Zen to have up to 32 cores?

It's not out yet. Besides,
>can have
Its like saying I
>can have consentual sex with a woman
Doesn't nean it's going to happen.

>>52916464
One of AMD's locations is in Israel.

>>52916531
As is Intel, and many others. These are multi-national conglomerates that have all manner of stations in every corner of the world, from marketing to R&D to manufacturing and distribution. They are chinks, spics, niggers, jews, gondola, nazis, pajeets, dindus and everything in between. What's your point?

x86 is dead
long live arm

>>52916464
I won't lie. This looks pretty attractive.

If they don't go stupid with the price, it should be more than competitive with intel options.

>>52916650
This was a thing with every generation. It looked good, but they fucked it up every time.

>>52916464
>we will have to buy DDR4 and a new mobo

Only thing I dislike about it. If it was DDR3 compatible, I would think about getting a Zen CPU.

>>52916500
Not with that spelling you're not.

>>52916464
Looks like the "MORE CORES" meme was real all along!

32 cores is for servers you mongoloidsxl, none of the desktop variants will have 32 fucking cores

Meh?

Broadwell-EP Xeon's with 20 cores + HT for 40 threads already exist.

>>52916464
why not HBM support?

>>52917033
Zen has SMT too.
Probably 2 threads per physical core .

>>52917001
All Zen CPUs will use the same socket.

>tfw your home computer has no purpose for anything more than 8 cores
>tfw with 32 cores, each core will likely be less efficient/slower than an intel core
>tfw my almost 2 year old quad core intel cpu will be faster than this

>>52917078
Its not meant for the consumer market but hey, u want to spend 1000 to play video games and watch anime, go ahead.

File: 1453261987691.jpg (178KB, 558x695px) Image search: [Google]

178KB, 558x695px

>>52916464
>What is your excuse for ever siding with jews?
pic related

>>52916464
you fucking idiots always fall for the marketing newspeak
"up to" is "guaranteed not to exceed" you fucking imbeciles; if they put out a 2-core CPU, it qualifies
what I want to see is "starting from 8 physical cores", that would mean something

>>52917033
Zen is reputed to be competitive with Broadwell, clock for clock, and 32 cores would be a considerable improvement over 20 for server workloads.

Note that indicates that Zen cores might be smaller than Skylake/Kaby Lake cores, which is contrary to what you might expect from the jump to SMT. Perhaps Keller worked his shit-wreckjng magic again.

Note that I think they're talking about the Opterons here. As far as I've seen, they're planning 4 and 8 core desktop parts (8 and 16 threads). Which is still very tempting if they're reasonable and competitively priced (something Intel haven't been doing recently, knowing how in the lead they've been).

For the anon up above wanting DDR3, not happening: DDR3 is obsolete and production is now winding down.

It's a pity it's landing in late 2016: I'm doing a new Skylake build this evening.

>40% improvement over past cores
yeah, AMD is finished

>>52916464
>32 cores
...wow how useful... can't get that IPC up to save your lives though, huh? I'll keep my 4790K for 5+ more years and enjoy technology that actually works well.

>>52916464
>builb more cores :DDDD
>ebin :D :D

Eight cpu core APU when?

If Keller actually managed to make the CPU market competitive on a few months of contract work, can we consider him to be among the greatest modern engineers?

File: 700.jpg (23KB, 600x606px) Image search: [Google]

23KB, 600x606px

>>52916464
>./configure
>make -j33

my body is ready

>>52917001
Broadwell-EP (also 14nm) is only expected to have 22 cores max.

If Zen cores are nice and small then the 8 core desktop Zens could be dirt cheap.

>>52917405
This holy shit.

Fucking bullshit marketing

By the time it comes out it will be 2017.

Intel will release XX core CPUs before.

No enterprise with right mind is still using AMD CPUs and won't change with Zen.

AMD is not worth anything on the market.

I wish AMD would come back with some serious competitive hardware, but I don't see that happening.

Would be good for the market though.

>>52917135
>tfw you realize you're a dumb gaymen
>>>/v/

>>52917216
Are you suggesting that $1000 wouldn't be a bargain for a 32 core CPU? Intel charge that much for an 8 core and will be charging $1500 for the 10 core Broadlel-E chip.

>>52917634
>this much retardation
Intel Xeon release plans have already leaked and they will not breach 30 cores before 2018:
24 cores in 2016 (Broadwell-EX)
26 cores in H1 2017, i.e. the Zen competitor (Skylake-EP)
28 cores in later 2017 (Skylake-EX)
30-32 cores in 2018 (Cannonlake-EP)

>>52917731
>muuuh cores

I can bet my money on it that the 24 cores will easily dominate the 32 Zen cores in 99% of programs used.

>>52917838
>Thinking servers only run one program

>>52917838
>I only play vidya

>>52917857
>>52917985
>implying all cores scale equally good and server programs will make use of every single core.

If it were this easy, Intel would have made decacore CPUs by now.

Or looking in the mobile phone departement, Mediatek SoCs would dominate the benchmarks with their 10+ cores.

>>52916847
As a normie faggot that fucked up my mobo on my first build in 2012, this is okay for me. But otherwise no.

>>52918021
>servers arent running multiple VMs per cpu

Since it's AMD, I think you mean to say 32 "cores"

http://www.theregister.co.uk/2015/11/06/amd_sued_cores/

>AMD sued: Number of Bulldozer cores in its chips is a lie, allegedly

>>52916464
i dont trust amd anymore

>>52918021
Most proper software scales with core count. Look at GCC. Meme programs like mobile apps do not.

>>52916464
Phenom II quad core here.

Waiting for Zen to upgrade. 16 core minimum for me, intelfaggoteers.

I'll upgrade to anything that can cut my 8 hour x264 encoding time down.

>>52916464
I really do hope Zen is competitive with Intel. Because right now Intel has no competition and zero incentive to innovate and progress. Maybe it'll force them to charge normal prices for their processors, too.

If Zen fails AMD will be in trouble. And we the consumers will be ass raped by Intel.

>>52918305
HAHAHAHAHAHAHAHAHAHAHA

>>52918305
that was their last good processors they made

>>52918305
>>52918388
I upgraded from a Phenom II X4 955BE to a 6700K about four months ago. My penis is still erect.

>8 channel ram

holy guacomole

>>52918491
Isn't that just 4 slots dual-channel? I'm not that educated on ram

>>52917373
Yes. Tesla picked his ass up knowing he could wreck shit.
He's the Engineering equivalent to Ron Jeremy. He comes in and he wrecks ass for hours with his fat nasty Italian dick.

>>52918570
kek

>Guys, we havent had a successful chip design in years. What do we do?
>Uhhh...try just throwing more and more shit into the chip.

>>52918491
>motherboards with 16 ram slots
My dick is diamond

File: b59428839a5e2e15687885792d5cc2c328e9067a.jpg (114KB, 1302x1323px) Image search: [Google]

114KB, 1302x1323px

>>52917405
This 32 cores thing that's been floating around better be the consumer grade figure...

>>52916464
Because I use a laptop. Jews are way more energy efficent.

>>52918359
Itnel has competition from ARM, so they are trying to make their chips more energy efficent.

>>52916681
Remind me where they fucked up with Athlon? Or better yet remind me when anyone has fucked up with what Keller has worked on.

>>52917405
>make -j33
>not make -j$((2*32))
u wot

>>52917302
>I'll keep my 4790K for 5+ more years and enjoy technology that actually works well.
Because Intel literally haven't made anything better than Sandy Bridge.

>>52918554
no. an 8 channel board would have 8 identical sticks installed

>>52918140
>AMD gets sued over allegations
>Nvidia still hasn't been sued over 3.5GiB

>>52918491
>dat bandwidth
It's pretty amazing from a technical point, but what would it be useful for?

>>52916464
>32 cores
MOAR CORES

Still, it would be amazing if they're competitive. I'm patiently waiting for the Zen 8 core, but having that kind of behemoth on a server would be so cash.

>>52919640
keeping pace with 32 cores

>>52919556
There is a class action lawsuit against Nvidia There is no way they can get out of it because they lied on their website and to journalists about the specs of their product. They can't just pretend it was an error and it took them 3months to notice it.

>>52919595
>>52919667

File: amd more cores.png (224KB, 778x960px) Image search: [Google]

224KB, 778x960px

Still accurate

>>52916464
A high performance server APU will have up to 32 cores.
Zen is the name of the core architecture, it is not the name of a chip.

>>52916847
>actually being dumb enough to think AM3+ or FM2+ would receive new CPUs with how absurdly limited they are


>>52917044
They do, for some parts.
OP's pic isn't from AMD, or anyone of note for that matter. Just some nerd rehashing already available information into a slideshow presentation.

>>52917078
All upcoming Zen based consumer chips use socket AM4, the enterprise SKUs do not.

>>52917289
>Zen is reputed to be competitive with Broadwell, clock for clock,
No its not. This is reddit garbage from tech illiterate retards.
Lisa Su came out and said that their new line of Opterons would address 80% of the X86 market, meaning that they'd be competitive in 80% of common workloads for the market segment. In other words its not competitive in 20% of them which are undoubtedly anything FPU heavy.

Steamroller and Excavator are both massively far behind intel's latest arch in performance per clock. Sandy Bridge is nearly twice as fast as Excavator in some metrics.

>>52916847
By the time Zen is released, DDR4 will have 90% of new PC sales marketshare so you would have to buy new mobo anyway. I say its a perfect opportunity to change sockets. Besides, the socket will be unified for all AMD desktop products so this is going to be neat. This may boost APU sales because of upgradability to CPU+GPU later on and we may finally get to build beefier AMD based ITX rigs.

>"hurr durr moar cores" memes still alive after intel released their 8 cores CPU
intel fanboys are truly retarded

>>52919684
So basically in heavily parallelized tasks? Databases, rendering, what else? I mean uses other than "it's a server".

>>52919784
I really doubt that. If I were a betting man I'd wager my left nut that we'll have mobos for "ITX APU" builds and they'll have utter shit for power delivery and it'll blow up the very instant you will try running anything with any horsepower to it socketed in it. I don't think AMD will gain enough mainstream traction for companies to put out such niche products as powerful ITX boards.

>>52917262
>what is concurrency

>>52919842
Most graphics APIs only use the 1st core and maybe run some simulation logic on the others if you're lucky.

>>52916464
>What is your excuse for ever siding with jews?

i heard intel is about to come out with a 64 core CPU with 50 threads per core and only consumes 4.2 watts DESU SENPAI

>>52919977
6 million cores

>>52919946
vulkan/dx12 are supposed(?) to fix that.
still
>playing games
>caring about games

>>52920002
>December 18th
>Nah
>Vulkan Soon™

I was really stoked for it too.
Fucking hell.

>>52916464
>32 cores, 64 threads
>supports 128 GB of RAM
Anyone know what you could do at home to justify purchasing that kind of hardware? Shitposting about it doesn't count.

File: 1452132196465.gif (3MB, 211x222px) Image search: [Google]

3MB, 211x222px

>>52916464

They are advancing much in the category of amounts of cores in processors. Meanwhile, the retard programmers and developers fail to plan properly for multicore and can not even do develop programming languages excellent use that amount of cores.

>>52920454
At home? r u retarded anon?
You buy this kind of hardware to use as servers for anything heavy load, not to play badly optimized AAA games on it

>>52916464
Fuck man, transcoding video with 8 cores is already pretty reasonable, 32 cores should be fucking awesome.
I am so converting all movies and anime typo h265 with Zen

>>52920501
My fucking nigga.

>>52920491
>AAA games are the only thing home computers get used for

>>52920501
This
I can't wait to see the applications of 32 cores in the hands of your normal consumer. I bet all the anime groups will finally switch to H265
Plus now every joe will be able to try his hand with neutral networks, because of the insane amount of multithreading

it can only be better than their current 4 cores + HT.

>>52917135
Keep dreaming /v/ immigrant
32 cores vs 4 cores for 3d rendering, or video converting or heavy calculations, guess who'll win? 32 cores

>>52920628
>every joe
This shit will be stupidly expensive.

What's the point with having 32 cores when almost every program there is so poorly optimized that you're using one thread to do all the computing.

The whole problem with the current gen of AMD processors is that they get poor performance in real life because applications don't take advantage of all the available cores.

>>52921001
Those programs don't matter, if someone is buying a 32 core CPU, they must have their own suitable programs to use with all those cores.
There are plenty of industries that rely on hugely multithreaded CPUs to function

I want /v/ to leave
>winderz
>vidya
>muh ipc

>>52919699
>5%

Pretty optimistic senpai.

Modern processors could easily have over 32 cores, but since it is so difficult to implement shit like parallel processing there is no need.

>>52920548
Besides rendering, what is the heavy task at home ?

>>52921380
Compiling and encoding in my case

>>52921380
Your ms excel needs a lot of cores too
Chrome is heavily multithreaded, hence for web browsing too.
Streaming media to other devices.
Converting media
With a suitable scheduler, it could make the general performance great too

>>52920501
>>52920628

You meant Daala right? Or maybe VP9 until it is out?

>>52921458
How does Excel fares about multicore/heterogeneous computing? I have been away from MSOffice for quite a while and calc is just in the OpenCL infancy, but is improving.

>>52916464
>up to
Probably for new Opterons

THIRTY VIRTUAL MACHINES

THIRTY

>>52919796
multitasking as well. Not being stuck waiting for I/O while another core makes a call for resources

>>52921762
Opterons and ServerAPUs are the most likely scenarios.

>>52919595
Massive ramdisks

A few of these and a couple of those new S1750 GPUs would be a nice compliment to just about any business.

File: sad cat.jpg (64KB, 600x443px) Image search: [Google]

64KB, 600x443px

>amd

https://www.youtube.com/watch?v=8lTGk0R12gs

>>52921001
I make use of the 16 threads on my dual socket xeon desktop all the time
32 would be even better

'32' meme cores

lel

>tfw one day CPU will be so strong we wont need videocard anymore

>>52921965
Seems like the opposite will happen first.

>>52921965
The opposite is probably more likely to happen.

It's nice and everything, but since much of software is compiled with Intel compilers, microcode optimizations will not kick in, meaning that every Intel compiler compiled program will run slower on AMD processors, rather than on Intel ones. Most of gaymes and commercial software are built with Intel build tools. It's painstaking to know that you have 32 cores, but many programs will just ignore them and disable all SSE features, seeing that you have an AMD core :(

>>52917262
11 cores.
lolwut

>>52918554
It seems they are using 4 channel per die but it will scale to two dies per socket. The 32 processor version will be a dual 16 processor one in one socket. Think of what Intel did with early Pentium D and Core Quad designs.

Single die versions should be quad channel and only up to 16 cores.

File: 1446910224368.jpg (2MB, 3010x2000px) Image search: [Google]

2MB, 3010x2000px

>>52922045
It's probably one of those Phenom X12's that just didn't cut it and repackaged it as a Phenom X11.

I still use a Phenom II from 2011. Is there any reason to upgrade this year if you don't game or recompress massive amounts of video?

>>52923633
If you do not need all the computational capacity it offers, there are today alternatives that are much more energy efficient. Otherwise no.

>>52923746
Anything I can run silent without a massive HSF?

>>52923844
Only if you want dual cores.

>>52923858
Nope. I guess I'll wait.

Chinks are making deca-core 2ghz cortex A72 SoCs for $10 before Zen will even come out.

>>52916464
I personally have been so disappointed by the current gen of AMD products that I will not buy from them again until there's actual word and evidence out there that shows it's not shit.

>>52923912
A single Haswell/Zen core is bigger than 6 A72 cores.

>>52916464
Any more news on the APUs?

File: 1447802494328.jpg (27KB, 391x506px) Image search: [Google]

27KB, 391x506px

>64 threads

Has AMD ever done SMT before?

>>52924056
No, K6 and all up to K10 are CMT

File: p8.png (566KB, 1024x768px) Image search: [Google]

566KB, 1024x768px

>>52916464
Reminder from 2014.

>>52918021

Wow, you're extremely fucking retarded. You don't know what you're talking about at all.

Servers run virtualised environments you fucking retard, and those who don't have server software which can quite easily scale over multiple CPU cores with worker threads.

Server software in general scales really well over many cores, as do compilers on compilation servers and servers for transcoding/encoding media also have real use for a lot of CPU cores.

Just please shut the fuck up and kill yourself. Your meme software on phones and vidya isn't a good example for the enterprise side of things

>>52919595

APUs, ramdisks

>>52918021
>Intel would have made decacore CPUs by now

They are up to 18 already. Besides, server motherboards usually have more sockets.

http://ark.intel.com/products/family/78585/Intel-Xeon-Processor-E7-v3-Family#@Server

Intel already has 120% more ST performance than AMD, 40% changes nothing really.

>>52924362
It's a good step in the right direction you fucking dipshit

File: 0nIkCAb.jpg (116KB, 636x812px) Image search: [Google]

116KB, 636x812px

>>52918388
You do know an fx6300 will BTFO the phenom ii x4, overclock higher, and has updated instruction sets?

Your phenom ii x4 is more like an fx4300.

>>52924439
They're still irrelevant unless they manage to undercut Intel's pricing by a very significant amount.

>>52924484
They've always been relevant.

>>52924362
>120%
In your dreams.

It's 50% at best.
And that's in extreme synthetics where the entire frontend and backend is running at the same time.

>>52924362
>>52924525

If they can indeed reach 40% that would be meaningful, see:
http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/9

>Haswell to Skylake (DDR3): Average ~5.7% Up.

>>52917262

that's my 5th core :(

>>52924128
Holy shit, IBM's finally coming out with a new POWER chip?
Fucking finally!

>>52925895
More than that, see: http://openpowerfoundation.org/

Huge heterogeneous effort. By the way, keep an eye on: https://raptorengineeringinc.com/TALOS/prerelease.php

>>52918491
Already running quad channel, and pedal to the metal, it really doesn't do anything special.
But it may be useful for an APU, which I lack.

>>52926031
YES
IBM is literally my fetish.
If there were a good dual socket 2U server or a BladeCenter that used these, I'd buy it as soon as I got the money.

I'd really love a mobo with two of these that's EATX.

>>52916464
Quality > Quantity

>>52926206
Is that why the fx 8350 can beat a third and 4th gen i5?

>>52926199
IBM itself seems to be invested into Rack mounts instead of blades. Maybe another OpenPOWER partner could offer one, probably Cirrascale.

http://www-03.ibm.com/systems/power/hardware/s812l-s822l/browse.html
http://www-03.ibm.com/systems/power/hardware/s824l/browse.html

http://openpowerfoundation.org/press-releases/cirrascale-joins-openpower-foundation-announces-gpu-accelerated-power8-based-multi-device-development-platform/

Who cares? I'm rooting for Zen, but 32 cores isn't a reason to buy one even if the engineers can pull it off (I'll believe it when I see it). You can get a 12 core Xeon, but they cost $1600+. A Zen chip with over 12 cores will be priced for a server, not a workstation.

>>52926594
HOLY SHIT
I thought IBM sold their server and BladeCenter line to Lelnovo in 2014?

I'd definitely get one of these, but at those prices, wouldn't I just be better off getting a mainframe?

only 40% makes it shit compared to i7

also: believing anything amd says. ever.

>>52927113
yes, that's why they sold it.

>>52927113
Depends on your workload, if it cant properly benefit from the Power architecture then there is no point.

They sold their x86 lines. Also they are now focusing more on licensing Power through OpenPOWER than doing the whole thing themselves.

>>52927292
Android power phone with 64 threads when?

>>52927168
Well, I guess that can make sense.

>>52927292
> Depends on your workload, if it cant properly benefit from the Power architecture then there is no point.
MUH OPEN SOURCE
That's actually one of the most exciting parts, actually.

> They sold their x86 lines. Also they are now focusing more on licensing Power through OpenPOWER than doing the whole thing themselves.
Oh, okay.
I still wish they didn't sell their fucking chip business to GlobalFoundries.

File: amd computing.jpg (63KB, 800x208px) Image search: [Google]

63KB, 800x208px

>>52916464
AMD adding more cores?
Colour me shocked.

>>52927360
...

Power is to x86 like x86 is to ARM. I cant imagine they fitting the Power architecture power envelope inside a portable form factor. Even if they did, it would probably not be an efficient solution.

>>52916464
>ADD MOAR CORES
Meme magic in action. Jesus christ.

>>52927465
Yeah, no.
Both the PS3 and the Xbox 360 use POWER chips.

They're fucking awesome. People made supercomputers or clusters with PS3s, even the US military. They got extremely mad when Sony stopped offering OtherOS on their PS3s.

>>52922045
Number 12 had the camera.

>32 cores

Just keep throwing what little money you have left on that fire AMD your graphics cards are hungry for more fuel.

>>52927522
Consoles have historically used custom power designs of some sort, notably so Nintendo. PS3 is actually a Cell microprocessor instead. None really compares to a smartphone or tablet. And that is probably the reason why Apple dropped them.

>>52922045
>cpu10
That's 10 cores you fuck

>>52927647
0/10

File: image.jpg (27KB, 650x600px) Image search: [Google]

27KB, 650x600px

>>52918412
I also have a 955be.

>mfw my 6600k and z170 both arrived today

Too bad I've been too busy puking and shitting magma to install. This made it bearable, though.

Now I need pascal to hurry the hell up.

>>52927601
Ah shit, I completely forgot PS3s used Cell. I thought they used POWER.
Anyways, yeah, Apple dropped them because POWER was too hot, and they wanted more GHz (I think). However, those problems have largely been solved, seeing how there are 2U servers that are being released with them.

>>52927725
Cell is a cousin of the mainline POWER architecture.

Also:
https://raptorengineeringinc.com/TALOS/prerelease.php
https://raptorengineeringinc.com/TALOS/prerelease_specs.php

>POWER8 in workstation format
>up to 12 cores (physical not logical)
>$3100 starting

Time to save up my neetbux.

>>52916847
I bet you anything that the lower end AM4 boards will support DDR3 or DDR3L. We have some H110 and Z170 motherboards that support 1.5V DDR3 even though Intel says it's not recommended.

>>52927864
Does it run Linux? If that runs Fedora and has at least 16 pcie and at least one pci I'm buying one.

>>52918359
>Don't forget about Nvidia.

File: 1455047106772.webm (2MB, 846x472px) Image search: [Google]

2MB, 846x472px

>>52919595
>8 channels x 2 RAM sticks per channel x 128GB per stick
You do the math
Once Samsung starts mass producing 128GB ECC registered RAM sticks, I will make the ultimate RAM disk, cached into NVME PCIe drives.

>>52927292
What sort of workloads does POWER excel at?

>>52916531
where in Israel? Haifa?

>>52927864
But why do that when you can have a P900 for about a third of the price?

>>52918570
Kek

>>52916464
So, since server Zen is 32-core, I think it's reasonable to say AMD's high end offerings are going to have 10 cores, and there will be that one chip with 2x the wattage that will have 16 cores.

Calling it now.

>>52928239
Ramat Gan which is near Tel Aviv

File: amdahls law.png (63KB, 528x430px) Image search: [Google]

63KB, 528x430px

>>52919842
Any task that inherently parallel is better off on the gpu. As for the rest...

AMD still engages in the same business practices as Intel. They both have signed microcode updates in their processors and there is the Platform Security Processor.
https://libreboot.org/faq/#amd

>>52928586
If AMD is actually as bad as Intel by making unwanted features in their hardware, is there any reason to use AMD over Intel? Should people just go for the most powerful hardware?

>>52928586
I remember reading something late last year going over the problems with Intel's ME (pretty sure it was called Intel x86 Considered Harmful or something like that) but there wasn't as much about AMD's PSP and the stuff I could find seemed to indicate that it wasn't nearly as bad as Intel's ME for what could be compromised with it. Is there more information out there showing otherwise?

>>52928540
>pulling random laws outta your ass

>>52928874
I can't even tell if people are pretending to be retarded anymore.

https://en.wikipedia.org/wiki/Amdahl%27s_law

>>52928540
The fact you can buy 8 core mobile phones and 4x18 core servers right now proves you wrong.

>>52928994
You don't even know what the argument is, yet you choose to participate in it.

>>52927563
if fire is what you need, nvidia are happy to provide

>>52927919
Pre-release version runs Debian, they still need to sort a number of things and optimize the software stack.

http://www.phoronix.com/scan.php?page=article&item=talos-workstation&num=1

>>52928307
Are you talking about the Lenovo Thinkstation P900? My understanding is that is Xeon only.

>>52928540
Except having more CPU threads available allows you to have higher tenancy regarding zones/jails/vms.

>>52927864
>POWER8 workstation
Muh dick

>>52929308
Yeah it is. It's also a dual socket workstation.

For some reason I'm not able to configure it on their website, but I think I got it's cost down to about 1200.

File: RAS.jpg (69KB, 528x486px) Image search: [Google]

69KB, 528x486px

>>52928217
The most meaningful advantage of Power is the memory sub-system and I/O. If you need to deal with lots of data (like large databases) or big data structures (like large matrices) then there is no comparison. Data mining and ERP probably are the markets IBM has been marketing Power more aggressively to.

For Virtualization scenarios you would benefit from POWER if you used their firmware scheme, KVM has support for it:
http://www.ibm.com/developerworks/linux/tutorials/l-ibm-powerkvm-system-bring-up/

They scale better to larger parallel workloads if you can parallelize your code correctly.

Power also offers lots of RAS capabilities like pic related.

>>52929447
PS.: Items 1 and 3 benefit each other greatly. And 2 may be related.

>>52927919
From the second link provided:

2 x16 Coherent Accelerator Processor Interface (CAPI) capable PCIe slots (8 shared lanes)

4 x8 PCIe slots

>>52927919
>>52929696

Oh, and:

1 legacy PCI slot

About CAPI:

http://www-304.ibm.com/webapp/set2/sas/f/capi/home.html

>>52916464
>M O A R
>O
>A
>R

>C O A R S
>O
>A
>R
>S

>>52929784
moar coars moar whores

>>52916847
That's for the high end opterons. The normal chips might have DD3. We don't know yet, and the damn thing is a year away minimum.

>>52917033
>Broadwell-EP Xeon's with 20 cores + HT for 40 threads already exist.

This is 32 physical cores.
Zen can do hyperthreading, so a 32 core Zen would have 64 logical cores.

>>52922034
As I recall there was already a tool out there that patches the cupid lookup of a runtime app. It's a beta though.

And they could always just change that line on the cpu to whatever Intel has it set to (which is what intel compilers check). If Intel sues them, they'll get fucked because it is proven that they create huge unfair advantage to their competitor.

>>52925895
Yeah, but it's something like a 220W beast that actually sucks shit in single threading, the only thing it is good at is performance per square meter.

>>52929711
I'm fucking sold.
Now where to get one in Australia?

>>52917731
>tfw got the 28 core chip but if I just waited 2 more months I coulda got the 30 core

:^(

>>52916464
That's 32 cores plus multi-threading anon

>>52916847
DDR4 costs like $10 more than DDR3, the prices have pretty much already equalized

File: AMD Vs Intel Updated for HSA.jpg (238KB, 1632x295px) Image search: [Google]

238KB, 1632x295px

can't wait to update the image sometime

>>52930130
DDR3L and DDR4 cost exactly the same here in aus.
At least it does from Achieva.

>>52930200
AUS is 10 kinds of fucked up on it's pricing and avlibility, haswell isn't even available in New Zealand, at least from the PC part picker

>>52930226
Pcpartpicker isn't as reliable here. I've found better deals by actually checking the websites.
Also no point getting an i5 6500 when the i5 4460 is 70 dollars cheaper and just as fast. Almost.

File: bg91x7H[1].png (112KB, 776x484px) Image search: [Google]

112KB, 776x484px

lol still relevant

>>52930588
>Implying that progress will come to a full stop at some point, so that picture won't ever be relevant again
There will be a time when there are 1024 cores, but there's no telling who will make the fastest processors then. If it's still not AMD then I guess they won't be around anymore.

finally 32 cores will go mainstream

thanks based AMD as usual