AMD Threadripper coming in fast!!! Theres going to be a 12 core

File: Vega.png (41KB, 1163x329px) Image search: [Google]

41KB, 1163x329px

4 Flavors of Vega on its way!!!!

Me, exclusively for shitposting and speccy threads.

File: inamd.png (131KB, 1161x512px) Image search: [Google]

131KB, 1161x512px

Intel will be using AMD graphics!!!
Is the end of Intel's graphic division??

INTEL BTFO

I don't need it, but I want it

I'm just hoping this'll lower the prices for 1800X so I can snatch one of those

Kikeripper has no mercy.

>>60823782
J I M K E L L E R

>>60823810
>Is the end of Intel's graphic division??
That was Larrabee.

>>60823822
1800X prices have already dropped, only by $25 but its a price drop never the less. So yes, the 1800x will drop even more.

>>60823810
Didn't Lisa Su specifically come out and just flat out say that there is nothing going on between Intel graphics and AMD?

>'We're Not Looking At Enabling A Competitor To Compete With Our Products'
-quote from Lisa on May 24

http://www.marketwatch.com/story/intel-and-amd-license-rumors-should-finally-be-dead-2017-05-22

Sounds like they just wanna go straight for the throat of intel.

>>60823901
If it's an Apple SKU, then there's no competition there.

>>60823901
I bet my ass it's some custom SKU for Apple which uses EMIB.

>>60823782
That second AMD Mandolin is very likely to go into 15W ultrabooks

>>60823901
AMD has to play a balancing act with intel, they want to sell to them but not make them lose investors. If i was an investor and learned intel will now use AMD graphics I would get worried.

>>60823782
12 core 16 thread? What the fuck are you talking about?

>>60823782
See that second letter in the strings being either M or D? That's desktop and mobile

>>60823782

>6 core cannonlake at only 2.6ghz

Mobile?

>>60823782
>6/12 3100 4200 KBL S DDR4 UDIMM EV CRB)
Those clocks are even worse than the last engineering sample. Just goes to show how unreliable this shit can be.

>>60823980
I don't think Intel is putting 6 cores into laptops.

>>60823782
Serious question: what do you kids need these MUH CORAZ for? Gayming? Benchmarks?
Holy shit, how old are you?

>>60823961
Fuck sorry, I got the 12 and 16 cores mixed up with threads.

*24 threads

>>60823982
I'll tell you some shocking news.

There are usually several engineering samples of the same chip.

>>60824010
>4 cores should be enough for anyone!
>2 cores should be enough for anyone!
etc etc. fuck off already.

>>60823782
I'm interested in one for the not-castrated PCIe lanes, but I'm only willing to pay ~$600 at most. If there's any Threadripper at that price or below I'll seriously consider it, even if it's just 8C or 10C.

>>60824034
With the rumor of 16 core at 850$, I think 600$ should probably get you the 12 core. Looking at those samples from OP it looks to be 3GHz minimum as well.

>>60824020
It still leads to blind speculation, you grade A cunt.

>>60823962
Thanks I did not know that.

>>60824010
>need
I don't need them. I want them for the same reason I want a car more horsepower than I could reasonable use without breaking the law, or the same reason why I stockpile ammunition that will likely never be used.

>>60824084
How old are you?
Is this what your priority is? Getting best PC config you can get? Do you live on your own?

>>60824105
>How old are you?
27
>Is this what your priority is?
Priority? No, but it's something I enjoy. I'm not going to stop simply because you don't like it and throw tantrums.
>Do you live on your own?
Of course, but there's nothing wrong with living with your parents either.

>>60824054
I hope so, it's price and how far it can OC that will decide if I get one. I've already got a custom water loop so cooling it should be no problem.

12c24t? Will the tsp on the low end be lower than the others?

>>60824288
This is quite possibly the dumbest argument for not upgrading I've ever heard.

>>60823782

OFF THE CHARTS EDITION

>>60823794
do I read IDs right?
so, 1.2Ghz vega performs in between 1080-1080ti
the heck is 1600Mhz vega then?

>>60823980
That's coffee lake
Cannon lake will be 2/4 and 4/8, and only for laptops.

>>60824545
Workstation card.

Non - frontier edition will be more optimized for muh gaymes.

AMD has already said this anon.

>>60824545
>the heck is 1600Mhz vega then?
FE card.

>>60824545
>>60824566
>>60824576

looks fine on the 16gb 1ghz version, higher oc headroom and confidence in baseline perf instead of oc'ing it to insane temps

File: jewshown.png (138KB, 354x504px) Image search: [Google]

138KB, 354x504px

>>60823782
>16c to be priced at $849

>>60824566
it will be same GPU with different drivers

600mhz difference doesn't disappear

1000mhz ID was floating around a few weeks back with 1070 scores, 600mhz bump puts it at who knows where

File: 1491102738152.png (13KB, 420x420px) Image search: [Google]

13KB, 420x420px

>>60823782
>AMD Jewripper coming in fast!!!
fixed

>>60824592
No, 1600mhz is 225w one. That's baseline Vega card.

File: Vega 4.jpg (109KB, 1280x665px) Image search: [Google]

109KB, 1280x665px

>>60824545

You cant compare AMD and Nvidia clocks.

>>60824566
>>60824576
this.

>>60823794
inb4 they are all bought up by crypto miners due to the current crypto bubble and you can't find a VEGA GPU anywhere (just like the RX 570 and 580 is currently sold-out worldwide).

>>60823810
>>60823901
She did and there won't be any Intel chips paired with AMD graphics.

>>60823993
Intel could very well be planning 6-core laptop chips because fully expect AMD to have a lower-clocked 8 core laptop chip. Ryzen 7 1700 has a 65W TDP. It's doable. Keep in mind that the 1600X has 2 cores less and that's a 95W part because of it's higher clock. Something like a 2 or perhaps 2.4 GHz 8-core laptop chip may be doable.

>>60824566
How much fucking wait for those? 2017 right?

>>60824614
i'm running 390
don't really believe that 225w

>>60824633
>tiled binning and larger L2
Pls read about Maxwell.

>>60824619
I'm comparing scores, 600mhz difference is very substantial for GCN

File: 1477771103515.jpg (336KB, 784x847px) Image search: [Google]

336KB, 784x847px

>>60824601

curious with premium mobo pricing, anticipating less than x99s

non-gaymen, just for reference

3m2 slots, 10gb aerion card included

802.11ad

>>60824633
There's about 3 architecture generations between your Hawaii and Vega, anon.

Also that 225W is in AMD's official slides.

>>60824631
Frontier will be released first. Consumer Vega shortly after. Read the news, my dude.

>>60824651
That mobo is going to cost upwards of $400, so it's a bad example

>>60824660
No shit? Intel's HEDT mobos go up to $700, no doubt the same will be for AMD's HEDT
Thankfully you can use the AMD mobo for 4 years, get CPU updates, but I'd still get the cheaper less overkill models.

>>60824651
thicc socket

>>60823948
I think that might actually be a 7W chip. The clocks are far lower than I would expect for a 4/8 config at 15W.

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

944KB, 1228x1502px

>>60823782
I'M LITERALLY SHAKING RIGHT NOW

PLEASE COMFORT ME INTELBROS

>>60824651
this socket doesn't get less scary every time I look at it

>>60824651

>the guy she told you not to worry about

>>60824683
That socket is the least of your worries, pretty sure in 2 years we will have a 32 core processor with a socket twice as big as that one.

Cant wait to cryptomine with these mega frontier cards, I'll buy 4 :D

>>60824697
coolers in couple years it's going to be x2 shrink, size should stay the same

>>60824697
The x399 socket and the server socket are quite literally the same. The chips will fit in each of the others socket. The difference being in that the EPYC chips are designed to run without a chipset.

>>60824680
Wait for 10nm, lol.

>>60824700
isn't it a bit on expensive side perf/watt wise?

>>60824713
then expect a 64 core processor with a socket 4x the size of a regular one.

there's nowhere to go but up on core count.

>>60824722
>Wait for a slower process
kek
Even Intel have said that their 10nm node will be slower for a significant amount of time, with only improvements being in power draw.

>>60824722
intel been disclosing things left and right recently, even tape ins
10nm canon only just taped in, yield news is yet to follow heh

>>60824718
EPCY are SoC?

>>60824729
in consumer space - impossible
consider the cooling, threadripper is already pushing it on cooler surface area size

>>60824697
Zen 32 cores for $1700? MMMMMMMMMMMMMMMMMMMMMMMMMMMM

>>60824754
Yes.

File: amd-one-socket-epyq-test-motherboards.jpg (56KB, 399x462px) Image search: [Google]

56KB, 399x462px

>>60824847
wow thats fucking crazy.
one chip delivers more performance than 2 xenons. I understand xenons are old and have not change much but it looks like intel is going to get fucked for the next 2 years or so.

>>60824893
No shit dude. 128 lanes on 1S boards is as anally devastating as it gets for Intel. They simply can't compete.

>>60824765
definitely dont agree but whatever, enjoy being a retard.

>AMD hating tripcucks nowhere to be found
This past year has been fucking awesome.

>>60824986
There's two still lurking around here shitting up the board sometimes.

>>60825008
its funny cause there really are 2 of them.

>16 cores of mediocre performance
klol

>>60825055
>>>/vr/

File: 1496588274604.jpg (470KB, 1060x1569px) Image search: [Google]

470KB, 1060x1569px

>>60825055
Don't you mean 18 cores of medicore performance .. from Intel?

File: realyy.jpg (38KB, 255x247px) Image search: [Google]

38KB, 255x247px

>>60825074
>that scaling
Hail bingbus.

>>60824010
>what do you kids need these MUH CORAZ for?

For me, 4C/8T is the sweet spot.

Going forward, I will be focusing primarily on single-thread performance, because that fits my workload the best. (99.9% of the time I have at least 4 threads idle.)

If a 4C/8T CPU offers better single-thread performance than a higher thread CPU (even by a little bit), I'll stay with 4C/8T.

I like 4C/8T better than any lower thread count because it gives me good flexibility in assigning thread counts when I'm running several virtual machines at the same time.

Basically, 4C/8T is a little bit more than I need, which is exactly where I want to be.

File: JUST krzanich.png (1MB, 1002x1480px) Image search: [Google]

1MB, 1002x1480px

>>60823782
oy vey anudda shoa! delid!

>>60825091
>Ryzed is bretty good
>Add moar bingbus

>>60823810
That would be a dumb as heck long term move because amd graphics ip will really enable them to kill intel on the mobile front

Why are there somany ppl using !!!!!!!????

File: 1495644450608.gif (505KB, 1080x1080px) Image search: [Google]

505KB, 1080x1080px

>>60825368

File: drake-strip-club-money-photos-06-480w.jpg (70KB, 480x373px) Image search: [Google]

70KB, 480x373px

>>60825368
I'm American.

I still have an 1100T and 750ti tb h fa m lol idc

jk ic

that socket she thicc she bad, can't wait to play infinite runescape sessions

>>60825646
Pretty much every HEDT motherboard comes with extra power connectors.

File: --------.jpg (225KB, 1512x879px) Image search: [Google]

225KB, 1512x879px

>>60824651

succ my power based AMD

i n f i n i t e s c a l i n g
n
f
i
n
i
t .
e .
p
o
o .

>>60825672

mm? I haven't seen any with more than a molex or something

>>60825751
Check the recent X299/399 motherboards, most of them either have 1x8pin or 2x8pin extra power connectors

>>60823993
>tfw most laptops still dual core

>>60825868
Ryzen Mobile to the rescue.
>>60825855
Asscock X399 have 4+2.

>>60825868
Which is why I don't think Intel is in any hurry to stuff 6 cores into a laptop. Would drain battery in two seconds. Ryzen Mobile on the other hand ..

>>60823794
how much would the frontier editions probably cost? TI-tier? Titan-tier?

>>60826277
Quadro Tier

>>60826277
$999.

>>60825677
what the fuck is dimm.2 anyway? literally just shove two nvme ssds into it and voila?

>>60827005
It's a slot for riser card with two m.2 slots.

>>60827017
oh. neat. thanks

File: 1496959284134.jpg (329KB, 960x686px) Image search: [Google]

329KB, 960x686px

>>60823782
>16C/32T base 3400 turbo 3700
and that's probably the highest binning. holy shit. Intel won't beat that with it's i9s (pic semi related)
I'm pretty hyped about this, but even more for Vega.

That said, how's Vega going to compete against 1080Ti? For the moment I'm stuck with a 2010 HD6870 and I'm few clicks away from ordering 1070. Wait for vega? yes/no?

>>60827115
>Wait for vega? yes/no?
Yes in any case. NVidia isn't Intel, they do drop prices.
> how's Vega going to compete against 1080Ti
Tbh nobody except AMD knows. Some say around 1080. Some that faster than 1080ti.

>>60827115
>Wait for vega? yes/no?
Yes

>>60827115
That 16G card has me salivating. I'm willing to wait and see.

>>60827176
>Some that faster than 1080ti.
Would it be cheaper tho? If so by how much? Will they go full cutthroat and go HALF?

>>60827176
and what I'm hoping for is that AMDs cards will be relatively cheaper compared to nvidias, otherwise I'm not going red.

>>60827190
>tfw that's 2x more than your system ram

>>60827196
Anon, I don't work for AMD. So I don't know, nobody knows.
>Will they go full cutthroat and go HALF
IF, NaVi will be as Epyc win as Ryzen -- maybe. inb4 Nvidia still having majority of market

Wtf ulv intel i5 mobile is now quad core?

>>60823794
just because there are four engineering samples doesn't mean there will be four different products. there could be more, or less.

>>60827235
I'm doing preliminary research on a new 1600x build (from scratch: monitor, peripherals, UPS, etc) and it's at $2600 currently with a 1080 TI placeholder. I think the final price would drop another $100 or so when we get Vega pricing. If we get another ebin release, hopefully the total would drop as far as $2200-2300

>>60827115
i'd wait for volta and vega

>>60827311
>on a new 1600x build
Have you considered going with 1700?

>>60827336
nah, for me it's either vega or nothing.
I've been waiting for few years like that going "maybe this will prove better at lower price" but it was exactly other way with everything getting more expensive.

>>60823782
Stupid question
A coworker of mine mentioned Intel has been using updated Sandy bridge cores in all their new CPUs
And that was why we keep seeing 5-7% IPC increases.
Does that theory have any foundation?

>>60827482
well okay then wait for vega

all i'm saying is pascal isn't necessarily a big enough jump over maxwell and polaris isn't meant to compete in the high-end segment

you do you, brandwise

>>60827546
And Volta is even smaller jump than Pascal. It's either Vega is good and all is okay (praise Pajeet) or it's bad (doubt) and you'll have to buy novideo.

>>60827570
it's still a jump though and will hopefully be priced alright

honestly i'm just excited to see what new shit all the companies are releasing. except for intel. babbylake-x is just a mess

it's also nice to see AMD returning to Athlon 64-esque days. they deserve it, after what intel did

>>60827518
He's completely right, but the core Intel is using goes back further than Sandybridge

>>60827518
Yes, and even then Sandy Bridge was an evolution of an older Pentium architecture (the not-Bentium-Hyber-Bibeline one). Most of the gains in IPC and efficiency was due to process shrinking, offloading tasks onto built-in hardware accelerators, and increasing core clock frequency by a few hundred MHz.
You could say that Kaby Lake is a remake of Skylake, which was a remake of Broadwell, which was a dieshrink of Haswell, which was a spin-off of Ivy Bridge, which was a dieshrink of Sandy Bridge, which was a remake of Westmere, which was a dieshrink of Nehalem, which was a reconfigured Yonah, which shared lineage to the Pentium Pro of the mid 90s.

>>60827680
>>60827675

So will they be forced to release a whole new architecture?

>>60827675
In 4 years.

>>60827705
Yup, see >>60827709 (misquote)

>>60827680
hold on, why only mention one generation of core2?

>>60827680
The difference is that before Sandy Bridge, they actually updated the architecture with new shit and more efficient solutions (Sandy doesn't have the same pipeline as Pentium Pro).

But AFTER Sandy Bridge, they kept the CPU part almost entirely the same. They added some new AVX instructions and bumped the clock speed up and that's about all the change they made with the computation part of it. Beyond that, they only change crap like the iGPU, the memory controllers, and moving more and more stuff on-die from the motherboard (I recall Haswell having the VRMs there or something like that), all in an effort to make it a better mobile SoC design. For two reasons: one is that they don't need performance anymore, they had no competition on that front, and 5% clock boosts combined with new AVX instructions can be marketed as a significant increase. Second is that the only potential competition they had were the mobile segment, so they had to make the chips smaller and more complete and easier to implement as a full system.

>>60827646
>priced alright
Its bigger dies and Huang will keep his margins.

>>60827772
Because I forgot. But there was only one generation I skipped (the 45nm Yonah) and the architectural changes were as significant as Sandy to Ivy.

Holy shit /g/ all the enthusiasts are going to get Threadripper, right?
I mean, who's gonna buy i9? It looks like a mess from every perspective.

If the rumors of the 16core costing less than $1000 are correct, it would really turn the market upside down

>>60828416
I'll buy the 16 core version just so I can compare it to the Threadripper flagship

>>60828416
Threadly reminder that AMD showed TR at Computex and it had near 1:1 scaling with an 1800X. https://youtu.be/NeuamYdfU4w?t=16m

Usual time on a 1800X is 24 seconds.

>>60828638

If thread ripper runs infinity fabric at 2:1 ram speed (which has been speculated) in some scenarios you'd get over 1:1 scaling from a 1800x.

>>60828636
Are you going to do a review on them?

>>60828638
We're gonna need clocks for that comparison, that TR could be clocked anywhere from 3.6 to 3.3 depending on different factors

>>60828978
Estimated to be running at 3.5. As usual, we need to wait for the final product or announcements.

>>60827412
I have, but I will mostly only do gaymen so I don't think a 1700 would be worth it

>>60827311
>$2200 shitposting machine
Kill urself my man.

File: 0017 - 63Ys8qp.png (75KB, 499x499px) Image search: [Google]

75KB, 499x499px

>>60829072
>caring about what other people spend their money on
It's probably also a 1800-1600 shitposting machine if you take out the monitor, KH speakers, and any possible price drops

>>60823782
Meanwhile the 6-core i9-7800X is making mincemeat out of the 1800X. No wonder AMD is dropping the price on their shitty chips.

>>60829597
I love it how you've deluded yourself into thinking this nonsense.

>>60829597
>proud of giving intel his shekels buying overpriced shit for years
good goy

File: 41b6e_AMD-Ryzen-7-1800X_Geekbench-3_4.12-GHz-366x410.png (43KB, 366x410px) Image search: [Google]

43KB, 366x410px

>>60829614
>>60829631
This has eight cores, but...

>>60829661
>geekshit3

Lmao, how about looking at the live benchmarks done at Computex instead of that cachebound junk?

File: ac124_Intel-Core-i7-7800X_Geekbench_Stock-366x410.png (48KB, 366x410px) Image search: [Google]

48KB, 366x410px

>>60829614
>>60829631
...It performs worse in every metric compared to the six core Skylake-X
And remember that both of these CPUs cost the same, but one has more PCIe lanes and twice the memory channels, as well as a higher overclocking headroom that's not process-limited.

>>60829671


Geekbench is SH*T.

It actually seems to have gotten worse with version 3, which you should be aware of. On ARM64, that SHA1 performance is hardware-assisted. I don't know if SHA2 is too, but Aarch64 does apparently do SHA256 in the crypto unit, so it might be fully or partially so.

And on both ARM and x86, the AES numbers are similarly just about the crypto unit.

So basically a quarter to a third of the "integer" workloads are just utter BS. They are not comparable across architectures due to the crypto units, and even within one architecture the numbers just don't mean much of anything.

And quite frankly, it's not even just the crypto ones. Looking at the other GB3 "benchmarks", they are mainly small kernels: not really much different from dhrystone. I suspect most of them have a code footprint that basically fits in a L1I cache.

t. Linus Torvalds


Makes you wonder why they tested their brand new hardware in an outdated benchmark, to inflate the scores obviously, cinnebench and cpu-z benches done at computex are both in
line with Intel's own performance claims of 10%

>>60829671
>cherrypicking benchmarks
Can it, AMDrone. The cheapest Threadripper isn't looking so hot from a value standpoint, since the 10-core has to compete against the eight-core i9-7900X, which not only has better single AND multi-threaded performance, but can also be overclocked higher for better performance.

>>60829676
>LN2 oced
Oh lol you came back.

>>60829742
Delicious tears.

>>60829742
You're not fooling anyone you retarded fuckstick.
GB3, rofl.
Even Intel claims their new shit is 10% faster in MT than equal core broadwell's, that's pretty much what they can do with Skylake clocked slightly higher and everyone knows that.
Keep dripping those tears, maybe you'll need a refill soon

>>60829750
It's stock speeds, you idiot. Even down to the memory.

>hwbot, cinnebench and cpu-z all show consistent performance increases
>meanwhile greatshit3 performance numbers are completely retarded and its a well known flawed benchmark
>Intel didn't even have the decency to run GB4

Nice cherrypicking, you filthy kike.
Get back to the gas chamber.

File: 1496768285799.gif (160KB, 749x578px) Image search: [Google]

160KB, 749x578px

>>60824651
>that socket
>10gb aerion card included

>>60824742
No wonder
The only thing Intel done on the last 5 years is reducing power on ultra low consumption skus. For the rest it's the same shit

>>60824697
Ebinyc and threadripper CPUs use the same socket so they wont need to change the socket.

I got a job which I love, and want to continue at home a bit. Having more cores really helps with some programs so I'm really glad for this. I currently have quad core and it's hell compared to the work machines.

Honestly is there any good benchmark out there anymore? Cinnebench and Blender scale too linearly and focus on the FPU, CPU-z is just generally shit, Firestrike and other similar type is far from realworld, Handbrake likes really fast memory, random photoshop/adobe benchmarks are about as reliable as an Intel atom.

That leaves the gcc subtest of SPEC, but nobody tests that besides anandtech once in a blue moon.

File: 1477228765510.jpg (298KB, 531x717px) Image search: [Google]

298KB, 531x717px

>>60829782
>stock speeds

>>60829742
>Faster single core performance
1/3 of pcie lanes due to greed
Thermal jizz and throttling

>>60829661
>>60829676
if I recall correctly, in some previous thread you were posting the same exact shit and you were called out for posting 32bit benchmark of ryzen and 64bit of intel

if you're going to use Geekbench (fucking lol) then just TRY to keep the build the same.

>>60829902
Too poor for LN2, goy?

>>60829882
All synthetic shit should be dropped from benchmarks, handbrake , blender and adobe premier and others are actual pieces of software that are used for work, so leave those in.

>>60829907
Not the problem is that is a little troublesome to work while pouring liquids over CPU. The advange is that you can also instant cool soda

>>60829736
I wonder how the dev feels when his precious got absolutely slammed by one of the most talented software developers on the planet and the owner of the most widespread kernel in history.

>>60830010
Considering how most 'modern' software developers are oversized manchildren, probably something along the lines of "linus is a big poo poo head and a bigot"

>>60829742
LMAO

Meanwhile in reality https://www.hardocp.com/news/2017/06/05/intel_core_i9_skylake_e_clocks_no_higher_than_broadwell/

>>60829805
>>that socket
>>10gb aerion card included

The sad things is that it's fucking theoretically redundant, since Zeppelin can just turn 1 PCIe lane into 1 10GbE port (limit: 4) but that AMD decided not to give low-end consumers a bunch of free 10GbE.

>>60831568
You need at least two PCIe3.0 lanes for a single 10Gb connection. 10Gb = 1.250GB, and a single PCIe 3.0 lane is about 980MB/s

File: this kills the SL-X.png (3MB, 1651x1475px) Image search: [Google]

3MB, 1651x1475px

>>60831656
prepare to be pleasantly surprised.

Zeppelin has 32x (plus 2x for some management shit) 12.5 Gbps general IO PHYs, up to 6 of which can be SATA and up to 4 of which can be 10 GbE.

PCIe 3.0 is ~8 Gb/s using 128b130b coding (~1.5% overhead), and 10GBASE-KR/SR/LR/ER are all ~10.3 Gbps using 64b66b (~3.1% overhead). The 12.5 Gbps PHYs allow up to 10 Gb usable bandwidth with lower latency 8b10b coding, which is likely what 2-socket EBYN is going to use for its xGMI Infinity Fabric bridging.

>>60831811
>AMD not fucking around with their PHYs
That's pretty EBYN, pham
I might ditch my Xeon-D for this if there are sub 45W TDP SoCs.

>>60831811
Man Zeppelin has so much redundant server shit built into it it's kinda insane.

>>60831855
Zeppelin is literally just an 8c Broadwell-D knockoff that can do Voltron mode.

It's main value in 1-die SKU is just that AMD isn't charging jewish pricing levels, so a switch would probably be a dubious sidegrade for you honestly.

>>60831950
I need more 10Gb connections, either SFP+ or Cat6As. Two isn't enough anymore.

>>60823832
J E W K I L L E R

>>60832017
C O N T R A C T K E L L E R

>>60823832
CERTIFIED

>>60832002
the caveat is that 2 of the 10 GbE controllers are "big" implementation and 2 are "little", which presumably means less offload acceleration and more software/CPU overhead for handling data on them.

why not just get a switch though? they've actually recently gotten rather affordable.
> Ubiquiti 12x SFP+/4x 10GBASE-T = $500

>>60832085
SHIT

>>60831950
>voltron mode
Oldtron or newtron?

>>60832022
Keller worked on K12 and Infinity Fabric, not Zen.

>>60832132
I'm an oldfag who just barely is aware newtron even exists, so either oldtron or maybe both I guess.

> assuming it's basically the same shit just with more SJW writing than the 80s.

File: IntelCorei77820XLN2Overclock_GeekBench3.jpg (389KB, 1680x1050px) Image search: [Google]

389KB, 1680x1050px

>>60829661
>>60829597
>>60829676
>>60829742
Nice cropped ln2 benchmark, faggots.

>>60829804
It's worse than that, they are cropping ln2 benches because Geekbench itself shows stock clocks in the app instead of the actual overclocked clock rate.

>>60832805
>>60832821
I wonder how many jewish tears will be on /g/ on 20th of June.

>>60824010
>Holy shit, how old are you?
old enough to need moar cores for work

time is money, goy

>>60832843
Keep in mind that Intel intends to price the 10-core at $999, where it will be competing on price against a 16-core Threadripper.

Also keep in mind that rumored specs suggest a 12-core Threadripper with 3.6/4.1 + 200mhz XFR sku that will be cheaper than the 10-core Skymeme-X

Also keep in mind that Threadripper is rumored to have the 2:1 infinity fabric speeds of Epyc which hilariously means it should have better single core performance than Ryzen at the same clocks.

>>60832923
I'm not talking about jewripper. June 20th is the day EPYC drops on market. Imagine the salt, imagine the tears of every jew crying in uniform.

>>60832923
>Threadripper is rumored to have the 2:1 infinity fabric speeds of Epyc which hilariously means it should have better single core performance than Ryzen at the same clocks.
Engineering sample Ryzen chips had it running at 2:1 speed, apparently it doesn't do anything for single core performance, but draws more power since the IF is running faster.

>>60832939
Fair enough, though I expect those tears to mostly be at Intel itself. The fanboys will be happy to ignore it unless the usual suspects start publishing comparison benchmarks of server parts.

File: ryzen-memory-overclock-geekbench-chart.jpg (66KB, 939x306px) Image search: [Google]

66KB, 939x306px

>>60832949
That doesn't make any sense to me. IF speed should directly impact cross-CCX L3 latency and RAM latency, and in fact appears to do so as faster RAM does in fact appear to improve single core performance on Ryzen.

I don't see how it could simultaneously be the case that faster RAM improves single core performance on Ryzen while doubling the IF speed relative to the RAM speed would not.

File: beardsley.jpg (177KB, 703x737px) Image search: [Google]

177KB, 703x737px

>>60823782
>Whitehaven
racists

>>60833020
Because only gayms actively use cross-CCX talk and no one cares about gayms.

>>60833044
Then why would RAM speed on Ryzen be affecting single-threaded benchmarks more generally? Do you have any evidence to show IF speed doesn't affect single-threaded performance on Ryzen?

>>60833128
RAM speed affects Intel's single-threaded results just as much.

>>60833149
Just so I'm clear, are you then contenting that increasing the IF rate on Ryzen independent of the RAM speed itself would have zero performance effect on Ryzen?

>>60833357
Probably.

>>60832939
>Imagine the salt, imagine the tears of every jew crying in uniform.
second dead sea when?

>>60824010
DX12/vulcan is made in such a way that it takes advantage of as many cores as possible.

Multi threaded programs shouldn't be 'choosing' to use X amount of threads; they should use as many threads as possible. The CPU clock game is pretty much done. 4.5 GHz to 5 is an 11% increase in output, but might require 30+ % increase in power consumption, and comes with reliability and longevity issues. the best way forward is to increase core and thread count.

All that said, prob if you're not gaymen/workstation/etc you prob don't need more than 4c/4t.

Highly considering the 16c/32t Kikeripper.

Considering that it'll be available in ATX formfactor, it'll enable me to rackmount it in my A/V server rack (~18" - 19" mounting depth) to virtualize all the stuff I want from a 4U chassis.

Up until now, to get that many threads you _needed_ to go dual socket, and with dual socket, you _needed_ to go SSICEB form factor, or similar which made it wider, not to mention that you often lost PCI-E lanes on these boards too.

Finally my dreams of being able to virtualize a game server hosting VM (ARK, TF2, etc.), and virtualize pfSense/Edge/etc for a router, a video capture PC, a guest PC for playing vidya when guests come over, etc will be able to be possible from a single highly space efficient chassis and the wonders of PCI-passthrough.

>>60834836
>Up until now, to get that many threads you _needed_ to go dual socket
not recommending, but 22c xeons have been out for a while

>>60823782
shouldn't this be really good for coin mining? like, can you split 8x pci-e lanes to 4x and have 12-16 video cards for mining, or are there limitations /prinicpal pci-e lanes?

Not interested, but trying to figure out an audience.

>>60834985
The 22c Xeons have been out, but in usual Intel fashion, it's like 2 grand and doesn't scale well.

Threadripper sounds more ideal for what I want in an all-in-one VM workhorse, especially considering pricing given how they used Ryzen 7 to blow Intel's $1K CPUs away.

>>60829736
>SH*T
Kys you're self

>>60833757
On 20th of June.

>>60825234
>that pic
Spilled my drink as a Pole.

>>60823782
I HATE AMD SO MUCH !!!

File: pissed.jpg (73KB, 600x600px) Image search: [Google]

73KB, 600x600px

>>60838116

File: 1490983023651.gif (932KB, 350x291px) Image search: [Google]

932KB, 350x291px

>>60837816
so you can confirm this too?

File: chrzanić.jpg (17KB, 328x328px) Image search: [Google]

17KB, 328x328px

>>60838170
Totally. There were some shifts in polish that ended changing kh to ch (which is just "h") and "cz" and "ć" are pretty similar to foreigners, so it's plausible they're actually the same words, just anglicised. Though, to be correct, "to fuck shit up" is the meaning of "Schrzanić". "chrzanić" is "fuck this shit", as in, not to care. Intel chrzanił AMD and now they're like >>60824601 >>60824680
Can't wait for AMD to dethrone Intel

>>60824033
Single core is all you need.

>>60832127
WRECKER

>>60839387
For most things I do, single core is plenty

But I like being able to have a lot of those things active at any given time...

>>60835007
Actually, ASRock has an entire series devoted to that.

>>60827176
I think that it mainly depends on the HBM2 speed - some stacks may reach 1000 MHz, some only 700 MHz (see >>60823794), so even the same Vega die may be used for different brackets (high range-mid range).

File: 20170529-5[1].jpg (315KB, 1200x600px) Image search: [Google]

315KB, 1200x600px

>>60835007
Imagine having to cool 16 GPU's (even if you manage to fit them in all 16 slots). Never mind the power they would require.
.

>>60845732

This is why 2kw psus exist.

>>60845732
holy shit they actually made a mobo for it

>>60845732
You can fit far more than 16 GPUs there with risers

>>60845932

>muh 480 exceeding pcie spec