SSD below advertised speed

[elivance]

I just got a Crucial RealSSD 128GB a week ago. HD tune shows it hovering very near 300 MB/s which is significantly less than the expected 355 MB/s. Its very fast, but I obviously want to get the most out of it.

Both the motherboard and cable are SATA III, so I do not believe this is the issue, and AHCI is enabled. Does anyone have any other suggestions? Any tips would be greatly appreciated.

Motherboard: ASUS M4A89TD AM3 AMD 890FX

SSD: Crucial RealSSD 128GB

 

[dEAne]

Their is a firmware update for this:
http://www.crucial.com/eu/support/firmware.aspx?ref=dynamitedata.com
Please read the pdf first.

 

[lafontma]

Crucial does not mention HOW they tested..

So most probable is that on some test, for a brief moment, the drive actually hit that number of 355MB/s so they can use that number..

 

[RetiredChief]

See if you can find some independent Reviews. sucvh as:
http://www.anandtech.com/show/3812/the-ssd-diaries-crucials-realssd-c300
Look at System config and see if you can find any performed on the AMD platform.
Reason I mentioned this is that if appears that The Pheonix Pro (SF1200 controller) p-erforms slightly better on the Intel chipset.

 

[elivance]

Thanks for the replies!

Unfortunately I'll have to erase it to upgrade, which isn't really a big deal since I only just installed anyway... however this concerns me:

"Note: The upgrade process will not work with drop-in HBAs (Host Bus Adapters) or 6.0 Gb/s ports.

A native 3.0 Gb/s SATA port must be utilized."

All the SATA ports on my motherboard are natively 6.0 GB/s. If I set the motherboard to 3.0 GB/s for those ports, will it work?

 

[RetiredChief]

Don't sweat the Sequencial reads as they are the least important parameter. Look at the small file Read/writes (ie 4 K random) and the IOPS. If these are close to Manuf specs then You are good. Quite often the Seq reads will be less than manf specs as these are normally under ideal conditions. Also bear in mind, you do not run defrag on an SSD, so even larger files may not see the boost of fast Seq reads as they may not be contigous.

It is the Random 4 K speeds that are important to boot times and program loads. The Sequential read/writes normally only boost applications that read/write LARGE chunks of data such as video files (1 Gig -> 30+ gig files) and working with Large Jpeg/Bitmap picture files. These types of files are normally placed on a HDD as opposed to being on the SSD anyway.

 

[sub mesa]

Try CrystalDiskMark, it tests on the filesystem with read-ahead instead. It may give you higher (real) scores than the HDTune test. AS SSD is also useful, as it tells you whether your partition is aligned ("GOOD") and what I/O driver you use (microsoft ahci or chipset ati/amd driver which DOES NOT GIVE YOU TRIM SUPPORT!).

So check whether you are using the Microsoft AHCI driver (msahci.sys) instead.

 

[RetiredChief]

Sub mesa
Trim is now supported using the latest Intel AHCI driver (Ver 9) Trim is also now passed if the SSD is on Raid, but not a member of a Array (Raid0 or 1) using the Intel RST drivers, ver 9. Trim is not supported using ver 8 of the Intel RST driver. The Pheonix Pro (probably vertex-2) with the SF-1200 Controller performs rather poorly using the default uSoft AHCI driver. Had to reload win 7 on My Pheonix pro as I had orignally set Bios to AHCI and used the default AHCI driver. Reloaded windows 7 using F6 and the Intel AHCI driver (recommended method).
http://www.intel.com/support/chipsets/imsm/sb/CS-031491.htm

Added. I think AMD setups are currently stuck using the uSoft AHCI driver.

 

[sub mesa]

But the OP has an AMD chipset, and thus he can only use Microsoft AHCI driver or the ATi/AMD chipset driver. Of those two, only the Microsoft driver passes along TRIM.

If you're on Intel platform instead then you're right of course. The Intel passes along TRIM in AHCI/IDE and RAID mode to SSDs just fine, as long as they are not in a RAID themselves. Essentially RAID mode is just an extension of AHCI, but it cannot be handled by the Microsoft drivers, and thus, until Intel modified their drivers to include TRIM in AHCI mode, there was no way to get TRIM support on the Intel controller running in RAID mode; but even in AHCI mode many people did not receive TRIM because they installed chipset drivers including the Intel driver which replaced the Microsoft driver.

Now at least these problems are gone on Intel platform with new 9.6+ drivers, but on other platforms like VIA, AMD, nVidia and add-on cards like Silicon Image (including SSDs like RevoDrive) you would have the same problem: the proprietary drivers don't support TRIM, and thus the Microsoft AHCI/IDE driver should be used instead to avoid degraded performance due to lack of TRIM.

On Linux/BSD the "RAID" hardware just registers as SATA controller, and thus TRIM works on all normal SATA controllers, FakeRAID or not. So it's purely a software limitation of the vendor-specific Windows I/O drivers. Kind of a shame reputable companies like AMD don't have TRIM support by now; they do have 6Gbps but no TRIM in their drivers. Many people may not realise this and run with AMD driver instead of Microsoft driver. All these issues with TRIM are rather messy, IMO!

 

[RetiredChief]

^+1, Should have gone back up to OPs first post.

 

[elivance]

So I shouldn't mess with the drivers? I didn't do anything to them, so whatever got loaded from my mobo drivers and from win7 are the drivers that are running. I am having trouble telling which ones to look for. In the device manager, these are the relevant ones I could find:

Under "Disk drives", the C300's drivers are: "disk.sys" and "partmgr.sys" both by Microsoft.
Under "IDE ATA/", the AMD SATA Controller's drivers are "amdsata.sys" and "amdxata.sys".

Does this mean I'm running the AMD driver?


About the firmware update:
I realize sequential reads are not the most important, but I would like the most performance I can get out of this drive. Additionally, this is what crucial says about the update:

"Release Date: 5/20/2010
Change Log:
Improved Power Consumption
Improved TRIM performance
Enabled the Drive Activity Pin (Pin 11)
Improved Robustness due to unexpected power loss
Improved data management to reduce maximum write latency
Improved Performance of SSD as it fills up with data
Improved Data Integrity
Note: This requires a Low Level Format to the SSD which will erase any data on the drive.
Please ensure that your data is backed up prior to performing the Firmware Update. We are hopeful that future Firmware revisions/updates will not be destructive."

This is a pretty important update that I definitely want to get, regardless if it changes my speed. Do you think the update will work in my SATA III ports if I set them on my bios to 3.0 GB/s?

 

[sub mesa]

That looks like you're using the AMD driver indeed; you can check with AS SSD.

You can use the AMD drivers, for your CPU and other things, like onboard video, discrete video, hdmi-audio driver, etc. But the AMD I/O storage/RAID drivers are optional, and you don't want them as they don't give you TRIM.

I'm not sure whether this works, but you could check Device Manager and try rolling back the driver to an earlier version; by any luck it will revert to Microsoft AHCI driver again and you have TRIM support again. Note that some performance loss may not be restored immediately; due to TRIM not working for some time. TRIM does not work backwards; only when you (or the system) deletes files when TRIM is working.

The most important thing is that you regain TRIM support now; try reverting back to Microsoft AHCI driver. Benchmark results may improve over time as the 'not TRIMed' areas in the past get overwritten and TRIMed again; so in essence it may take some time for the SSD to regain all usable space for itself. That's no real problem though, it will solve itself over time.

 

[elivance]

Rolling it back worked, now I have the Microsoft AHCI driver. Thanks for that, I wasn't aware that TRIM wouldn't work on the AMD driver.

EDIT: I tried AS SSD.

It looks like I have firmware 0002, so that is great. The sequential numbers are looking much better too. However, the 4k random is now concerning me. Only 27.27 MB/s and 57.59MB/s - or ~7000 and ~15000 IOPS? Why is write higher than read? Both are very low. The specs for this drive are "Random 4K Read/Write: 60K / 30K IOPS"... Any ideas on improving this?

 

[elivance]

Best answer selected by elivance.

 

[elivance]

Thanks for all the helpful replies!

If anyone has any advice on the 4K random read/writes, I'd really appreciate it. Am I misunderstanding my numbers? Could something else be affecting that?

 

[sub mesa]

Low random read 4K is normal, 20MB/s is what you can do; this is what a single channel does. This is comparable of running a single threaded program on a multicore CPU; only one will be used. No way to increase this by RAID0, either in the SSD internally (multiple channels) or when using onboard/hardware RAID0.

The 4K writes shouldn't have this problem, but these are very good scores you should be happy about them! The writes may be higher as you overwrite stuff that wasn't TRIMed previously. There are tools to force this manually, but not really needed the problem will fix itself over time. Your scores are already very good i'd say!

Oh and the rated IOps are at queue depth 32 or 64; not 1. So look at 4K-64 score to compare to specified IOps.

 

[elivance]

Low random read 4K is normal, 20MB/s is what you can do; this is what a single channel does. This is comparable of running a single threaded program on a multicore CPU; only one will be used. No way to increase this by RAID0, either in the SSD internally (multiple channels) or when using onboard/hardware RAID0.

The 4K writes shouldn't have this problem, but these are very good scores you should be happy about them! The writes may be higher as you overwrite stuff that wasn't TRIMed previously. There are tools to force this manually, but not really needed the problem will fix itself over time. Your scores are already very good i'd say!

Oh and the rated IOps are at queue depth 32 or 64; not 1. So look at 4K-64 score to compare to specified IOps.

Sweet, thanks a lot! I was just confused with how the information was given - I saw that the 4k-64k numbers looked great, but the specs listed were random 4k. I didn't really know what to compare it too. I'm extremely happy with my performance in terms of how fast it feels - everything starts up instantly. I've heard people complain about loading times in Fallout: New Vegas - mine are ~2 seconds.

Thanks again for clearing all this up, and the advice from before!

 

[rwpritchett]

I think AMD setups are currently stuck using the uSoft AHCI driver.

You can use the AMD drivers, for your CPU and other things, like onboard video, discrete video, hdmi-audio driver, etc. But the AMD I/O storage/RAID drivers are optional, and you don't want them as they don't give you TRIM.

SM and RC,

Have either of you SSD guru's confirmed whether or not AMD has finally allowed TRIM commands to pass through in their latest AHCI drivers? There's a lot of buzz about it being added in 10.9.

Here are some threads I've been following on the topic:

http://www.ocztechnologyforum.com/forum/showthread.php?80581-AMD-AHCI-10.9-and-Trim

http://forums.amd.com/game/messageview.cfm?catid=261&threadid=142003&highlight_key=y&keyword1=trim

http://www.rage3d.com/index.php?cat=75#newsid33970627

Any thoughts on the subject?

 

[sub mesa]

Could actually be that AMD implemented TRIM, but this can mean two things:
- TRIM only on non-RAIDed SSDs, while controller is running in RAID mode (like Intel supports)
- full TRIM support on both non-RAID and RAID configs with the AMD drivers.

If AMD actually implements the latter option, then i would be hugely surprised. More likely AMD implemented the first option, making them on-par with the Intel drivers as far as TRIM support is concerned.

Would be good news, but i've yet to see something official and directly from AMD. But it looks plausible.