Anvil Storage Utilities

Has anyone here used the Anvil Storage Utilities to bench their HDD/SSD's with?

See:
Anvil's Storage Utilities


I find it interesting because it shows all relevant information on a single screen, automatically. Unlike other benchmarking programs that just report the 'score' and leave you guessing as to what equipment was used to produce it.


This is also the software that is used in the write endurance testing here:

See:
SSD Write Endurance 25nm Vs 34nm



What I use this software for is for evaluating different HDD's (partitioned) for the fastest possible, overall, O/S and program use.

This is an easy way to see that a Seagate XT Hybrid 500GB partitioned to 100GB is not as fast as a the first 100GB partition of a 3.5" 1TB WD Scorpio Black (raw platter performance, not the SLC nand).

However, in actual O/S and program use, the XT Hybrid is faster when used with the following method to move the users folder to another drive (benchmarks only guide, real-world testing still required. ):

See:
http://forum.notebookreview.com/win...gramdata-folder-separate-drive-partition.html

because after the O/S and programs are installed, it is mostly used in a read-only scenario - and that is where the SLC nand shines in speeding up the response of the drive. Compared to running this system from the Scorpio Black with the same moving of the users folder (to the XT Hybrid as the 'D' drive), the XT seems at least twice as responsive if not more. But still not as responsive as an SSD with the same 'users' folder 'tweak' done to it.

If I had been comparing normal HDD's to each other (I did, but did not keep the screenshots), the information shown by Anvil Storage Utilities would have been enough to rely on to pick one drive over another (rather than simply testing each and deducing which was faster, overall).

If I had been comparing SSD's to each other, I still would be testing 'real-world' to see what was actually faster - but I'm guessing that the 'response time' of the reads and writes is what I would be using to guide me if I had to choose one drive or the other.

I would be very interested in seeing others with their HDD/SSD scores for comparison with this useful software.

See attached screenshots:

 

And this is on my U30Jc notebook and an Intel 320 160GB SSD:

The comparison here is that I tested while on the high performance battery profile with ThrottleStop 3.20 beta1 with and without the settings as shown in the post below.

See:
http://forum.notebookreview.com/7755405-post728.html


Ignore the notes (in the screen shots) that say High Performance Battery mode - these were tested when plugged in (meant to say high performance profile...).


What is interesting here is that the desktop is much more responsive with the XT Hybrid than the notebook is with the SSD. Regardless of what the benchmarks might indicate.

However, the notebook storage subsystem can do more 'work' faster (with ThrottleStop running), even if the notebook system is only an i3 350 @ 2.3GHz vs. the Q9450 @ 3.4GHz. ('Work' here means Windows and program updates...).

Both systems 8GB RAM, both running Win7x64 SP1 and both with the 'move users folder' tweak. Desktop should have the advantage as it has 2 physical drives installed and the 'D:' drive is the 3.5" Scorpio Black 1TB drive.

I also tested the ThrottleStop tweak with the desktop (above) HDD's, but there was minimal difference except for about 35% faster Random Read 4K at a queue depth of 4. (1.05 vs. .77 MB/s 4K QD4 'scores').

 

And this is from a SNB workstation with an Intel 510 Series 250GB SSD (only 100GB allocated) and a vRaptor 300GB 10,000 RPM drive as the 'D:' drive (using the same 'move users folder tweak' above. System running at 4.5GHz on 'air'.

I haven't tested the ThrottleStop tweak on this one yet, but I still might in the not too distant future.


Note here that while the 10K RPM vRaptor is posting lower 'scores' than the Scorpio Black is at 7.2K RPM, it is doing so over the whole platter area - and is not partitioned as the Scorpio Black is (at 100GB).

Yeah - God bless these vRaptors - true workhorses.

The 510 doesn't do too shabby either.

Again, the response time tells the story between the 'feel' of the Intel 320 Series vs. the Intel 510 series. What!?! But they're not in the same system? I know, I know... but the 'feel' is something that a system has or doesn't.

The feel of the SNB platform with the Intel 510 is noticeably snappier than the XT Hybrid on the O/C'd Q9450 system. The 'work' that the SNB platform can do though is in a league of it's own (stay tuned...).

Final note: all systems in the three posts above running Intel RST 10.6.

 

Here's my old Latitude D830 which has a Core 2 Duo T8100 and an X25-M 80GB hard drive.



Three of the SSD Read tests are competitive with Intel's newer technology while in the 4K, 32K and 128K tests, the old technology really smokes the Core i5. The newer chipsets and CPUs are more efficient when they idle down and turn parts of themselves off but this seems to hurt the hard drive benchmark scores.

The X25-M 80GB drive is limited to Write speeds of only 80 MB/s so in the Write tests, the new drive kicks it's butt. It would be interesting to see your modern drive running on the old Core 2 Duo and old chipset for a fair comparison.

Thanks for the link to this new tool from Anvil. It looks like it is going to be very popular for hard drive testing.

 

Thank you unclewebb, for ThrottleStop and for the benchmarks too.

Here is my old P8400 Core 2 Duo platform running the same drive (only 0.16GHz faster than yours), an Intel 320 Series 160GB SSD. With ThrottleStop 3.20 Beta1 and without ThrottleStop running.

And, although the storage subsystem is 'obviously' superior on the older platform - the work the i3 is capable of (same clock speed, btw) puts the older platform to shame in a productivity setting like photo editing (also cooler, better battery life and quieter too).

The 'feel' between the i3 vs. the P8400 with the same 8GB RAM, same clock speed and identical installation (except for system specific drivers) definitely goes to the i3 - in spite of the P8400's system superior scores in the ASU benchmarks as shown here.

Anyway, here are the scores you asked for:

 

Here are the results with the Intel 510 Series 250GB SSD and ThrottleStop 3.20 Beta1 on and off.

Actually, this is a little bit different because the results on the SNB platform O/C'd to 4.5GHz did not change significantly with or without ThrottleStop.

What I did is added a load to the system (as this is what would be 'normal' anyways) - by running SuperPI to 32MB - at the same time as I was running the ASU program. With and without ThrottleStop running.

The differences are minute (0.4% less write speeds with SuperPI and ThrottleStop but 4% faster read speeds, overall), but I wanted this testing to be complete (and in one place).

While the overall score is only ~2% faster (with a load and ThrottleStop running), it is important to note that depending on how a system is used, the overall score may not be as important as the overall read or write scores (or even, specific subscores in each).

4% faster reads doesn't seem like a lot (and it really isn't, on an older platform...), but it is something to keep in mind when I'll be using the system for actual photo editing work to compare to in a real world with/without (ThrottleStop) scenario. In simply using the computer for around 1/2 an hour with TS running, I did feel the difference (even in browsing) when I turned TS off (yeah, sensitive to the little things...). I'm wondering if I'll be able to feel this small difference when actually editing images at full tilt? (More testing ahead. ).


unclewebb, what I was curious about is that the SNB platform with ThrottleStop 3.20 Beta1 would not let me disable/choose any C-States for the package? Is this not controllable with ThrottleStop on SNB? Or is this because it is a desktop platform (Z68 chipset + 2600K)?

The scores:

 

My Crucial M4 256GB for SATA II

 

Honzik1,

Thanks for the screensaves.

With a SF based drive, could you (please!) redo the scores with the Compression set to 67% (assuming you have data on the drive)?

Thanks.

 

Intel 310 80gb, same controller as X25M G2.

Turning off CPU power mgmt in the BIOS makes a big difference in ssd speed, but it's not practical since it kills battery life.

Forgot to add notes:

Left is with cpu pwr mgmt On, Right is with it Off.

 

vinuneuro,

Nice improvement, but not so nice to battery runtime, huh? (How much does it cripple it)?

You may want to see if ThrottleStop 3.20 Beta1 set as show below gives the same kind of performance with less drastic battery life loss.

See:
http://forum.notebookreview.com/7755405-post728.html


What notebook do you have that you're able to turn off power management in the BIOS?

 

Nice to see that you have downloaded my app

Will try to follow your findings.

Here's my Latitude E6400 using the Vertex 2 60GB.





edit:

Will be posting some from my Lenovo as well.
Running one SF drive and currently an Intel 320 in the UltraBay.
(the current signature listings aren't updated, last time updated was when I got the W510)

 

Anvil,

Thank you very much for your new and immensely useful program!

I see from the screenshots that Beta5 is available ... downloading now...

(Actually, can't find Beta5 yet... only B4 is available for download right now... ).


Great work on the other forum with the SSD Write Endurance testing!

Could I ask where Vapor got a SF product with no LTT enabled?

See:
SSD Write Endurance 25nm Vs 34nm - Page 47


Are these for sale? Anywhere?

Sorry, just a little excited about the possibility of a SF drive running without silly restrictions.

Thanks for your benchmarks and waiting to see ones from your Lenovo as well.

 

Thanks!

I'll make Beta5 available later tonight, not much left except for testing and a few small cosmetic issues that was reported earlier today.

One can't order the normal SF based drives w/o LTT, yet.
I agree that it's a bit silly, it is more of a nuisance than a practical limit though.

 

Silly? No, more like downright stupid (should be in all caps!).

See:
http://forum.notebookreview.com/solid-state-drives-ssds-flash-storage/509106-slow-ssds.html


Beta5, okay - later.

Did you see this (speaking of cosmetic issues):

See:
http://forum.notebookreview.com/att...l-storage-utilities-_gb_1gb-20110730-1950.png


The bottom righthand side is not displaying properly on a P8400 Core 2 Duo notebook. (Just thought I'd mention it...).

More importantly - it is showing as 0KB alignment - this is wrong (default Win7 clean install - just like all my other systems).

Hope this helps you to fix these little bugs a bit quicker.

 

Anvil,

Playing around with the software, I noticed that it can display the minimum and maximum access times.

I would love to see this info displayed on the main benchmark page! This is exactly the reason why I think I felt all (pre-gen2...) SSD's were slow.

Although this screen shot is from a mechanical HDD, it clearly shows that the avg. access time can be, uh... misleading by a factor of 100+:

Min. acc.: 1.98762ms
Max. acc.: 2076.88441ms

Avg. acc.: 18.453ms


Two order's of magnitude difference is something I want to know about!

I'll try to post the SSD's scores with this same metric (1GB Test File with 1K @ 8 QD) by tomorrow - curious if they have such a magnitude difference too.

I would also like to customize the main benchmark's default tests too.

Something along these lines:

From:
The SSD Manufacturers Bluff - The SSD Review


Hope I'm not bombarding you too much with my requests?


See:

 

Hi

No, not at all.

The other block sizes will be covered in other tests, there will be a few more tests aimed at different "audiences".

The Top 5 is not a good example though, 80% of all I/O on desktop computers are reads.
Looking at that list it looks to be the other way around
I'll dig up some references for our analysis.

You can display the min/max access time by hovering the mouse over the Resp. time results.
Most result fields have "Hints", so, do some hovering. (requires that a test has been run)
I'll try to make the min/max access times as a part of the visible results.

All tests are stored in a database and you will be able to export the results to Excel. I'll make sure that the min/max access times are stored as well.

The new beta was uploaded last night but it has yet to be made available, I'll make a post when it's confirmed that it is available.

If the issues you pointed out are still present I'd like to know a bit more about you setup.

 

I read about this program on Les's site in which he loved it. I finally got the time to run it on my Samsung 470. This is SATA 1.5.
I would love to hear your comments

 

What do you think of this.
Interestingly there is only little difference between power profile balanced and high performance.

 

It looks fine, considering that it runs at 1.5Gb/s.
It's a great drive, I'll post a test of mine shortly.

 

Nice, not very often to see raid-0 on laptops

I've almost given up on getting performance like in the good old days, my Latitude E6400 is on par with the desktop chipsets. (from that same age)

So my W510 is unmodded, no tweaks at all, it's still a "mean"-machine for work related stuff. It runs 3-4 VM's easily.

 

Lenovo W510

No tweaks and running the "Power Source Optimized" profile.

First off the boot drive, a Vertex 2 240GB, I used 0-Fill so the result is optimistic.


My standard SSD for VM's, the Intel 320.


And finally a Samsung 470 256GB, which I like a lot.


I expect there is quite a lot to gain in benchmarks by doing the standard tweaks. (I used to tweak but not currently)

 

@Anvil
Why don't you use the Intel Rapid Storage AHCI driver?
In my opinion, in real life it performs better than the Microsoft driver.

 

I do use IRST but due to testing I sometimes have to run other drivers.
10.6.0.1022 is the one I'm currently using.

 

Just a little update with Anvil Storage Utilities Beta5:

I was just running the Threaded Mixed IO benchmark @ 8 QD and see that the at the 256K size, the maximum access time can hit almost a quarter of a second (~250ms). This is on the Intel 320 on my U30Jc with ThrottleStop 3.2 Beta2 running also.

I knew I could feel this...

While not technically a 'stutter' - it sure feels like that to me in certain scenarios.

Do other SSD's/Controllers exhibit similarly high maximum access times?

Curious to see if I'll be upgrading or not.

 

My results with a 256GB C300:



This seems to be a pretty comprehensive benchmark.

 

Thanks for the replys guys, sorry it's taken me awhile to get back here.

Actually my laptop is not that old even though it is SATA 1.5. It is a Panasonic Toughbook, in which Panasonic throttles the CPU back due to the laptop being fully enclosed, no fans. I'm not sure how they do it, but I can't tell by "hand's on" the difference. To me that is what counts, the hand's on.

This Samsung 470 is without a doubt the most stable, consistant SSD out of all my SSD's. I have 5 of them right now.

 

Here's a little tidbit:

See:
FancyCache Overview

Running FancyCache with 2GB RAM used as a R/W cache with the write delay set to 30 seconds, the lowly XT Hybrid (see first few posts for benchmarks...) becomes the new king of the hill.

What is interesting is that the computer feels just a little faster/more responsive; not the 102x increase the 'score' might have us believe over the normal 4GB SSD Hybrid enhanced Seagate's partitioned 'default' score.

I think I will be using this beta software to show clients how an SSD might improve their computing experience - it is definitely an eye opener and clearly shows that any improvements past even a Hybrid SSD/HDD setup needs to be orders of magnitude to show up in 'normal' day to day computer usage patterns.

And like most caches, remember that the first time a program is run, it will run at the slow (i.e. 'normal') speed you're used to - there is no improvement until you need to run it a second time (or more).

What is interesting though is that it is obviously increasing the responsiveness of this forum! (Noticed during posting/previewing...).

 

You might try increasing the test size, should still result in gains.

 

Well, judging by the 'feel' of the system I would say no - so what I did is push this expectation the most I can:

By setting FancyCache to 1GB RAM and the same 30 seconds write delay (deferred writes) and also testing with a 4GB test size I forced the benchmark to show what happens (translated: what I 'feel') when the cache is exhausted or not filled yet with the program/operation I am trying to initiate.

Although it is still (barely...) above the non-FancyCached benchmark run (244.47 vs. 281) it still feels slower than the non-cached bare drive and looking at the Read/Write 'scores' we see why: the Writes with FancyCache are what is boosted (~50% higher), but the reads suffer by about 14% and thus my 'it feels slower' impressions - when the cache is not used optimally.

I was wrong to think the benchmark would be slower (overall), but I am right to see that the benchmark backs up my 'feel' of the system too (in the reads).

Setting FancyCache to 4GB and ASU to a 4GB test size does show a good increase still, just as you suspected Anvil. What this run also shows though is how far a drop the sequential 4K Reads drop when the workload (test size) is equal to the cache size (RAM used).

One way this manifested itself is how long it took for Windows to draw the window controls to save the screen shot right after this run - almost 5 seconds! And this is with over a GB of RAM free too - so it couldn't have been a shortage of RAM.

Hmmm... I guess FancyCache can also be used to make HDD's stutter too.

And, it gives me greater appreciation for the optimizations (and tradeoffs) that MS has done with SuperFetch and other memory tweaking enhancements in Win7x64.

First (left) thumbnail is 1GB FancyCache with 4GB test file in ASU.

Second (right) thumbnail is 4GB FancyCache with 4GB test file in ASU.

 

I really like this utility, where can I find Anvil's homepage or is there not one yet cause the utility is still in Beta?

 

First page, first post, first link.

 

Not Anvil's homepage but whatever.

 

So, it's not good enough for you?

I'm sure that is all he has right now.

 

Hey tiller,

Here are my ASU results for the 7k750

 

Nice increase in the write's. Faster than the XT Hybrid.

I notice how the 3.5" desktop (first post) Scorpio Black 1TB is still faster though (I guess without nand complementing a 2.5" HDD such as the XT Hybrid, good 3.5" models will always be faster).

Thanks for the update.

 

Yeah, I'm very pleased with it. More room for games!

 

I just got a couple of OCZ 60GB AGT3's (bad choice looking at posts on this forum ) but looks like I'm not going to get much chance to play with them for another month or two (going away) but here's a quick test with C-State enabled/disabled FWIW. RAID0

No C-States




C-States enabled


What are your thoughts?

 

Looks very impressive.

Curious what version of Intel RST you're using? (v11?) I can only find 10.6 on the Intel site.

Does the system feel as fast as this would indicate it should be?

 

It's an IRST alpha version, so a little experimental on my part.

Can not give an opinion to speed as this is my first SSD, so a learning curve for me there, and due to poor timing looks like I will be going overseas before installing the OS on it as I haven't decided exactly how I want it setup yet.

 

Thanks Dufus.

Hope your overseas trip is enjoyable and looking forward to your impressions on this setup when you return.

Safe trip!

 

Intel Rapid Storage Technology 11.0.0.1015 Alpha Version

can be downloaded at

Intel Rapid Storage Technology 11.0.0.1015 Alpha Version - PCTreiber.Net

as already said, it's alpha and not WHQL.

It installed without any issues on my HP 8530w.

 

Thanks for the link.

Can't understand that language... does it mention any specific benefits (seems like a major version change)?

 

maybe the discussion at this forum can help to understand.

Technical Discussion Intel 11 Series Orom

The package contains also the ReleaseNotes with the changes.

 

Always curious about things like this. I ran the test on my Intel X25-m three times. Once with my current RST, then with 11 alpha, then back to RST 10.6.0.1002

I found that some numbers improved drastically, while some stayed the same. Also, a few were significantly negatively impacted. 4k, for example, was super low on RST 11, for my system.

 

I don't think the scores are dramatically different.
With every run of the Anvil Bench my scores are also different.
Of interest for me is, that Anvil Bench wasn't able to determine the ntfs cluster size and the driver version.
And Dufus post 37 shows iaStorV as driver, my screenie shows iaStor!
Can someone explain?

 

For some strange reason it fails to read some of the WMI stuff on Anseio's setup.

@anseio
What is your exact setup and have you removed any services or performed any particular tweaks. (any error msg's would be interesting as well)

@maximinimaus
I'm not sure why there is a different wording on the iaStor driver, could be chipset/generation related as I've got both iaStor/iaStorV. (on different computers of course)

 

I've got a few non-essential processes disabled, like tablet pc and a few other things.

edit: my hardware specs are in my sig. I've also done JJB's tweaks for SSD's. Those apply to CPU states though. Any more specific information you'd like from me? Otherwise, I'm not sure what to look for.

I do get this error upon launching Anvil:

 

Very unlikely but thanks all the same


Seems like the driver installation wasn't 100% as the Windows driver wasn't updated. Had to do that with the device manager.



All the posted screen shots were made using DDR3 set to 1333 and CPU at 4GHz. Overclocking does increase the benchmark a few %.

I also have a bit of SW I wrote that creates a block of data in RAM and writes it to disk multiple times until a 1GiB file size is reached. Results there show it took 843ms which is about 1.18GiB/s (1.27GB/s). Idk how much of this is cache effect though.

Disabling C-States makes about ~8% difference in the benchmark, maybe some exit latency from deeper C-States having an effect?

Running with a lower BIOS option ROM (10.5) doesn't seem to make much difference in performance with 11.0 drivers, maybe the newer ROM is primarily for fixes of errata.

Don't know yet if running RAID0 to speed things up will make a big difference or not for real world so might split the drives for use on 2 different systems. Not sure what to make of it all, so many unanswered questions and no time for now to answer them.

@maximinimaus, with the updated windows driver it's now iaStor without the V.

Edit:
Using Compression 46% setting


67% gives 3225 with r/w ~1600 ??? QD4/16 write IOPS drop to ~44000

Compression 100%

 

iaStorV is the Raid driver contained in Windows 7. It will show when the Intel Raid packages are not installed.

Intel Matrix Storage Manager and Intel Rapid Storage Technology contain iaStor.

@Dufus, as you had to reinstall the Alpha version, now it's clear for me about iaStor(V)

 

Thanks!

Could you post a screenshot or PM me a copy of the following info

Select "System Info->WMI" from the main menu (in my app) and click on DiskDrive.
(could take a few seconds and there is a lot of info so you might need to maximize the window and/or scroll a bit to get to all the info)

I'll try to recreate the issue and will post a special build for testing this issue if we can narrow it down a bit.