The #ExtremeEdition Brand is about to get killed. What a big mistake

twitter.com/FPiednoel/status/1016360673318612992

>The #ExtremeEdition Brand is about to get killed. What a big mistake.

F

Attached: processor-box-core-i9ee-x-series-1x1.png (550x550, 145K)

Other urls found in this thread:

git.videolan.org/?p=x264.git;a=summary
obe.tv/about-us/obe-blog/item/50-avx-512-what-s-useful-for-us
arxiv.org/pdf/1704.08579
github.com/jeffhammond/vpu-count,
felixcloutier.com/x86/VZEROUPPER.html
twitter.com/NSFWRedditImage

It already died with Shitlake-X, along with Intel's credibility in the HEDT space.

i dont get it, it was born dead, how can it be killed again?

#shintel

But but nobody but threadripper

S

Gonna shill me a meme build I9 9900k, cirno approved. Inb4 >Jewtell. Lol being this poor

Not really. It started with the Pentium 4 Extreme Edition, and continued through the C2D era since Intel was trying to compete with AMD. The name's just run out of steam now since Intel can't do sane product lineups.

#ExtremePricing is dead.
Even Intel doesn't know if there will be 48 cores Threadrippers on 7nm.
Won't prevent them from 2 new platforms in between.
The level of cucked Intel is getting at is pretty impressive.

It's also not a good fit for the i3/i5/i7 and new i9 based branding system. It makes more sense to integrate any new 'special high end enthusiast' chips into that naming scheme.

scold me

Attached: Speccy64_2018-07-09_13-28-17.png (807x648, 42K)

Pooripper cant into avx so looks good

no need, just touch the heatsink while it's running

It was dead since Pentium 4.

>Pooripper cant into avx
>gotta do those avx512 FLOPs for muh high scool biology class

>too proud to buy kikeripper
>too poor to buy 7980XE

Why was the 7980xe $1800 when the 7960x was only $1300 for literally 2 less cores and faster clock speeds? Who would have been dumb enough to even buy the 7980xe?

Yeah what is even the point of a 512 bit register value in that context? There are not a whole lot of practical applications you can perform against a 512 bit value in that would be better than 256.

Also, what mainstream applications (with any appreciable market share) will be coded against such a fragmented feature set? It would be a huge waste of resources to code, test and support features that will only be used in 1-5% of your customer base.
>pic related

Attached: Screen Shot 2018-07-09 at 4.58.10 PM.png (1189x565, 169K)

i actually do avx512 research and no they are not "512 bit values" they are "512 bit vectors" meaning you can load in a full 64-bytes worth of a cache line and perform vector operations on it.
Meaning you are working with 16-32bit(ints or floats or whatever) values at once in parallel can be a huge speedup especially when working with matrices and image/video manipulation and simulation and such.

Its a niche "need" but a lot of code bases have conditional code paths to use avx512 should it exist. a lot of audio software and video software already have such code paths if they detect AVX512F or AVX512VBMI and such in your CPUID bits

Attached: Crotch.gif (249x166, 1.64M)

for specific software that use it, pretty much every video encoder including ffmpeg and most x264 encoders/decoders:
git.videolan.org/?p=x264.git;a=summary

some other video and signal processing firms:
obe.tv/about-us/obe-blog/item/50-avx-512-what-s-useful-for-us

and its accelerating some other already-existing algorithms even. And most compilers out there are emitting it where ever appropriate if the tuning of the target architecture allows it.
arxiv.org/pdf/1704.08579

also that's just 16 parallel operations for 32-bit values in a 512-bit register.

there are 32 total AVX512 registers. Meaning you can potentially do 512 32-bit vector operations very very very fast

(the most VPUs out there are 2, github.com/jeffhammond/vpu-count, so you arent exactly dispatching 32 opcodes worth of vector operations

What's with the coreclocks and multipliers? Is it automatic or something you can controll? Heat control, or power savings?

It performed worse than the normie chips. What did they expect?

>What a big mistake

Literally no one cares that the 0.1% of people who used Extreme Edition chips won't get new ones.

For extremely specific cases, I can see the benefit, but i can also see some performance concern. If AVX is anywhere in the execution pipeline, the CPU may throttle substantially depending on which AVX operation is being used. This could significantly slow down code paths which are tightly nested around AVX (loop counter increment instructions, etc).
>pic related

Attached: Screen Shot 2018-07-09 at 5.18.46 PM.png (763x551, 197K)

it's automatic. all processors do this. They proportionally downclock themselves when a core is finding itself "idling" more often and due to other factors for the sake of power efficiency.

I'm literally just browsing chrome with discord open so i'm not exactly putting all the cores to work for them all to be at 4+ Ghz right now.
Every multicore device does this for power efficiency and you can only really configure the min-max frequencies in certain ways but theres no reason to make all cores run at 4+ Ghz

open task manager right now and youll see that it's not always running at full-speed all the time.

if laptops did this then they would all run out of battery in only like 2 hours.

Attached: 1360908126545.jpg (1916x800, 149K)

Pretty sure it's a TDP limit so it will clock depending on the percentage.

if you're talking about power-gating this is all true and everyone is kind of annoyed by this fact but it's also due to the fact that AVX512 isnt popular enough for much of the upper lanes to be powered on all the time so it's an efficiency measure taken on intel's behalf and it's why instructions like this exist:

felixcloutier.com/x86/VZEROUPPER.html

Most uses of AVX512 are centered around tight AoS useage where AVX512 usage is "dived" into for a code path, used for how-ever long. and then opted out of. No one at the moment is sprinkling random avx instructions among regular x86 instructions. and most vectorized work using simd instructions make use of being "simd for a little while, then back to the usual x86" anyways. SO at the moment its a transitional phase where the workloads arent "popular" enough among the current code that is out there to keep those lanes alive and powered all the time.

Idling increases latency and puts the processor into higher power saving states regardless of clock speed and it has been this way for quite some time. Lowering the clock speed may reduce the voltage slightly but with how efficient Intel processors are lately, this is only really useful in extreme cases where you want to use as little power as possible, I.E. No modern laptop made in the last 4-5 years downclocks to save power, they do it if they get too hot.

Attached: wew.png (960x531, 156K)

(You)

I would but I just bought an 8700K. I've got no right to scold anybody.

lol

>8700k

Good God, user.

>I.E. No modern laptop made in the last 4-5 years downclocks to save power, they do it if they get too hot.
[citation needed]

That's scald you moron...

AMD > IBM > chinese AMD copiers > qualcomm > intel

why do the jews even try?