Skylake-X is going to be big, projected of a 15% IPC increase

>>59894596
provide a source on the pricing that isn't a pajeet site

>>59894596
Just like how Kaby was a 15% increase over Skylake?

>Intlel fags happy for 6 cores that cost twice the competition's.

File: Intel-8th-Generation-Core-i7-8000-Series.png (402KB, 1683x948px) Image search: [Google]

402KB, 1683x948px

15% or "15%" ?

>>59894596
Source on those prices.
To produce a whole 8-core product without the use of core complexes is insanely expensive. Hence why Intel sells them for $1600

15% IPC increase with no major architectural improvements is near enough impossible unless you want to compare one specific synthetic.

Extra lanes are never a bad thing, especially as Naples is looking to rape Intel's enterprise line in terms of PCIe capability and core count. And efficiency. 3 things that are huge to enterprise customers.

Sceptical would be too far from the truth.

> Skylake-X is probably the most exciting HEDT product coming from Intel since the original Core i7, and I'm sure they want to make an impact here. One friend told me some applications/games are seeing 10-15% performance per clock increase compared to original Skylake, that's probably because of the new cache architecture and quad-channel DDR4. Expect the entry level 6C/12T to become a very attractive all-around choice for productivity and high-refresh/multi GPU gaming choice.

>>59894676
Is this another "My Dad works at Nintendo" source?

>>59894596

i doubt its a 15% ipc increase. probably equal perf to skylake. AMD will have a hard time competing no matter, zen only matches broadwell in specific workloads, a shit uarch that regressed in terms of oc headroom and per clock perf.

File: Screenshot from 05-04-2017 23:59:51.png (94KB, 879x714px) Image search: [Google]

94KB, 879x714px

Shitlake is dead on arrival.

Only thing Intel has going for them at the moment is single core performance. You add more cores, that advantage goes away. And it'll still cost more than AMD's Ryzen. The "15%" is just more marketing horseshit.

>>59894697
I think a lot to do with Ryzen's somewhat underwhelming OC results are simply due to node immaturity, and motherboard vendors still getting their shit sorted.

Even since Launch, we've seen massive improvements in how Ryzen performs on the desktop. I think it may well be some time yet before motherboard vendors finally get everything in place and allow Ryzen some room to move around.

>>59894596
>projected of a 15% IPC increase over standard Skylake

It is standard Skylake, shitposter-san. That's quite literally how Intel's HEDT platform works. That's quite literally why it's called Skylake. It's the exact same archetecture, and won't reach the same sort of clocks as the quad cores either.

>>59894695
my uncle rather

File: 1484686763369.gif (36KB, 265x200px) Image search: [Google]

36KB, 265x200px

>>59894596
>competing
In what? Stuttering even at 720p when overclocked to 4.9Ghz?
https://www.youtube.com/watch?v=qswrFCb5sMQ

Yeah, it'll have trouble competing there since it runs all games smoothly.

Skylake-X launches at IDF 2017, with probably another shitshow mobo situation like every HEDT platform in the last 5 years, that's way too late, both since Pinnacle Ridge is 4-5 months after and because everyone who already got a Ryzen or Broadwell-E(heck, even Haswell since similar IPC) won't upgrade.

>>59894676
>that's probably because of the new cache architecture and quad-channel DDR4

Hmm, funny how no previous HEDT rehash of an arch has ever seen these improvements, despite benefiting from those exact same things. It's almost like it's completely made up or something...

>>59894769
Never underestimate the wallet of a fanboy.

Value for money is a concept lost on all but a select few it seems now.

>>59894769
To add insult to injury the 12 to 16 core Zens are launching in June.

File: Screenshot from 15-04-2017 00:52:23.png (35KB, 590x244px) Image search: [Google]

35KB, 590x244px

>>59894596
>5 months after everyone and their mother bought Ryzen

>>59894817
No one will buy Intel's overpriced garbade anyway.

>>59894817
>>5 months after everyone and their mother bought Ryzen
Lel, it's the same situation as Fermi vs Evergreen half a year late, 10% faster, twice the power use and 50% more expensive.

>>59894799
Nice that AMD is finally viable now as a replacement.
At least to the average CPU consumer.
I've growing tired of Intel these past years as well.
:^]

>>59894596
>""'''"15%"""""
So the improvement will be negligible then?

>>59894596
What's the purpose of Skylake-X besides a non-XeonPhi entry point into HEDT Optane and AXV512 programming ?

>>59894875
They have to produce milk somehow.

>>59894596
>only 10 cores
>not even 6 channels like the Xeons
>no ECC

What's the point of this abortion?

>>59894817
>launching at the Israeli Defense Force
Can't make this shit up folks

>>59894896
Ehh, they are pretending they can compete against AMD.

>>59894914
>10 cores
>competing against 16 with ECC support
don't think so.

I hope Ryzen announces the 1900 and 1950X pricing the day Intel launches Skylake-X.

>>59894896
>he thinks Skylake xeons all get 6 channel memory
Only the 20+ core ones have that, memory controllers are spread through different core complexes, even Haswell-EX is like that, it's not a monolithic memory controller, that's insane.

>>59894925
They are pretending. No one actually takes this launch seriosly.

>>59894925
>I hope Ryzen announces the 1900 and 1950X pricing the day Intel launches Skylake-X.

>1900/1950X(tentative) is released in June, Skylake-X is middle of August (IDF) to last week of August(gamescon)
With availability probably 1st of September

15% IPC is possible only in specific cache bound applications due to Skylake-X 4x the L2 but 3x less the L3, with no register, load/store/retire changes, needless to say it'll make a great dedicated CPU-z architecture!

>>59894596
>JUST WAIT(TM)
Intelfags remain on suicide watch.

>>59895025
My-my, the times have changed.

File: AMD-Ryzen-1800X-Intel-6950X-1.png (29KB, 617x837px) Image search: [Google]

29KB, 617x837px

>>59895023
Reminds me that Ryzen has situations where it demolishes the Broadwell-E in certain single threaded tasks even at the same single core clock.

>>59895061
Wow, it's almost like IPC isn't static or something.

Fucking witchcraft.

>>59895061
That's the fucking 10 core BDW-E and it's still doing so good against it.

>>59894596
>the 6 core projected at $400
Meanwhile, you can get a Ryzen 1600, motherboard and RAM for $400, then OC it easily with the included cooler.

Unless you're retarded, a fanboy (same thing really), or filthy rich, it makes zero sense to go for Intel even it it does cost $400 just for the CPU.

>>59895129
You won't even be able to get the Skylake-X six core until September, I dunno what OP's thinking.

If I had to JustWait(TM) for 5 more months I could JustWait(TM) another 4 months for Zen2

>>59895156
>I could JustWait(TM) another 4 months for Zen2

there's no way this would come that early, right?

>>59895156
>I dunno what OP's thinking
>"This'll get me so many (You)s hehehe! :^)"

He was right.

>>59895170
maybe they'll pull the radeon move and just rebrand shit with slightly higher clocks each year

>>59895170
Pinnacle Ridge Q1 2018, AMD has been releasing Bulldozer upgrades every year consistently for an arch that nobody cared for since 2011, and they even promised a Tock Tock Tock release cadence for Zen.

>While Anderson’s responsible for bringing Ryzen to market—“you don’t have any idea how many hours I and my team have spent on this,” Anderson said—it’s Papermaster who has to think of the future. When asked how long Zen would last, compared to Intel’s two-year tick-tock cadence, Papermaster confirmed the four-year lifespan and tapped the table in front of him: “We’re not going tick-tock,” he said. “Zen is going to be tock, tock, tock.”


And considering I can find at least 10 different things to improve for Zen2 compared to Zen on the top of my head, we can very well expect a good IPC increase.

>>59894952
WHAT?
I thought Skylake-X would be the one in June/July and 1900/1950X(tentative) would be in August or September.

>>59895025
>JUST WAIT(TIM)

>>59895061
Holy crap that VirtualBox. Why is that so good? I'm probably going to be running that on my Ryzen system.

Is Skylake-X's AVX-512 actually 512bit width, or 256bit * 2?

>>59894596
>the 6 core projected at $400
but will require a $450 motherboard or some shit

File: Screenshot from 15-04-2017 01:26:12.png (152KB, 601x186px) Image search: [Google]

152KB, 601x186px

>>59895279
see
>>59894817

>>59895279
The 32 core Opterons are releasing in late Q2, why would you think the 1900/1950X(tentative) are in Aug/Sep?

>>59895298
It's Skylake so 256*2 afaik.

>>59895298
>512b FPU pipe

That's literally meltdowner-tier

>>59895368
>That's literally meltdowner-tier
It's fucking Intel, they always resort to housefire whenever someone competes against them.

>>59894596
>Remember how Intel always says new gen is 10%increase when it actually 1% episode.

>>59895373
It was just a joke, there's no way they have a FPU pipe that huge, it's skylake so 256*2 is logical

>>59894596
>15% IPC increase
Just like Kaby Lake, no?

>>59895310
Does that say that production weeks will be week 22-24 of 2017 for just some of the wafers? lol.

>>59895353
>The 32 core Opterons are releasing in late Q2, why would you think the 1900/1950X(tentative) are in Aug/Sep?
because apparently it takes mobo manufacturers months to make boards.

>>59895061
>all the synthetic benchmarks and crap that I never use
>ryzen equal or slightly worse
>all the crap I actually do use
>ryzen better or equal

nice

>>59895357
Oh, I see, thanks. No improvement from AVX2, then.

File: Screenshot from 15-04-2017 01:35:14.png (72KB, 939x891px) Image search: [Google]

72KB, 939x891px

>>59895061
It gets even better on a nonshit system not using ICC

>>59894596
>>59894635

So 1.5% increase? not bad, last time it was only a 1% increase.

>>59895299
desu $400 6-core is DOA when I can get a $250 6-core from AMD

And by the time Intel's 6-core is out the R5's will be $20-$30 cheaper than they are now.

>>59895400
Motherboards are made 2 months in advance, how else would you otherwise get microcode and EFI up to snuff?
Also these 16 core ones will be using actual good shit from Supermicro and not gaymer shit, so that will be neat.

File: 1488745053176.png (290KB, 579x713px) Image search: [Google]

290KB, 579x713px

>>59894596

>>59895448
God I hope Supermicro has a 2P board for these 16 core Zens, I know moving to 32 core Naples boards will be way too expensive.

>>59895353
>>59895448
In all seriousness, because I assumed that the HEDT platform would come significantly later after the server ones.

Like I figure Snowy Owl would launch around Naples, but I figured they'd just pace out the launch more for the 16 and 12 core HEDT SKUs.
I still feel this is probably true and it'll be like August or September for "X399" or whatever.

>>59895402
It's not all that fair a comparisons as it's being compared to the $1700 Broadwell 10 core.

File: Screenshot_20170415-094557.jpg (261KB, 1080x1920px) Image search: [Google]

261KB, 1080x1920px

>>59895429
>>59895516
>>59895530
Pretty fucking funny.

Intel has been wrecked.

>>59895516
>In all seriousness, because I assumed that the HEDT platform would come significantly later after the server ones.
It's usually the other way around isn't it? The server ones need much more validation and that takes time.

The only reason the HEDT ones would release after the server ones is because server demand is even higher and they're re-purposing the HEDT for certain server SKUs

Which doesn't make all that much sense because server SKUs are made in conjunction with server customer demand, if they want a SKU, AMD will make them, there's rarely ever unused SKUs found in that market unlike the consumer ones, which seems to majorly focus on like 2-5 SKUs per line.

OH BOY A NEW SOCKET THANK YOU INTEL PLEASE TAKE ALL MY SHEKELS!

>>59894596
So you mean AMD actually made Intel drop their insane prices and make an actually measurable IPC gain? Wow, AMD is doing good this year.

File: delid.png (234KB, 882x758px) Image search: [Google]

234KB, 882x758px

>>59895569

>>59895429
>desu $400 6-core is DOA when I can get a $250 6-core from AMD
and the AM4 boards aren't as expensive as intel's HEDT shit

>>59895597
Unverified rumor horseshit.

>>59895597
15% IPC outside of canned benches that don't exit the caches seem to be extremely questionable just from cache size changes, in fact It's fucking impossible.

>>59895061
Outdated. Prime, Doom and AotS performance was fixed.

>>59895642
Oy-vey delet this.

>>59894596
>15%* IPC increase over standard Skylake
>measured *In CPU-Z

>>59895569
>meanwhile kiddies cherrypick benchmarks for gaming
intel btfo, naples when.

>>59895616
and as usual, socket change for every new intel cpu.

>>59895616
How much is an AM4 mobo compared to a mobo for a 6900K.

File: Ryzen-5-1600X-6700k-init.png (23KB, 617x617px) Image search: [Google]

23KB, 617x617px

>>59895647
It's a day 1 bench, but S:A never did a retest with the 1800X, here's with the 1600X, and yes, there's obvious changes now with Prime and WinRar

>>59895061
Doesn't the 6950X cost $1400, and the AMD 1800X only costs $500?

That's ridiculous that it has comparable performance.

>>59895664
If you don't want to overclock(and if you got a 1800X you don't need to) you can go with a $90 AM4 mobo.

>>59895584
>In all seriousness, because I assumed that the HEDT platform would come significantly later after the server ones.

>It's usually the other way around isn't it?
Usually, but Naples is literally just 4 1800X CPUs put together under the same IHS on the same socket, no?
So Ryzen 7 launch was able to test and iron out the kinks for both Naples and Snow Owl so they can get those server contracts they'll need to fulfill before they think much about selling 12 core or 16 core desktop.

>>59895664
$80 vs $250.
You can find X99 as cheap as $150, but they're garbage. You can get an actual good B350 board in the $65-$105 range.

>>59895569
I want to build a little NAS box with a ryzen power sipper cpu

>>59895717
>If you don't want to overclock(and if you got a 1800X you don't need to) you can go with a $90 AM4 mobo.
What? You can overclock on a $90 motherboard.

>>59895734
Yes, but there's no need to overclock a 1800X at all, overclocking (as in moving the multiplier) disables all


>When the "OC Mode" is activated the SMU will disable the voltage controllers, which among other things disables the automatic voltage offsets. This can create an illusion that the power consumption is heavily increasing due to the use of "OC Mode". While technically it is accurate, it is more of a consequence than the actual reason. A vast majority of the increased power consumption comes from the now disabled automatic negative voltage offsets, which causes the actual CPU voltage to increase anything between 50 and 150mV. Because of this behavior, it is advised that the user doesn't increase the CPU voltage right away (when overclocking), but only upon actual demand (as usual).

>One of the major down sides of the "OC-Mode" is that upon activation both Turbo and XFR will be disabled as well. Basically, this means that unless you are able to reach at least the default MSCF / XFR frequency on all cores, then you will essentially be losing single threaded performance compared to stock configuration.

>>59895702
Holy, is that a fucking $330 Skylake i7? Jesus Christ.

>>59895664
where I live (straya cunt), X370 Taichi is $329 and there's only 3 boards more expensive.

there's B350 boards for under $150 if you're poorfagging it (and combined with the 6 or 8 core ryzen cpus you get a huge amount of performance for not much money)

cheapest shit-tier X99 boards are ~$350-450 while the ones worth having are all in the $500+ range (with the meme ANUS shit going for like $800+)

>>59895782
Yup, it's pretty amusing that some tests actually go way over the theoretical 50% increase in performance from 50% more cores, note that the 6700k is higher clocked as well.
It just goes to show that measuring IPC in a single application is retarded, IPC is not static.

The Blender, Virtualbox, VeraCrypt and Foobar results are nothing short of hilarious.

>>59895805
Here in canuckistan the x370 is literally $400

I don't know how the fuck anyone justifies a price like that for what you're actually getting. In comparison the Gaming 3 and Asrock Pro b350 boards are both ~$130 - $160.

>>59894705
>12 and 16 core
So three & four CCX units on a chip?
Here's hoping it still plugs into the AM4 socket.

>>59895717
How much is a mobo for a 6900K?

>>59895810
Can someone answer though, I don't get this:
Why is VirtualBox, virtualization software, so much better on AMD?

>>59895837
>12 and 16 core
>So three & four CCX units on a chip?
No. It appears it'll literally be 2 1600X or 2 1800X put under the same IHS.
>Here's hoping it still plugs into the AM4 socket.
nope

>>59895839
>How much is a mobo for a 6900K?
>>59895722

>>59895839
Non garbage ones start at $200

>>59895837
No way, those things are on SP3 or SP4, which are 3700+ pin LGA sockets, they're not even PGA

>>59895856
>Why is VirtualBox, virtualization software, so much better on AMD?
Some architectures are very potent at certain applications, the hell if I know why Virtbox performance is so high without some profiler tools.

File: feuerfaust.png (622KB, 1272x713px) Image search: [Google]

622KB, 1272x713px

>140W

>>59895726
I need some r3 shit for a sweet little box
wonder how much the cheapest r3 cpu is going to be.

>>59895856
>>59895881
Darn

>>59895859
Shit

Maybe Intel really is fucked this time.

>>59895898
AMD said they even had some error with their virtualization technology AMD-x or whatever it's called
(intels is vt-d or whatever)
and that is sad that it's getting btfo'd

>>59895959
Eh, Broadwell-E and X99 was a small market anyway, Intel's HEDT is too expensive, I've talked to some retailers in East Europe and Broadwell-E sales and secondhand Haswell-E sales have literally halted overnight when Ryzen was released.

>>59894596
This thing is a shitshow, due to the insane amount of private L2 cache per core this thing won't overclock any better than Broadlel-E

File: bipelines.jpg (253KB, 1310x888px) Image search: [Google]

253KB, 1310x888px

>>59894596
FTFY

>>59895836
fark, aren't our currencies basically at parity?

do you have fuckhuge sales taxes or are your chinks just more greedy than ours?

>>59896704
Uh what? No. Canada currency tanked.
It's like worth 30-40% less.
Google it, retard.

>>59895702
Not to bash on results, but isn't S:A just hate intel? Like recent optane stuff and other before that.

>>59897612
Who likes Optane? tard

>>59897484
not burger, i'm the strayan you'd replied to earlier

our currencies are like 1% apart senpai, combination of us tanking a bit and the eternal burger regaining their strength in the last few years

>>59894596
>50% more cores
>only 15% more performance
OY VEY THEY DID IT AGAIN

>>59899059
Oh. Yeah Canadian and Aus currencies are comparable.
But if they make prices TOO high in Canada, they get too much shopping migration to the USA.

All you Aussies have to do is push your country up against California.

>>59894596
>up to 10 cores
>while amd is doing hedt with 12 to 16 cores with basically two 1600 or 1700 strapped together, and it's going to be cheaper than the 6950X

Intel is truly finished

File: 1491054816249.jpg (20KB, 264x270px) Image search: [Google]

20KB, 264x270px

>Intel caught with their pants down
>can't easily increase core counts in the short term to match Ryzen's at the same price points

File: spurdo intel 2.0 again.png (992KB, 801x1500px) Image search: [Google]

992KB, 801x1500px

>>59894596
Why is there a long line in that image reminiscent of a long pipeline between northbridge and southbridge.

>>59895061
>foobar2000

chukled

>>59894596
(((Projected)))

Sure, it's another 3.5% increase and then 2 more rebrands of 2.5%

>>59894596
>6c up to 10c
>140W
>4c only
>112W
kek

>AMD promises 40% IPC increase
>It's 62%
>Intel promises 15% 5 times in a row
>4 of the 5 times it's been 0-5% increase
>This time is different though!

>>59899245
apparently we just need to wait a cool million years or whatever and the pacific ocean will disappear through plate movements and we'll be up there

for the moment we rely on the pc parts business being dominated by chinks who keep their markups low to compete on price

>>59899292
fuck i want a 16 core so badly for compiler and vm abuse

>>59894635
"""""""""""15%"""""""""""""

>>59899524
j-just w-wait! 10nm is coming!

top fucking kek

File: 1489100922662.png (246KB, 882x758px) Image search: [Google]

246KB, 882x758px

>>59894766

>>59894635
>Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors
>SYSmark
lmao

>>59899524
>It's 62%
Nope. It's more like 25% compared to Excavator's ST results and 35% compared to its multi-threaded results. Nowhere near 62%.

>>59894596
>140W
dropped

>>59899750
It's so easy to lie on the internet, huh.

>>59899790
Check Passmark and Anandtech's benchmarks. It's almost impressive that Excavator managed to beat Prescott from 10 years ago.

>>59899836
>shillmark
kek

>falling for hype this early

are you stupid or what?

>>59899836
>Passmark
How have you not managed to accidentally kill yourself by now through your mere stupidity getting you into a fatal accident?
I guess it's since you're underage. It'll come soon.

>>59899836
>>59899750
>this shit architecture which is not competitive
>and now this other arch that is only marginally faster is competitive

which is it?

>>59897612
S:A has always shat on Nvidia GPUs and AMD CPUs. This Zen fellating is a new development. He also pointed out in the Optane (which is a poor product all around) review that the biggest reason why he's so upset with Intel is that they've recently changed for the worse, which is true.

>>59900183
What is S:A?

>>59899750
>Nope. It's more like 25% compared to Excavator's ST results and 35% compared to its multi-threaded results. Nowhere near 62%.
prove it.

The improvement comes from the 256kb 8 way L2 to 1024kb 16 way L2 cache

>>59900535
>4x the cache but only 2x the associativeness
oh look it's fucking nothing. should've stuck with 512KB and 16-way.

File: 1466220830384.gif (1MB, 512x384px) Image search: [Google]

1MB, 512x384px

>>59894766
>games

>>59900571
We already knew it loses at productivity and everything else.

Losing at games too is the last nail in the coffin.

>>59894596
OY VEY! A NUDDUH FEEFTEEN PUHCENT! BETTEH BUY IT AND SEE THE ONE PUHCENT, I MEAN FEEFTEEN PUHCENT INCREASE!

>>59894596
Yes... Myeeeesss

Let's get this competish ball rolling

>>59899641
>10nm is coming
kek
So 7nm when, 2025?

>>59901891

Glofo is claiming initial run of 7nm in 2018. While it is still not as dense as Intel's nodes it would still be a 18-24 month lead on Intel.

>>59902086
Intel's density comes at the cost of atrocious yields.

>>59902122

Gotta make those monolithic 24 core dies my man!

>>59902171
It's 28c for Skylake Xeons now.

>>59894596
I guess I'll Just Wait (tm)
>for my system to stop stuttering :^)

>>59902122
It doesn't even matter when it's that late.

Can you imagine a third/fourth gen Zen with 8 cores per CCX and two memory channels per CCX?
Scale that up to 4 MCMs 64 cores and 16 memory channels, where the fuck would you even fit the DIMMs there?

>>59902807
Two memory channels per CCX doesn't seem difficult anyway, Raven Ridge is two channel memory with 4 cores (one CCX)
They can fit over twice on 7nm

>>59902807
That would be the end of Intel on every possible market. They have no chance of competing against that. If they actually can fit this in 7nm Zen.

>>59902828
>>59902807
Number of memory channels is limited by the IO density of the chip, the package and the motherboard.

tl;dr 16 channels is possible but it's a mess getting the 1300 signals for 8 channel memory working (150 signals per channel), doubling that will be a massive feat.

I think AMD won't go over 12, tops simply due to the difficulty of getting the traces out on the motherboard.
But that also means they'll have all the bandwidth they can, so they can completely skip half the work on the memory controllers, which is bandwidth and focus on latency.

File: Supermicro-H8QGi+-F-Overview.jpg (196KB, 548x439px) Image search: [Google]

196KB, 548x439px

>>59902935
>when you need to have a custom motherboard to fit all the possible DIMMs on it

>>59902967
That's 4P board. Fitting all the traces on 2P board would be a nightmare.

>>59902967
The funny thing is that the dual socket Naples has 32 DIMMs already and 2DPC , that's quad socket.

File: AMD-Naples-Project-Olympus-OCP-Server-Motherboard.jpg (283KB, 800x559px) Image search: [Google]

283KB, 800x559px

>>59902982
Well, they did manage to fit it on a 2P motherboard.
But going any higher than this will be somewhat difficult.

>>59903015
Unholy fuck that's a fuckton of memory for 2P system.

>>59903022
4TB maximum, using the largest 128GB DIMMs currently.
Though outside of O&G and big web services like Google and Amazon, nobody has that much money to blow on DIMMs, so I figure most of these will be used with 16-64GB ones.

>>59903041

>Though outside of O&G and big web services like Google and Amazon

Which is exactly why AMD is going for the throat - Intel has lost in every market except x86 they have competed in and AMD is looking to be a strong position to start eroding Intel's total dominance of big servers (well, the big stuff IBM doesn't have sewn up). What makes the zen design so dangerous isn't so much what is doing now, but what it can scale to and zen has been design to scale fron teeny tiny sub 10w deployments to massive MCM setups. It won't best Intel's extreme server chips due to lack of I/O but anything that isn't at the extreme ends of the scale is fair game for zen to take big wet bite out of Intel's backside.

That said for some of the current hot topic worklaods (machine learning for example) Nvidia is the bigger threat than AMD, though Vega is seemingly AMD's attempt to get a foot in the door there as well.

>>59903161
>due to lack of I/O
>64/128 PCI-E 3.0 lanes per socket
Wut.

File: 12122451787l.png (106KB, 1597x868px) Image search: [Google]

106KB, 1597x868px

>>59903161
Lack of I/O? Naples has 128 direct attached PCIe lanes, some 115 after the mobo gets its fill of SATA/BMC/USB/M2/Inifniband/etcetc

Intel has 44 direct attached PCIe lanes, rest is high latency stuff from chipset or a power hungry and expensive PLX chip

>>59903203
Holy shit the wonders of scaling, monolithic 650mm2 dies are truly fucking finished.

>>59894596
>AMD will have a hard time competing.
No, you say 'AMDFAGS ON SUICIDE WATCH'

>>59903203
what fucking package are they even gonna use for this shit?

>>59903237

Playing cyrsis. That or handling hundreds of thousands of users at one time.

>>59901891
After am4 next step is 7nm. So around 2021?

File: AMD-Naples-Zen-Platform_5.jpg (2MB, 2728x1816px) Image search: [Google]

2MB, 2728x1816px

>>59903237
SP4.

Somewhere around 4000 pins on LGA

>>59903250
>package
Can you read?
>>59903256
SP3 actually.

>>59903250
the fuck are you talking about

>>59903256
awesome, finally credit card sized CPUs lmao

>>59903262
Wasn't SP3 for SnowyOwl or the 16 core ones?

>>59903285
Nah, SP3 is server/HEDT LGA socket, SP4 - BGA for Snowy Owl, SP5 for laptos afaik.

File: 118424-AMD-Naples-Processor-3.png (21KB, 716x265px) Image search: [Google]

21KB, 716x265px

I guess more is always better.

Look at those clocks on a 32 core test board from June 2016(release date is June 2017)

File: 5cbe0f537cb8f4c4f51fe0b62ed35d13.png (182KB, 326x327px) Image search: [Google]

182KB, 326x327px

>>59903256
>128 big core threads per node

This is clearly illegal in some places.

Well, at least we'll get a new socket.

Cheap broadwell Xeons when?

If I can get a 6C (or even 8C - but not likely) Skylel CPU in the 4.6-4.8+ GHz range with 44 PCIe lanes I'll be very happy. I'll have to wait and see, so far Ryzen hasn't been impressive enough for me to want to upgrade from a 4.7GHz 4790K.

>>59904755
Skylake-X isn't overclocking that high, in fact due to its private cache sizes I it shouldn't overclock better than Broadwell-E

>>59904828
We'll have to see, so far Intel's HEDT shit has mirrored the mainstream counterpart to a certain degree in terms of OC potential, i.e. Haswell-E OCs better than Broadwell-E, just as Haswell OCs better than Broadwell. Skylake definitely OCs better than Broadwell and maybe even better than Haswell, so if Skylake-X doesn't have more potential than Broadwell-E it would be a reversal on the general trend.

The CPU market is pretty fucking disappointing though, Haswell is a 2013 arch and neither Intel nor AMD have managed to produce a CPU with at least 6C that also has a single-threaded lead (OC vs. OC). I'm going to be pissed if Skylake-X is another disappointment like Broadwell-E and Ryzen and I sure as fuck won't pay for an upgrade to a slightly faster Intel 4C CPU like a 7700K.

>>59904972
That's because SB to Broadwell HEDT lines were the same architecture at higher core counts on a lite version of a server chipset.

However that changes with Skylake-X, Skylake-X has 4 times the L2 cache of regular Skylake(and any other Intel design since SB, which were 256kb L2 per core), which unfortunately means lower overclocking headroom, L2 caches are run at core clock speeds and are very power hungry, so per core power consumption will notably increase with Skylake, so will voltages required to run the core.

On the other hand the L3 cache per core decreases by 3 times over previous Intel designs, but the L3 cache is run on its own separate clock domain so it doesn't really mean all that much for overclocking.

tl;dr
Skylake-X , bigger, hotter, more power hungry core = less overclocking

>>59905027
Didn't know that about Skylake-X, haven't really kept up properly. So how are they keeping the same TDP if every core draws significantly more power? Just by cutting down on L3? Skylake IPC isn't that much higher, so they can't really afford to drop clocks unless the L2/L3 tradeoff provides magically huge gains, which I kind of doubt.

>>59906995
Cutting down the L3 is your best guess..
Also they're not keeping the same TDP, the Skylake Xeons go up to 205W, Broadwell-EX topped up at 160.


There are actually 4 core Xeons with 60MB L3 cache, these things are rated at 150W TDP, and are only 3.20/3.50 GHz

Intel's L3 cache is very power hungry, and I guess they figured they couldn't scale it much further so they went with higher L2s

>>59907038

>Intel's L3 cache is very power hungry

I would wager that is the side effect of it being inclusive. AMD going the victim route probably helps a fair bit on getting power draw down.

File: pajeet_laughing.jpg (30KB, 650x500px) Image search: [Google]

30KB, 650x500px

>>59894596
>15%

>>59894596
>projected of a 15% IPC increase over standard Skylake

Nope. It's the same Kabylake cores, that are the same as the Skylake cores, so expect at most 1.5% increase.

File: Screenshot from 27-03-2017 18:07:18.png (208KB, 458x572px) Image search: [Google]

208KB, 458x572px

>>59907058
With such a low L3 cache with Skylake, it's not going to be inclusive anymore.

>>59907038
>Also they're not keeping the same TDP
I'm talking about what I may actually buy as a consumer, not Xeons. Everything I've seen says ~140W for Skylake-X 6-10C CPUs, which is the same as Haswell-E and Broadwell-E.

>>59907127
Ah, that.
That's true.

>>59895279
Amd is targeting the virtualization market, they have very good virtualization features including SEV, SME. I don't think Intel will have any of these features soon.

File: 1473016065470.png (1MB, 1280x720px) Image search: [Google]

1MB, 1280x720px

>>59903412
>Henkuma art

>>59894596
A whole lot of enthusiasm for this platform on /g/ and generally everywhere.
As fucking if, this thing will launch with a very long sigh from everyone.

What's even the point of this? Everyone who already needed high core counts upgraded to Ryzen and they won't sidegrade to Skylake-X for 10% higher performance for 30% higher power draw.

>>59894596
>>59894621

This. Intel has been giving us 5% increases that dont actually exist in real world use.

It's going to be glorious when all the E5 v4 xeons are dumped on ebay.

>>59907607
>A whole lot of enthusiasm
Really? Because all I see is a bunch of manufactured shills and actual excitement for Ryzen, which is here right now and provides some much needed relief from Intel's bullshit train.

>>59908242
Did you just read the first line and missed the "As fucking if" ?

>>59907109
>meanwhile at AMD, naples will use less power than HEDT uses (so less than 180W),while boasting a massive L3 cache of 64MB and 32 cores
>intel has 4 less cores, barely more than half of naples's cache, while using a lot more of power (205W)

Intel got lazy, thinking that AMD would never catch up on the server market, now THEY failed to catch up to AMD

I think I slipped to an alternate timeline at some moment

>>59909934
Anyone who was paying attention could see Intel was flailing around in other markets and ignored their core development process. You don't get 25% improvement in 5 years and call it progress.

>>59894596
What's the point of Kaby Lake X? You have to pay for the much more expensive motherboard but don't get any of the benefits like Quad Channel memory or the extra PCI-E lanes.

Also
>24x PCI-E on the chipset DMI'd through only 4x PCI-E
Why do they they insist on gimping their HEDT platform with this consumer tier shit. At least give it 16 lanes.

>>59910274
Kaby Lake X is just a way to make money from stupid kids who will replace their brand new 7700k for the 7740k because
>"wow, 100MHz higher base clock! Totally worth it! 7700k is truly obsolete!"

>still no real performance increase over an i5-2500k from 2011
Top kek

>>59895462
updated

>>59899836
>assmark
>poointheloo applel shill tech
what unbiased sources you got there