Jump to content

System Testing and Benchmarks


NoelC

Recommended Posts

I feel that occasional benchmarking helps ensure a system is working in peak shape.  I benchmark my system after updates to ensure Microsoft or a driver manufacturer hasn't botched something up.  Nowadays with Win 8 one no longer gets a direct WEI score from Microsoft (the old "7.9" type readout), though you can still get detailed results with winsat on the command line.  There are some good system testing tools out there, many of which are free, or at least free to run for a short time.

 

My own personal preferences for software to test and benchmark with are:

 

  • Passmark PerformanceTest (great overview, but also some excellent advanced tests)
  • ATTO SSD benchmark
  • RealTech-VR OpenGL Extensions Viewer

 

What are yours?

 

Don't be bashful to post your scores if you'd like.  Such threads sometimes seem like a competition, but hey, show it off even if it isn't next week's new model.

 

-Noel

 

Link to comment
Share on other sites


I don't like ATTO for SSD benchmarking, it was designed for spinners and just 'blows a gasket' above ~1000MB/s. Much prefer AS SSD Benchmark, great little program. CrystalDiskMark is fine too.

 

Aida64 has many useful tests, like its Cache & Memory benchmark:

 

W829TkCl.jpg

 

Linx for CPU GFLOPS horsepower.

 

Cinebench.

Link to comment
Share on other sites

Hi TELVM, thanks for your response.

 

Not sure what you mean by ATTO "blows a gasket"...  It seems to work okay above 1000 MB/s to me. 

 

Note that AS-SSD uses uncompressible data, so it generally measures smaller numbers than ATTO does for drives that internally compress data before writing it.  With drives from 2 or 3 years ago (the age of mine) that matters more than it does with modern drives.  ATTO uses compressible data, which yields better write times and has in my experience been more representative of real performance.  ATTO seems to me to be the tool to use to uncover developing problems (which so far, thankfully, I have not seen).

 

Also, low level caching (such as is done by Intel RST drivers, which I do not use) causes AS-SSD to read higher values, while that does not appear to affect ATTO's measurements when ATTO is set for Direct I/O.

 

AS-SSD_05_29_2014.png

 

  ATTO_05_29_2014.png

 

Thanks for the tip on Aida64.  I'll check it out,

 

-Noel

Link to comment
Share on other sites

I don't use any benchmark tools. If I have issues, I trace them with xperf/xbootmgr and sysinternals tools.

 

I appreciate your mad skills for sure, Andre, and thanks for your thoughts.  But how do you know whether you are developing an "issue" if you don't get objective readings of your system performance from time to time?  If you sense one, how do you know when it developed?

 

Because they don't watch their performance numbers, most folks never realized that Windows 8.1 actually slowed down the file system performance (though thankfully that seems to have gotten somewhat better with the latest Windows Updates).

 

-Noel

Edited by NoelC
Link to comment
Share on other sites

... Not sure what you mean by ATTO "blows a gasket"... 

 

VPTtgaG.png

 

"... ATTO is an old school benchmark, developed well before 64-bit computing. ATTO uses 32-bit integers. Since ATTO uses bytes to display the performance on the right side, any number larger than the largest 32-bit number wraps around and rests. The writes wrap around at the 128KB transfer size and the reads wrap at 2048KB ..."

 

http://www.tweaktown.com/reviews/5629/samsung-840-evo-250gb-ssd-review/index4.html

 

 

 

... Note that AS-SSD uses uncompressible data, so it generally measures smaller numbers than ATTO does for drives that internally compress data before writing it.  With drives from 2 or 3 years ago (the age of mine) that matters more than it does with modern drives.  ATTO uses compressible data, which yields better write times and has in my experience been more representative of real performance ...

 

Problem being that many real world files (music, photos, video, etc.) have a nasty habit of being incompressible  :)  . If that defeats controllers like the early Sandforces, I prefer the benchmark showing it (even if that ruins our bragging rights :D ).

Edited by TELVM
Link to comment
Share on other sites

OK, I see what you're talking about.  I suggest that perhaps ATTO is "blowing its gasket" above 4 gigabytes / second, not 1.

 

4 billion and change just happens to be the biggest number one can put in a 32 bit word.  Coincidence?  I think not.  32 bit is SOOOO last century.  :-)

 

What kind of I/O subsystem do you have that marches right up and passes 4 GB/sec?  I like.

 

-Noel

Link to comment
Share on other sites

I tried PC Mark 7 this days (huge but not as much as PC Mark 8). I monitored its benchmark with PC Wizard and the use of the processor was never maximized. I did it two times, one with EIST enabled and the processor running at 1.2 GHz and one with EIST disabled and the processor running at 3.2 GHz. I have done the same in my old laptop with a Pentium M 740 processor. The results:

 

Pentium M 740 - PC Mark 05

EIST enabled      EIST disabled

800 MHz             1733 MHz

1844 Marks         2942 Marks

 

Intel Core i5-3230M - PC Mark 7

EIST enabled      EIST disabled

1200 MHz           3200MHz (with turbo boost)

1612 Marks         1599 Marks!

 

So the processor is indifferent to this thing! How do they sell it?

 

Passmark Performance Test 8 seems good but it is not free, I am staying with Novabench and Crystalmark (thought a bit outdated - hopefully a new version will be released at some time) for the moment.

Edited by HarryTri
Link to comment
Share on other sites

Also there are toys nowadays able to deliver over 4GB/s sequential straightforwardly  :ph34r:  :

 

Areca 1883

 

Do you use an Areca controller?  With what OS?  Do you find it stable?  The controller I have is a mid-level one with decent performance but no fancy cache features.  Not that I find I wait for a whole lot, but faster is always better.

 

Also, can you get to the raw SMART registers of the individual drives?  I can only get to the standardized/normalized set, which isn't bad but I can't read things like total data written/read, etc.

 

-Noel

Link to comment
Share on other sites

Passmark Performance Test 8 seems good but it is not free

 

I honestly think Passmark is worth paying for.  It's cool for two reasons: 

 

1.  It's a decent way to monitor a system's performance over time, primarily because of its database features and screens on which you can compare past results with current.  Assuming you leave yourself notes, you can see when certain things got slower or faster and you can easily watch for trends.

 

2.  Their web integration makes it a VERY handy tool for doing research on what new systems are actually delivering re: performance.  You kind of have to ignore the very top group of results, because they're either rabid rich folks looking to top the charts, or (more likely) just plain cheats by people looking to top the charts.  But you can easily do things with it like ask for results from a particular kind or family of process or system, e.g., "E5-2687" or "Dell Precision", if you're looking to buy into something new and want to know ahead of time what it actually delivers.  Enough different people use it that those results are very useful.  It also tells you whether your setup is performing up to the levels set by others with substantially the same setup.

 

Back a few years when they came out with version 7 I decided to buy it so I could use it long-term and I am glad I did.  I paid for the upgrade to version 8 when they came out with it for use in much the same way.  8 happened to come out right when I got a new workstation, so I didn't have to worry about whether I was comparing version 7 results to version 8 results.  I've been happy with it.

 

Thanks for all the feedback, guys.  Keep it coming!

 

-Noel

Link to comment
Share on other sites

Do you use an Areca controller? ...

 

No no that pic isn't mine  :)  . I wouldn't mind having an Areca but don't really need it.

 

 

 

... So the processor is indifferent to this thing! How do they sell it? ... (PC Mark)

 

Then you're gonna love this  :w00t:  :

 

"... My my. Swap CentaurHauls for AuthenticAMD, and Nano's performance magically jumps about 10 percent. Swap for GenuineIntel, and memory performance goes up no less than 47.4 percent ..."

 

http://arstechnica.com/gadgets/2008/07/atom-nano-review/6/

Link to comment
Share on other sites

 

e81b34b1_areca1883ias-ssd-benchAreca840P

 

 

By the way, I noticed you have a 1.7 version of AS-SSD, so I went and downloaded the new version.

 

I'm not sure we should be impressed by a benchmark that delivers better results when you run the latest version (compare these numbers to those I got before with 1.6).  There's a certain feel good factor in that but I'm not sure it's useful.  Or maybe it's just inconsistency - I've seen inconsistency from AS-SSD before.

 

AS-SSD_05_30_2014.png

 

-Noel

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...