Extra AMD Ryzen 7 1800X Linux Benchmarks

[phoronix]

Phoronix: Extra AMD Ryzen 7 1800X Linux Benchmarks

Assuming you have already checked out this morning's Ryzen 7 1800X Linux benchmarks, here are some more data points while putting the finishing touches on the Ryzen 7 Linux gaming benchmarks being published later today...

http://www.phoronix.com/scan.php?page=news_item&px=Ryzen-7-1800X-More-Data

[eltomito]

Muchos kudos to AMD!

[balouba]

Ryzen's results are higher than I expected compared to the Windows benchmarks (basically it seems to be a little slower per core in many work loads, but has more cores than most CPUs - on these bench though it seems faster in most cases)

[ruthan]

No games, no fun.

[Adarion]

It depends on the software to a certain degree. Optimized for something else? Bad luck. Made to use max. 2 cores? No luck with all the additional cores. This is something to consider when looking at benchmarks. If a benchmark was optimized for other architectures, or not doing good in parallel computing one has to keep that in mind.
On the other hand gcc can put the cores to good use for a kernel compilation, so there you can see the device really shine.
So I am very careful when it comes to gaming benchmarks. Besides, there seemed to a be a multitude of influences, firmware versions on the mainboards, memory config, HPET on or off and more - and they'd all influence gaming experience visibly. Give that area some time for optimisation...

[Danny3]

Hi Michael,

Our Phoronix.com server is still a bit slow from all the traffic today, but at least OpenBenchmarking.org and all these links are off on a separate, more powerful server.

A few years ago you wrote an article about G-WAN web server:
http://www.phoronix.com/scan.php?pag...tem&px=MTE0MzM

Since I've had heard about it some time before your article and made me really curious if such a performance it's possible I run a benchmark myself to see and added a comment at the news you posted.

About G-WAN

Hi guys,

I have more than 1 year experience with G-WAN. I discovered this server searching the web for web server benchmarks and I found this site:
http://nbonvin.wordpress.com/2011/03...server-to-use/ and I was intrigued by the speed of G-WAN. I downloaded it and runned some bechmarks:

On laptop - Core2Duo @ 2GHz (dualcore):
G-WAN ~ 36 000 RPS (requests per second)
Apache ~ 11 000 RPS
On desktop Core i5 @2.6GHz (quadcore):
G-WAN ~ 170 000 RPS (with weighttp 303 000 RPS)
Apache ~ 45 000 RPS (with weighttp 62 000 RPS)
.................................................. .................................................

In short, testing it on my 4-core computer with weighttp (which the developers claim it's multithreaded) showed ~300,000 K RPS, while Apache ~60,000 K RPS (G-WAN 4.8 times faster than Apache).
At the time, G-WAN didn't support PHP, so i had to test only with small HTML file (around 100 bytes, if I remember corectly, provided by G-WAN as example.
Since then, I think they've added support for PHP as their benchmark suggest.
http://gwan.com/benchmark/babel.html.
It's a shame that I don't have time and the same computer to repeat the benchmark to see how the performance changed over time.

The major thing that I don't like about it is that is closed source and the developer doesn't have any plans to open its source, but I found a post by a phoronix user which seems to have reversed-engineered it and made a similar server for which he has realeased its source code on github.
He claims on its github page that its performance on a Core i7 laptop is about:
320,000 K RPS for requests without disk access, and without pipelining and
290,000 K RPS when disk I/O is required, for files up to 16KiB.

A little while ago I was intrigued by GWAN's performance and decided to understand how it works. Using strace and other tools, I figured some things out and began working on an opensource alternative using just the general idea of its architecture (event-based, 1 thread per CPU, etc)... it's been a while since I've ran GWAN, but there were times my web server was actually faster -- but it isn't as feature packed, so that's not really a fair comparison. I'll perform some tests comparing with GWAN again another day.

In any case, feel free to grab it at http://github.com/lpereira/lwan (be sure to check the 'wip' branch) and try it for yourself. It's quite small (~2500 LOC of C code), runs on Linux/glibc only and does not support SSL, but I've been working on it only during my free time.

Maybe you can compile the PHP to static HTML files and serve those files with one of these two servers. to take advantage of the performance they claim.
Anyway, maybe you or somebody else here can do some benchmarks on either server of these two and post the results.
I am really interested to see if there's a way to serve PHP faster than Lighttpd and NginX.