Buy i7 7700k

>Buy i7 7700k
>Hear about coffee lake
can I survive being a corelet?

Other urls found in this thread:

git.io/v5Fcl
godbolt.org/g/ZpkmhS
github.com/suaefar/ryzen-test/blob/master/kill-ryzen.sh
twitter.com/SFWRedditVideos

>got a 7900
>64 gb ram
>gonna be set for a few years with that sweet avx512

>>Buy i7 7700k
i was so close to making this mistake.

luckily i have a ryzen 1700.

corelets just cant compete with the 8 inch BBC cores from amd.

Don't worry bro, just upgrade to Coffee Lake when it comes out.

too poor

>can I survive being a corelet if intel releases a six core processor ignorant of the fact that AMD already has an eight core CPU that is faster than both?
Just sell your computer and sit in the dark because you were too dumb to buy the AMD processor.

AMD sucks though

>Retard who's too stupid to research current and upcoming CPUs thinks AMD sucks

It's like pottery.

why are you so angry

>SKL-X
>Only 512bit wide unit is the shuffle one. Arithmetic units are all 256bit wide.
>Run a single avx512 instruction and the entire CPU underclocks to avoid a housefire
>x264 gets a 5% boost at best using avx512 because of the throttling.

B-based Intel.

>being a corelet when Ryzen exists
dumbass

>Run a single avx512 instruction and the entire CPU underclocks to avoid a housefire
[citation needed]
it doesnt underclock the worst it does is powergating. And even then it's because 99.9% of applications out there do not use all the lanes.
AVX512 operates as a superset layer over AVX2 for the most part, which is too a superset layer over SSE4. The difference being that now you can have huge multi-lane operations done atomically, and on top of that with 32 registers, each of which being 512 bits. meaning you can apply fast arithmetic to 2048 bytes worth of register data.

I regularly work with low-level accelerated shit like this so this is quite a treat but for the brainlets that treat corporate activity as TMZ gossip they'll bark "but Sup Forums told me its only a 5% increase"

Depends on your needs. Laptops won't go past four cores anytime soon, and since most people only use laptops most consumer programs will be made to be able to run with four cores. You'll be able to do most things, but if you need superpowers I don't know.

>I regularly work with low-level accelerated shit like this so this is quite a treat
So do i, and it is quite a treat indeed. But it's just not yet there because of the limited amount of full width units and powergating/throtling or whatever the fuck these chips do as soon as an instruction touches a zmm register.

>but for the brainlets that treat corporate activity as TMZ gossip they'll bark "but Sup Forums told me its only a 5% increase"
Take a look at the #x264dev logs. The guy that wrote the 20+ avx512 functions himself said he got about that much boost on his SKL-X. Probably because a lot of the functions just take advantage of the k mask registers, implicit broadcasting and such, rather than the full 64byte regs.

>willing to upgrade
>coffee lake won't support 200 series boards

fucking dropped

what if my needs are games?
>inb4 manchild

what kind of low level work you do nigga

Writing simd optimized shit for multimedia use. What else would i do with these instruction sets?

>People tell you for months going that the ryzen cpu's knock the socks off pretty much everything
>you could have had 95% of the singlethread performance and 150% multithread performance for half the price
>or 95% singlethread performance and 300% multithread performance for the same price
>but "MUH GAYMES" won out over common sense and now you're faced with not only owning a deprecated CPU within 3 months of buying one, but if you do upgrade, you'll have to pay out the ass for an entirely new motherboard on top of it
>and you still have to delid and buy a separate cooler

You did this to yourself.

>multimedia use
you writing plugins for photoshop or some shit

Video codecs.

You'd be more depressed once you find out that Coffee Lake is already rendered obsolete since Ice Lake/Coffee Lake+ will come out next year with 8-cores

It's Z390 too so it will be the same 14nm+++++ and your Z370 mobo is already a useless brick before release

If you're an Intelfag you should upgrade next year

>faster than both
(lol

Why do AMDrones lie?

The 7700k shits on the entire R7 lineup for stuff that matters (gayming)

Don't even bring up the "It's good enuff" argument, with that logic no one should be buying high end GPUs

>Buy DDR4
>Hear about DDR5

Enjoy your compiling failures son.

>Spewing months old memes

>buy 1080 ti
>hear about volta

Most games don't even utilize all the cores available

>Volta
Not this year m8

Nvidia can easily wait until halfway of 2018 to shit out Volta since AMD's GPU division are inept hacks

Even if they announce Volta on next year's CES your 1080Ti still served you for like 4-5 months

wanna crit my dumb ass code:
git.io/v5Fcl

I wanna see what that user says so.. bookmark? Interesting discourse fellas

is DDR5 actually coming?

>he thinks AVX512 is useful and won't blow up his house
ohsweetie.jpg.bmp

This was pleasant

all right now, I'm calling firefighters for you user, hold on there

>simd to speed up array reversals
>simd to possibly do very fast merge sorts
>simd to bundle together sorting network operations

you're onto something here

Underrated. That is something nice. What if you add a "donate" button?

why would this merit donations

Have you considered submitting this to gcc so they may include it on their std::reverse implementation of libstdc++? Based on your benchmarks, it looks like they don't optimize it at all.

std::reverse absolutely does not have that optimization since its just implemented using iterations of std::swap I believe
godbolt.org/g/ZpkmhS

I wouldn't know how to propose something like that besides just tossing a link over and going "wacha think"

Well, they will very much welcome a patch sent to gcc-patches cleanly implementing your optimized simd code into their tree. But aside from that i don't think they will do anything with a link to a paper or even an external implementation.

Maybe try asking for pointers about how to actually implement this into libstdc++. 30x speed up for avx2 will surely sound appealing.

A sheckel for the good goy.

>higher latency at the expense of lower voltages
>need to run at higher frequency just to get the same speed as the previous generation's RAM
>but muh new RAM!! new = better XD

>avx512
>desktop
Nice meme.

Want to upgrade but new hardware punishes performance or will not work unless you have Windowns 10

>intel shitpost thread
>learn about simd

the fuck, is this shit for real? HBM based L4 cache can't come soon enough

>The 7700k shits on the entire R7 lineup for stuff that matters (gayming)
Have you tried playing more than one game at once?

>he uses Linux
Kek'd

>TFW bought 1700 in May
>TFW waiting for replacement from AMD after confirmed segfault and libx264 issue
You may have had a point like 4 months ago, but only retards like me who buy new stuff early can get shit on for that, it is binned out of the new silicon so now there is literally no excuse anymore.

Besides, it's not like Intel has a perfect track record for this crap either, at least AMD didn't just straight up gimp the processor via microcode like Intel did in order to fix theirs. Can you imagine the rage if a similar fault developed with Intel's memetastic AVX512 and instead of fixing it and replacing chips they just disable the instruction set as they've done before?

Any idea which batches have the problem? Or how to identify if your CPU is in th esegfault batch?

might as well end it now

Kill yourself back to Sup Forums

Sauce

Run the kill ryzen script.
github.com/suaefar/ryzen-test/blob/master/kill-ryzen.sh
Apparently all the chips made after June are OK.
Supposedly they just started checking for it in the QC process, they probably didn't know about it before so they didn't account for it in the binning process.

I think any 4c/8t i7 from the last 5 years will be totally fine for the next 5 years.

Blender

Eh, I think the Ryzen 5 6-core is like this gen's Core 2 Quad, while the i7 8700K and the Zen2 6 core next year will be the 2500K.

to be real i think 2core 8 thread will be fine for now until the 8K res meme comes along.

I think the 1600 and 8700k will be prime for 8k IMO.

Resolution only cares about GPU, not CPU.

Oh, sweetie

>buy i7 gaymen build two years ago
>crippling hikki depression
>no video editing done
>No gaming done
>nothing done
>literally a browsing machine

Should have bought a kabini.

Sweaty, you forgot to cut out his icon, now your image is double-paid for.

It's funny how the Kaby Lake i5 is not even shown

Intel is dead, user...

Ur right tho

But would an improper cpu would surely bottleneck the gpu(s).

Well, enjoy your single core performance that will most likely crush everything coffee lake has to offer.