32 fucking cores?

...FreeNas w/docker anyone?

>>60467533
and /g/ told me global foundries was shit wtf

>>60467723
Dude it's full of IBM foundty stuff now lmao.

>>60467533
I mean, the process is pretty mature, but hell it's nice for a new design.

>>60467708
Yes, 32 fucking cores and 128 PCI-E lanes.

>>60461881
>https://www.amd.com/en/products/epyc
Isn't it 16 cores and 32 threads? Intel is up to 24 cores and 48 threads. Regardless, Intel has the more advanced fab plants so can put more transistors onto a die the same size. This has always been the issue with AMD.

>>60462278
After taco bell i can release threadrippers too.

>>60467887
32 -physical- cores. 128 threads.

>>60467887
16/32 is Threadripper aka HEDT Ryzen. EPYC is their server offering. And it's 32/64 with fuckton of memory and I/O.

>>60467887
8 cores per die, 4 dies per package (for Epyc/Naples), 2 packages per machine. 64 physical cores, 128 threads per machine with an unholy fuckton of memory and I/O on top of that.

>>60467911
>32 -physical- cores
>128 threads
kys yourself

>>60467897
Thats hot if you're a girl.

>>60467985
(｡>﹏<｡)

>1S system with 128 PCI-E lanes
ALL THE GPUS.

>>60467998
(｡>﹏!｡)

>>60468013
M2 ssds as well.

>>60467977
Not saying this guy is right, but my ultrasparc t2 machine has 4 cores and 32 threads, so it's not non-existent in high-end server offerings(esp. risc stuff).

>>60468068
Talking about EPYC here

>>60467241
60 FPS? 120 FPS?

Are intel/amd wars the stupidest thing on /g/ at the moment?

>>60468208
Yes. Who cares?

File: 1486510190271.gif (155KB, 640x360px) Image search: [Google]

155KB, 640x360px

>>60468208
Yes, the future of x86 in the server room is "stupid".

Back to /v/ now.

>>60468239
x86 rules the server room today. CISC has proven its worth over RISC.

File: moarcoars.jpg (60KB, 678x240px) Image search: [Google]

60KB, 678x240px

>makemoarcores.jpg

They're doing it again.

>>60468327
And it works. Servers love more cores. And more memory. And more I/O. And non-gimped 1S systems.

File: sandra01.png (127KB, 602x423px) Image search: [Google]

127KB, 602x423px

>>60468349
>Servers love more cores
Those cores need to be actually working, and not twiddling their thumbs thanks to Infinity Fabric latency.

>>60468386
>different CR
Wew.

>>60468386
This is the new 'hurr durr Ryzen sux' image that's going to get posted twice an hour every hour for the next century, isn't it?

https://www.pcper.com/reviews/Processors/Ryzen-Memory-Latencys-Impact-Weak-1080p-Gaming

>we can narrow it to say gaming is more sensitive than most of the non-gaming workloads that are utilized by reviewers and analysts to inspect, measure and gauge the performance of a processor and platform.

tl;dr it's nothing unless you're a gaymur fag. And since we're not talking about gaymes here ..

>>60468386
There is only a few applications that NEEDS ultra-low latency and those will stick with XPoint Memory.
Your point is completely useless.

>>60468386
Could you please show me the workload where IF actually chokes?

>>60468453
>he thinks Xpoint is faster than DRAM
>he thinks people will "stick with" a technology that hasn't shipped yet

>>60462590
its boosts to 4.1 ghz like the 1800x it is literally 2 1800x's MCMed onto the same substrate

>>60468386
memory latency doesn't matter when *INSUFFICIENT MEMORY TO LOAD*

>>60468589
Of course he can't. He can only repost images from the "AMD is doomed" article of the week. Without actually linking the article.

>>60468722
8 channels? Why would you need 6 channels for?

>>60468727
Like, I'm genuinely interested in ways to make IF choke under somewhat realistic conditions. Also faggot forgot that IF in EPYC is clocked much higher.

>>60468205
25FPS, 30FPS, and 60FPS

File: 1494809801547.png (206KB, 600x446px) Image search: [Google]

206KB, 600x446px

>>60461881
Yeah megatasking!

Wait humans can'treally multitask... Hmmm and wait, most software can't be highly threaded. Even games can't really go beyond single digits.

But sure, whatever. I mean great for datacenters I guess.

File: Suharto.jpg (49KB, 630x960px) Image search: [Google]

49KB, 630x960px

>>60469196
What the fuck do you think Epyc is for you mongoloid /tv/ pedophile?

>Wait humans can't really multitask
Hang yourself faggot.

>>60468386
just install a database server like postgresql there. it will use 100% cpu and ram

>>60468589
You'd have to create something with a main thread that spawns slave threads that communicate with each other, this is fairly normal.
But if you want to make it choke IF you'd have to tell it to constantly copy the main and slave thread to different dies for no reason whatsoever.
In fact this would impact Intel's performance as well since you're doing core to core comm which induces quite a lot of latency, it would just choke AMD harder.
Needless to say nobody makes such retarded code

>>60462278

I wanted to build two baby servers for my closet. One low power nas for on all the time. Maybe freenas again. Then another for linux which I can turn on and off over the lan and do linux shit with it.

>>60470319
It's pretty much server hardware, so yeah.

>>60467533
>About 80% of Ryzen dies have all 8 cores fully functional.

can we get unlocked r5?

so if they double cores every year then we will reach 1000 cores in only six years
thank you based amd without you we would still be stuck with 4

>>60471008
That's not how it works.

>>60471008
2560 cores on my gpu

File: 1475564961272.png (246KB, 882x758px) Image search: [Google]

246KB, 882x758px

>>60471036
not with intel but with amd we are making progress

>>60471036
Yeah. We'd reach 1024 cores, not 1000.

>>60461881
if it's going to cost as many shekels as intel's top line, it's the same as nothing

i'm glad it exists, but i wont buy it.

>>60471055
Your 2000+ shitcores can't even do VFMADDPD

>>60461881
intel fags will denie amd superiority no matter what

>>60461881

OP I hate to break it to you but you but fizzbuzz can't use 32 threads

>>60468733
Everyone knows 4 channel is all you needs, besides this shitpost was made with dual channel.

>>60471124
Intel is charging $7k for 22c/44t 2.20GHz with quad channel memory and 40 PCIe lanes. No way AMD charges anywhere near $7k for their top end, and it's more capable. And this isn't even including E7.

>>60471761
Maybe yours can't.

I'm psyched, having to work on massive projects the compile times are fucking abysmal, even running a 4.0GHz octacore processor. I'm readddyyyyy

>>60467692
Jeebus! AMD might actually pay off their debt of this is true.

File: 41815A_ATHFXchip_FR_BK_low[1].jpg (103KB, 480x268px) Image search: [Google]

103KB, 480x268px

>>60461881
I want an overclockable dual-socket motherboard for these, QUAD FX style

>>60461881
EBYN :D:D:D:D

>>60472687
This is third time I see this joke, and I still laugh from it.
:DDDDDDDD

>>60468733
>>60471785
single channel is good enough goyim

Why did amd ditch their own name for their servers sku's, I perfer the name Opteron.

>>60475604
me too
epic sounds like rice gaymer edition

File: 1494410921788.png (119KB, 320x252px) Image search: [Google]

119KB, 320x252px

>>60461881
NOOOOOOO

>>60472608
quad fx was shit

Why do you faggots have these threads EVERY SINGLE FUCKING DAY?

Both are soulless companies trying their hardest to sell you new shit with the least possible amount of new in it.

I'm telling you now, the human eye will never be able to see more FPS than a 2500k can produce.

I just want a big shuriken 2 adapter for ryzen pls

>>60478093
\(^o^)/

>>60467533
>80%
It's possible I guess.
The Zen Arch in relative terms, is really quite simple. Combine that with the very mature 14nm node it is built on.

>>60471761
>not making forkbuzz instead where it forks itself for maximum cpu rape

>>60478093
It's this or C vs Rust

Shit's a welcome break from the daily language autism threads of recent weeks

>>60461881
>moar cores!

File: 1445873750669.jpg (15KB, 250x250px) Image search: [Google]

15KB, 250x250px

>>60471874
>$7k for 22c/44t 2.20GHz

File: 1299967672057.jpg (118KB, 307x485px) Image search: [Google]

118KB, 307x485px

Can't wait for my 16c/32t under 500 bucks.

>>60478616
>moar bingbus!

>>60462548
Would be great for the few HPC apps that can actually use AVX512 (and for those you'd definitely want the new high end Xeons), but is pointless for general purpose servers and is a bitch to compile due to ridiculous errata - they've had since Knights Landing to try to get it right and it's still flaky and clocks down, barriers etc.

Naples engineering samples run very nicely indeed for general purpose 1S servers (don't know personally about 2S) with plenty of fast IO and Epyc will likely be extremely price competitive with Xeon - if it is they may well have actually a few big contracts, not just taunting Intel to get cheaper prices this time.

Apparently over 80% yield too.

Daily reminder that Intel are developing 32c CPU too
http://wccftech.com/intel-skylake-e5-2699-v5-32-core-geekbench-score-leaked/

>>60484696
>165W
Maybe at 2 GHz and without so much as touching AVX. That thing generates more heat than a nuclear reactor.

>>60484809
EPYC will be rated at 220W+

>>60484932
Socket specification says up to 200W

>>60484972
>Socket specifications
PCI-e is rated at 75W but rx480 drew more than 150 watts

>>60484988
It did draw more than 150W, but not from the PCI-e slot.

File: aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9BL04vNTkxMzU5L29yaWdpbmFsLzE4LUdhbWluZy1CYXJzLnBuZw==.png (22KB, 600x450px) Image search: [Google]

22KB, 600x450px

>>60485038
Yes it did

>>60485081
>82 watts from the PCI slot

File: aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9BL0svNTkxMzU2L29yaWdpbmFsLzE1LUdhbWluZy0zRC1QRUctT3ZlcndpZXcucG5n.png (105KB, 600x450px) Image search: [Google]

105KB, 600x450px

>>60485113
On average

I knew AMD was going to rape the Intel server market but I never expected to see Intel faggots arguing about how good a 7700K server would be

Jesus

>>60485168
The butthurt is real.

>>60485141
That's the reference card, which later got patched.
The custom vendor cards never had this problem AFAIK.

>>60485168
>Ctrl+F 7700
>1 of 1 match
Try harder

>>60485141
Yes and the PCI-e is rated for 75W... on average. It's not a max.

>>60468386
>>60485204

>>60485210
It's clearly stated 75w limit

>>60485180
no kidding

Gluing CPUs together to sidestep difficult manufacturing is stupid effective

>>60485210
>>60485204
>>60485113
>>60485038
>>60484972
Full damage control from AMD fags
https://community.amd.com/thread/202410

>>60485251
Going over 75W on average is a breach of the protocol and there are lots of motherboards that will die under those kinds of loads since they're not designed for more than the specs.

Saying that 'hurr durr 155W is being drawn from the 75W rated PCI-E slot and is causing explosions' is dumb fanboy speak.
Leave the hyperbole in the toilet.


You hear this? I'm being reasonable and you are not.
Also kill yourself

>>60485219
let me rev up my OCR software

>>60485251
>everyone I don't like is an AMD shill
what's going to be next? "it's not true if I ignore it"?

AMD's 32 core EPYC doesn't really matter for anyone here.

AMD has moved from being a budget bang-for-the-buck brand to targeting the high-end high-margin market segment.

If you're buying a AMD Ryzen then you're giving 30% of what you pay directly to AMD as gross profit.

>>60462539
Intel's 32 core CPUs don't really matter, just like AMD's won't. You're not going to pay $5000 for a CPU unless you're into something very profitable that absolutely requires it.

>>60485282
>You hear this? I'm being reasonable and you are not.
>Also kill yourself

>>60485306
He's right though. You should remove yourself from the gene pool.

>>60485168
It scales down very well since power is frequency * voltage^2.

E3-1505L 4C/8T 2.2ghz 25W with an igpu
4*8 = 32C
25*8 = 200W

>>60485292
>AMD's 32 core EPYC doesn't really matter for anyone here.
Yeah, no one here is a sysadmin, hardware guy, solution architect, etc. No. We are all NEETs here. The memes are 100% accurate.
Also server market is a joke. Everyone knows gayming is the biggest market 4ever and ever.

>>60485306
That's a reasonable request tho

When will Apple Macs have this?

File: 1494252206252.jpg (111KB, 900x675px) Image search: [Google]

111KB, 900x675px

>>60485405
>Muh datacenters

>>60485292
>If you're buying a AMD Ryzen then you're giving 30% of what you pay directly to AMD as gross profit.
Intel's margins are way higher.
>>60485292
>You're not going to pay $5000 for a CPU
Hold up there, dipstick. Where exactly did you hear it's going to cost $5000? Sounds like Source: My Ass to me.

>>60485804
>MUH dwarf fortress

Meanwhile in the real actual world:

>there is no programming language on Earth today that makes multiprocessing easy
>threads and processes remain extremely heavyweight abstraction as far as kernel resources consumption go
>problems solved by regular applications remain serial in nature
>only highly specialized applications will even begin to benefit from threads support let alone more physical cores
>in fact software architecture as a whole is moving towards single-threaded event loops for I/O including user input
>video games are literally just one giant single-threaded event loop
>memory and network I/O remain as the dominant bottleneck in modern computer architectures

File: 6b6.png (21KB, 625x790px) Image search: [Google]

21KB, 625x790px

>>60486287
Oh look, another entirely evidence free shitpost.

>>60486325
You mean the same servers whose HTTP servers run a single threaded event loop that accepts zillions of connections while keeping memory usage at a minimum because they don't incur the overhead of threads and their stacks? Data centers whose hardest task is storing and moving humongous amounts of data all over the place in the most cost-effective manner possible?

You don't need clock speed for either of those...

>>60486332
Oh look another fanboy

>>60486287
>>60486457
Can't tell if trolling or just retarded.

>>60486477
>i must justify my AMD purchase
>0 arguments
>u r tard

>hardcore gay-memers are going to waste money on this shit even though they don't have any workload to saturate even 8 cores

lmao

Why do Xeons egist when they're bad at gaming?

>>60486528
Are you shitposting or rally this stupid?

File: retard.png (160KB, 568x1023px) Image search: [Google]

160KB, 568x1023px

>>60486503

>>60486546
He's making fun of the "muh single coarse!!!!1!" kike shekelchaser I think.

File: apprentice and second firing door in foreground.jpg (331KB, 1067x1600px) Image search: [Google]

331KB, 1067x1600px

>>60486287
>>60486457
>MUH SHINGLE COURSE!!!
>In a thread about server CPUs
MFW kiketel shekelchasers have been reduced to this.

File: intel thermal designed final solution.jpg (28KB, 550x550px) Image search: [Google]

28KB, 550x550px

AMD stole the future from us. We should've been on 10 GHz a decade ago. Damn you AMD, you have to ruin everything with your COARS! https://en.wikipedia.org/wiki/Tejas_and_Jayhawk

>>60486528
They're good for gaming, for professional gaming, xeons are error free and their for gamers who mustn't allow a single error to happen to their games

>>60486608
(You)

>>60486563
Except I'm not pretending dumbass. What I said is literally the truth, 32 fucking cores and you have no recourse but to say 'but intel is doing it to' and 'what about X', 'stop trolling'. The roles are reversed here fucker

>>60486598
So what if it's a "server CPU"? I/O is still vastly more important for performance. Functionality is compartmentalized. Programs are fully virtualized now in order to isolate failures and decrease attack surface. Do you actually run all your programs on a single god damn server? I pity you if your company does this.

What the fuck makes you think you actually need this 32 core piece of shit? If you have such highly parallel workloads that it could benefit you, I bet your needs are likely better served by GPGPU.

stay mad you unemployed fucker

>>60486735
> I/O is still vastly more important for performance
Which the 32C chips will have an unholy fuckton of. 128 lanes of PCI-E, 8 memory channels, integrated sata controllers, and who knows what else AMD has crammed onto each of the 8c dies going into each one.

I/O is something AMD is not lacking in this generation for their server chips.

>>60486735
>I/O
Oh, you mean that stuff that EPYC has a shitton more of than any Intel offering?

Actually, there pretty much isn't a single metric on which EPYC doesn't BTFO every Xeon

>32 core
Oh don't worry there will also be a 16-core EPYC to BTFO all the cheaper Xeons too.

>>60486780
>>60486865

>128 lanes of PCI-E, 8 memory channels, integrated sata controllers, and who knows what else AMD has crammed onto each of the 8c dies going into each one.

All of that is great. Those features have everything to do with I/O performance and nothing to do with multiprocessing.

If it were me I'd get rid of these idiotic cores and cram even more sweet stuff in the chip. I don't know, even adding more cache would substantially improve things and would be a better use of the chip real estate.

Leave multiprocessing to the GPU. Parallel problems are already leveraging it to great effect. A measly 32 core CPU is laughable really.

>I/O is something AMD is not lacking in this generation for their server chips.

It deserves praise for that. Not for it's huge amount of stupid ass processor cores.

Computer architecture needs to move the fuck on. Nobody cares about processors anymore. They're good enough. Focus on improving the ridiculously slow memory and network. Processor caches can only help so much.

>>60486923
If it were you, these companies would be bankrupt.

IBM's coming out 176 thread POWER9 chip soon, Intel's coming out with their 28-32 core chip soon, AMD has its 32 core.

They don't build these chips for kicks, their customers demand more compute power and more efficiency.
Ignoring the memory capacity and I/O, because those are pretty important nowadays too.

>>60486987
>IBM's coming out 176 thread POWER9 chip soon, Intel's coming out with their 28-32 core chip soon, AMD has its 32 core.

Yawn.

>their customers demand more compute power and more efficiency.

Sure.

Lowering power consumption is a great way to improve efficiency. Today's world is all about low power computing. "More computing power", especially higher clocks and more cores, are the enemy of low power. This isn't just about mobile battery-powered devices either. Server farms consume absurd amounts of electricity just to run all the hardware and they're also spending money in refrigeration costs to dissipate all the heat the hardware generates.

Using more expensive materials and manufacturing processes to ensure longer lifetimes is a great way to add value. Consumers deal with planned obsolescence but companies use things for a pretty long time. "Server processors", just like "server hard drives", are generally built to run 24/7/365 whereas consumers frequently turn off their computers. Reliability is key here.

>Ignoring the memory capacity and I/O, because those are pretty important nowadays too.

"Most important" is more like it.

Industrial chips have more of everything that matters, including processor cache, memory addressing lines, better expansion options, ECC memory, the list goes on.

Quite frankly if you decide which processor to purchase based on the amount of cores it has you deserve to be fired. Buy a fucking GPU if you really need that stuff. It can just as easily be virtualized in order to power several servers worth of applications using the same technology that allows Linux gamers to play on a virtualized Windows guest.

>make -j33

my body is ready

>Ryzen processors seem to be reaching strong values and are reaching numbers passing 80% for fully the functional 8-core parts.

intel dreams about it.

>>60487272
Intel's yields are high too for small dies but their fabs go into the price, and fabs are fucking expensive.

So it was true that Ryzen doesn't have all the I/O lanes enabled, while each die has 32 on Naples.

>>60487288
intel can't make 8 cores with 80% yields, though.

>>60487139
I don't know what you got against cores, but you do know there will be 16 and 24 core versions with the same amount of I/O, memory bandwidth and channels as the 32 core ones who don't need the compute performance and want less power and something cheaper?

>>60487351
>I don't know what you got against cores

They cost money, take up chip real estate that could've been used for better things, use actual electricity to run and dissipate heat even while idle.

>16 and 24 core versions with the same amount of I/O, memory bandwidth and channels

Sounds great. No need to spend money on features that are useless at best and outclassed by graphics hardware at worst.

You keep saying "compute performance" as if you were going to crunch numbers with the thing

>>60487402
It's a general term for higher throughput, but it is a fact that more cores per node give you the ability to run more VMs or docker instances.

>>60487432
>more cores per node give you the ability to run more VMs or docker instances.
Nah most datacenters now are constrained by space since they overbuilt the power delivery and cooling for projected future needs but those needs never materialized. Now it's all about stacking units butt to nut and having a short run to the storage units. Every major center is going to be getting almost free power from supply contracts with power providers since they're in on the renewable energy bonds scheme.

File: 1366912581266.png (356KB, 724x3267px) Image search: [Google]

356KB, 724x3267px

>>60486287
>there is no programming language on Earth today that makes multiprocessing easy
literally the dumbest thing I've read today.
congratz

>>60486287
>>there is no programming language on Earth today that makes multiprocessing easy
Erlang is literally made for extremely fucking durable, extremely distributed systems with a downtime less than a second in a year.

>>60488122
Almost forgot the educational video that comes with every Erlang distribution: https://www.youtube.com/watch?v=uKfKtXYLG78