SSD: Poor Boot Times and Low WEI

[idata]

I just got an Intel X-25V and I installed Windows 7 Ultimate on it. It's currently taking up about 8GB. I've hardly installed any software other than the drivers and firefox.

When I ran that Windows performance test, my score came out to be 5.9, which is due to the SSD. I don't remember the numbers exactly, but I'll try to remember them:

• RAM random access memory - 7.5

• CPU central processing unit - 7.5

• Hard disk - 5.9

• General graphics performance on the desktop 7.4

• 3D graphics capability - 7.4

I haven't run any benchmarks. I have done everything on this page except for RAMDisk. I tried that, but it caused too many problems so I installed it. Even though, I got the info from OCZ forum, it still applies to SSDs:

http://www.ocztechnologyforum.com/forum/showthread.php?63273-*-Windows-7-Ultimate-Tweaks-amp-Utiliti...*

So, I:

installed the latest firmware

bios and OS set to AHCI

SSD is connected through SATA Port 1

and lots of other tweaks that are found on the page above

I used this software to determine my restart/boot time:

http://www.sevenforums.com/tutorials/720-restart-time.html

When I used it, the result was 48 seconds. That number represents the restart time, which includes shutting down and starting up. It took my computer 7 seconds to shutdown, which means it took about 41 seconds, give or take, to boot up. I read about other people who said that it only takes their computer 12 seconds to boot up. Others have said about 17, and a few even said 6 seconds. One of the main purposes for me getting the SSD was so that I'd be able to boot up my computer within 15 seconds, like everyone else with SSDs.

As for the WEI, I thought that was low because the Intel's SSD X-25V had low read and write speeds. But others with the same SSD have reported getting scores around 7.7 and 7.8.

Someone mentioned changing the driver to RST (post # 9 and # 10 on this page):

http://communities.intel.com/thread/11286?tstart=0

So, do you people have any suggestions as to how I can improve my SSD to be like yours, faster, speedier, more impressive, and hopefully, so I can get my money's worth. So far, I fee like I got a "slightly" faster hard drive.

This is my current setup:

OS: Windows 7 Ultimate CPU: Intel Core i7 860 MB: MSI P55 GD80 RAM: Gskill 4 GB

PSU: Corsair 650TX

Case: Antec Sonata Elite

Video: XFX 5770 Sound: integrated HD: Samsung F2 500GB (storage) Samsung F3 1 TB (storage) Intel SSD X-25V (OS installed on SSD)

Update: This one applies to Windows Vista, but he says that when he disabled his 7200 RPM drives and only used his SSD, that cut his boot time by 10 seconds. Does the same apply to Windows 7?

http://social.answers.microsoft.com/Forums/en-US/vistaperformance/thread/ea530723-c3fe-4817-9de1-68c...

Update 2: Well, by upgrading the RST driver, I managed to increase the WEI of the hard drive (SSD) from 5.9 to 7.7. Now my WEI score is 7.4. But I'm still not satisfied with the boot time.

[idata]

That was most generous of you sir, Thanks very much! I did not find the labels for the individual test scenario's to be descriptive of their intent, and now that I consider that, why did I ignore them? I should have noticed the magic number 64 in the one test as maximum queue depth. So AS SSD writes less than other benchmarks, when it writes one GB? Wow!

I just read that in SSDs, NCQ is used moreso to keep the SSD busy, while it waits for the rest of the PC to catch up to it. If so, that is really something.

Thanks Again!

[idata]

Here is a link to a thread where Intel explain the benefit of using AHCI for SSD's.

/message/97615# 97615 http://communities.intel.com/message/97615# 97615

The one thing I disagree about in that in AS SSD test description is this:

The 4k 64Thrd test is most heavily weighted in the results, as this is, according to experts, the main operation mode of an SSD.

That might be true for enterprise use, but for desktop use it is not much use as queue depths rarely go above 2 when using SSD. (Maybe a bit higher with HDD)

What I look for is performance at queue depth 1(in particular read speeds) and latency, because that is predominantly where desktop performance lies.

[idata]

Robert your AS SSD benchmark should have been coming out with 100 to 110 MB/s for sequential writes. ~80MB/s is what I would expect for an X25-M 80GB, which has slower write speed specs than the 160GB drive. My X25-M 160GB consistently comes in above 100MB/s.

Your access times are not that great either. Are you using the latest firmware? (The first release firmware was superseded with a version that increased the sequential write speeds specs).

[idata]

Yes, I would agree with your statement, particularly regarding the statement in the description, "... the main operation mode of an SSD".

Unless we are missing something here, the motivations and optimizations with NCQ in HDDs have little to do with the manner that SSDs operate. For example, placing read requests in the order that they are sequentially stored on the tracks and cylinders on HDD platters has no analog in SSDs. Also, given the IOPs performance of SSDs, when would the desktop user have outstanding requests in an NCQ queue of more than three or four at most, if even that?

A queue depth of one in the NCQ implementation is the most trivial state, and actually equivalent of non-NCQ operation.

The following is a link to a white paper written jointly by Intel and Seagate about the benefits of NCQ. It's a few years old, but was published when NCQ began to be used.

http://www.seagate.com/content/pdf/whitepaper/D2c_tech_paper_intc-stx_sata_ncq.pdf http://www.seagate.com/content/pdf/whitepaper/D2c_tech_paper_intc-stx_sata_ncq.pdf

[idata]

Below you can see the difference between ATTO and AS SSD Benchmark when I use hIOmon to monitor activity during the benchmark. ATTO is set at QD 4.

With Crystal Benchmark Diskmark it is possible to set the size of the test files. This can be as low as 50MB.

If you have hard raid with say 2GB of cache you need a test file above 2GB if you want to avoid testing the cache rather than the SSD.

The figures you get from a benchmark are only really relevant if your usage patterns match the test pattern. If you use IO Meter you can make a test pattern that suits your particular area of interest, but that requires an understanding of your usage patterns.

I have found hIOmon to be a lot more informative than benchmarks. You can monitor how your SSD performs on the actual tasks you run and get a much better understanding of performance that matters.