[ / / / / / / / / / / / / / ] [ dir / animu / asmr / rel / strek / sw / travis2k / vore / zoo ][Options][ watchlist ]

/tech/ - Technology

You can now write text to your AI-generated image at https://aiproto.com It is currently free to use for Proto members.
Name
Email
Subject
Comment *
File
Select/drop/paste files here
Password (Randomized for file and post deletion; you may also set your own.)
* = required field[▶ Show post options & limits]
Confused? See the FAQ.
Expand all images

File (hide): 9ec9bec8378b857⋯.png (2.04 MB, 2027x1524, 2027:1524, boardlarge.png) (h) (u)

[–]

 No.890426>>890472 >>890559 >>890827 [Watch Thread][Show All Posts]

Dual 8 core POWER 9 verses 96 core ARMv8

>John the Ripper faster on ARMv8

WEW

https://www.phoronix.com/scan.php?page=news_item&px=Raptor-TALOS2-Initial-Tests

 No.890428>>890432 >>890435 >>890599

>96 core chip performs better than 16-core chip(s)

You don;t fucking say

It gets more hilarious when you read between the lines and see how abysmal the ARMv8 is in terms of IPC compared to the Power9

Get the fuck out of here with your retarded chink shit


 No.890432

>>890428

ARM is such a POS, it honestly took me by surprise.


 No.890435>>890439

>>890428

>more cores means better

I bet you also think GPUs are better for everything


 No.890436>>890438 >>890827

File (hide): 31669d6aa4c4836⋯.png (292.26 KB, 600x517, 600:517, 31669d6aa4c48369c50aaabffd….png) (h) (u)

Now it just needs to be cheaper, have a better ISA, and have more than eight programs available that is able to use fucking 96 cores.


 No.890438

>>890436

Not to mention Talos II is mostly, if not all libre firmware and isn't gay chinkshit made by a tranny.


 No.890439>>890441

>>890435

The Talos II and Cavium Thunder X are similarly priced hardware so you would expect them to perform similarly. It will be interesting to see how it performs against Xeon and EPYC CPUs.


 No.890441>>890442

>>890439

>A CPU should have similar performance for GPU tasks


 No.890442>>890443

>>890441

Cavium Thunder X isn't a GPU you brainlet.


 No.890443>>890445

>>890442

>GPU tasks have to be run on a GPU

you can't read can you


 No.890445>>890446 >>890447

>>890443

>A CPU should have similar performance for GPU tasks

Where did I say this, bitch?


 No.890446>>890449

>>890445

>The Talos II and Cavium Thunder X are similarly priced hardware so you would expect them to perform similarly


 No.890447>>890449

>>890445

>A great many weak cores doing GPU tasks should have the same performance as many fewer strong cores because they cost the same.


 No.890449>>890450

>>890446

>>890447

Ah, I see. You are retarded. Good day, Sir.


 No.890450

>>890449

>GPU tasks are how CPUs should be measured

lel


 No.890472>>890475 >>890713

>>890426 (OP)

>$2000 motherboard


 No.890475

>>890472

>he prefers gimmick gamur rgb motherboards for his government spook ISA


 No.890559>>896800

>>890426 (OP)

>POWER9 compiles the Linux kernel in 73.9 seconds

wtf I love gentoo now.


 No.890595

Damn, the Talos is really kicking ass.


 No.890599>>890601

>>890428

Did you read the article? Talos II still won all but 2 benchmarks.


 No.890600>>890704 >>890807

File (hide): 03a6a3590a34114⋯.jpg (1000.29 KB, 3264x2448, 4:3, sac59lqtdio01.jpg) (h) (u)

>posts the only bench where the Talos loses against a machine with 2X the core count

>hurry look how bad it is guyze

Go back to halfchan


 No.890601>>890797

>>890599

What's the other? I only saw one loss.


 No.890704

File (hide): c8c84e58e60c762⋯.png (848.79 KB, 1024x722, 512:361, MATHS-INSIDERS-blog-14-102….png) (h) (u)

>>890600

>loses against a machine with 2X the core count


 No.890713

>>890472

It's low-volume niche hardware. What did you expect.


 No.890797

>>890601

You're right, It only lost in blowfish, I misread.


 No.890807

>>890600

It's six times also it only fails on jack the ripper all other tests show the 16 core ibm power9 raping the 96 core thunderX cpu.


 No.890815>>890818 >>890827

>preloaded with debian testing

what the fuck


 No.890818>>890830

>>890815

Too new to have support in stable maybe?


 No.890827>>890830 >>890834

>>890426 (OP)

The only reason john the ripper is slower is because it is designed with high frequency CPU's with less core's in mind. Except with the GPU functions via opencl and the like where it would perform better in high ram and extremely high parrellization workloads like the power9 is built for.

>>890436

Install gentoo you faggot and use opencl applications and go nuts with parrellization. Granted even gentoo isn't meant for this high of parrellization due to glibc being centered around high freq low core count. Needs its own OS to git gud. so this >>890815


 No.890830>>890832

>>890818

probably

>>890827

the speed improvements in gentoo are undetectable with powerful hardware. gentoo is pretty much obsolete garbage in 2018. too slow to compile things on old computers, speed undetectable on newer ones. the only thing it has going for it is customization, which servers don't really need.


 No.890832>>891098

>>890830

>speed undetectable on new ones

Prove it with benchmarks faggot. At the very least you will save energy/electricity in the very long term with a better idle from applications using less proccessing time. That and by installing gentoo you aren't limited to software devs compile for say debian. You can just compile it and run it yourself. For obscure architectures like power9 its awesome. But in the same breath glibc is shit for such parrellization so its better and more secure just to write a new OS from scratch.


 No.890834>>890859 >>891183

>>890827

It also needs its own specific compiler and libraries

https://developer.ibm.com/linuxonpower/advance-toolchain/

These benchmarks for CPU intensive tasks are pretty pointless unless it's being done with software designed to work with PPC beyond having it compile and run in a QEMU VM for a quick test. Which is the case with John the Ripper.

As an ex-PPC user I'm glad many projects make sure the software is portable but you're often stuck with old builds or terrible performance. There hasn't been cheap PPC hardware on the market in almost a decade, the people using old G4 and G5 macs have moved on long ago.


 No.890859

File (hide): 299b5e6a32c0871⋯.gif (1.94 MB, 282x157, 282:157, moses-kun.gif) (h) (u)


 No.891098>>891189

>>890832

>spend hours compiling every DE update

>save energy


 No.891183

>>890834

Not to mention, newer versions of some programs have removed Altivec support because of bugs, lack of interest, or programmers familiar with the platform. I was really irritated when I wanted to take my G4 Powerbook and realtek SDR stick out for a drive to see what I could receive, and VOLK (Vector Optimized Library) configured with generic support.


 No.891189>>897011

>>891098

>using a DE

OH NO NO NO NO


 No.896800>>897012

>>890559

With what options? That statement is meaningless without context.


 No.896823>>896851 >>897027

And on single threaded loads that 96 core ARM is slower than a 20 year old Pentium III. Power 9 meanwhile is so fast that x86 emulation on Power is currently the fastest x86 "processor" for single threaded loads on the market.


 No.896826

File (hide): d1b253fd04dc69c⋯.jpeg (317.4 KB, 2000x1000, 2:1, talos2.jpeg) (h) (u)

>16 cores BTFOs 96 core in all but one benchmark

>Another benchmark shows the Talos competing with a $3,700 x86 CPU

>this somehow makes the Talos II bad

x86 shills go away, freedom is here


 No.896851>>896863

>>896823

How? ARM-based smartphones have long since outclassed old i386 Pentium computers.

Anyway, buy the TALOS if you want. I don't see why you have to shit on ARM boards that are good enough for me or other people who don't need a big workstation.


 No.896863>>896890

>>896851

Faster memory and add-on chips like GPU's. And even those only go so far. Have a look at places like https://openbenchmarking.org/ that have results for the same benchmarks over several architectures and generations. (Or do your own benchmarks.) Even all four cores of the latest RasPi still get beaten by single core Pentium III's and AthlonXP's from the beginning of the century in benchmarks where those things don't help.


 No.896890

>>896863

Do not use Raspi as a benchmark for all ARM stuff, raspi doesn't even have out of order execution.


 No.897011

>>891189

That was just one example of a yuge package with a fast update schedule. Substitute GIMP, any web browser or whatever else fits the bill.


 No.897012

>>896800

The same ones as the other CPUs benchmarked. That's all the context you need.


 No.897027

>>896823

>And on single threaded loads that 96 core ARM is slower than a 20 year old Pentium III

doubt.jpg

>Power 9 meanwhile is so fast that x86 emulation on Power is currently the fastest x86 "processor" for single threaded loads on the market.

You are so retarded it hurts. P9 is still typically ~30% slower than fastest-clocking Xeons under single threaded workloads. It completely outclasses them in heavily threaded loads though.




[Return][Go to top][Catalog][Screencap][Nerve Center][Cancer][Update] ( Scroll to new posts) ( Auto) 5
44 replies | 6 images | Page ?
[Post a Reply]
[ / / / / / / / / / / / / / ] [ dir / animu / asmr / rel / strek / sw / travis2k / vore / zoo ][ watchlist ]