SSD Thread (Benchmarks, Brands, News, and Advice)

And you refreshed the WEI score in performance information and tools and it's still 7.8?

 

yes, just re-ran WEI assessment again, and still get 7.8. no clue why WEI offers this more than P256.

 

i've 7.6 on my g1

 

i've seen G2 is reported with WEI from 7.4 to 7.7...

 

i always said i don't care about g2 (i would like to have all g2, of course). it's an improvement to g1, but g1 delivers all you want very well

 

I ran the winsat disk and then cleared all WEI scores and re-ran the assessment and got 7.2. Comparing this winsat score with right after flashing with new FW, it looks like it puts a lot of weight behind the 16k reads.


This is right after flashing when it got 7.4

 

"An example problematic sequence consists of a series of sequential and random I/Os intermixed with one or more flushes. During these sequences, many of the random writes complete in unrealistically short periods of time (say 500 microseconds). Very short I/O completion times indicate caching; the actual work of moving the bits to spinning media, or to flash cells, is postponed. After a period of returning success very quickly, a backlog of deferred work is built up. What happens next is different from drive to drive. Some drives continue to consistently respond to reads as expected, no matter the earlier issued and postponed writes/flushes, which yields good performance and no perceived problems for the person using the PC. Some drives, however, reads are often held off for very lengthy periods as the drives apparently attempt to clear their backlog of work and this results in a perceived “blocking” state or almost a “locked system”. To validate this, on some systems, we replaced poor performing disks with known good disks and observed dramatically improved performance. In a few cases, updating the drive’s firmware was sufficient to very noticeably improve responsiveness."

http://blogs.msdn.com/e7/archive/2009/01/19/engineering-the-windows-7-windows-experience-index.aspx

 

micron C300 256G AS SSD benchmark:
http://www.micronblogs.com/2009/12/as-benchmarks-for-realssd-c300/

 

heh, they do report quite detailed their knowledge i like that.

 

very likely, and the 4k random write you ran is also interesting, i did it several times and got this: (the number from winsat--74MB/s-88MB/s-- is much higher than AS SSD benchmark for 4k, which is only 25.23MB/s for 4k and 61.28MB/s for 4k-64thread ), for 4k random read, the number of winsat 112.92MB/s is between AS SSD's single thread number 14.73MB/s and 4k-64thread 133.67MB/s.

 

Yeah, the winsat 4k random write only runs for 5 seconds and AS runs it for 2-3 minutes. I don't know about the Intel drives, but the Samsung drives start out at like 15mb/s(probably the 128mb cache) on AS and gradually decay to about 4mb/s by the end of the test. Not sure how useful running 3 minute sustained 4k random writes is for typical retail consumers.

 

same here in P256, but intel changes only between 25mb/s-26mb/s during benchmark time. controller algorithm makes difference i guess.

 

best SSD for $$: http://pchardwareblips.dailyradar.com/story/best-hard-drives-and-best-ssds-for-your-money-holidays/

 

I don't know your sources (but would like to read them), but as I hinted Anand stated it himself in the 3rd SSD anthology he wrote:

Quote:
"This could mean that in the real world, Indilinx drives stand to gain the least from TRIM support. This is possibly due to Indilinx using a largely static LBA mapping scheme; the only spare area is then the 6.25% outside of user space regardless of how used the drive is."


From here:
http://www.anandtech.com/storage/showdoc.aspx?i=3631&p=9

 

Phil,

Good to see battery life is normal again. Did you do anything to specifically fix that?

Just scratching my head here a little... this is the same drive you stated as 'lightning fast', yet you want to return it because it benches 'wrong'?

You may want to try the different disk caching options (they may improve the 'score' for you), but I'm really puzzled why actual/real performance is taking a second seat to benchmarks in your decision.

See:
http://forum.notebookreview.com/showthread.php?p=5664935#post5664935


Hope your Holidays were great!

 

No, I don't change my 'theory' at all!

No GC is not TRIM. They both do the same thing (effectively) but the methods are markedly different.

A TRIM command is sent when we delete, format or move a file. The O/S itself tells the SSD exactly what to 'trim'. The key point here is that the O/S knows what is needed (valid data) and what is not (non-valid data as opposed to 'clean' nand) while GC does not know this important distinction.

GC is basically a defragmentation strategy for SSD's. GC does not know what is valid data or what is non-valid data. It only knows used nand pages and unused (clean) nand pages. What it does is re-arrange the used nand pages so that the unused/clean nand pages will be the first thing it writes to with the next write request.

As you can imagine, GC is very 'write-wasteful' compared to TRIM and not only does it eat into the power savings that SSD's promise (because the drive is working whether we are using it or not), but it also writes/cleans and re-writes needlessly because it is a 'dumb' command that is not in sync with the O/S which 'knows' what is needed data and what is non-needed data.

With all the above information (which you can find in the 3 part anthology of Anand's SSD trip down into the rabbit hole... ), I not only prefer TRIM as Windows 7 is already my O/S of choice (and of course supports TRIM inherently), but I would also go out of my way to not get a GC SSD because of the above deficiencies noted above.

Okay, your turn again... is my 'theory' believable now?

 

In this post a few pages back:

See:
http://forum.notebookreview.com/showpost.php?p=5665476&postcount=2398


vostro1400user posted his winsat disk scores for the Intel G2 (160GB, I think) SSD. (Sorry, can't quote his post and still show the attachment in this post...).

There is a lot of talk in the last few posts of how benchmarks are weighted/biased and I too am interested in this because I don't care how they are biased - as long as I know by how much!

Comparing to the 7K500 winsat disk scores here:

See:
http://forum.notebookreview.com/showthread.php?p=5664935#post5664935

I'm actually quite amused that the mechanical drive scores even higher than the Intel in some of the sub-tests. (To me, just shows how unreliable these 'benchmarks' are at predicting RWU - that is why its amusing).

What it does tell me though, is that each subscore (for the disk WEI 'overall' score) is biased/weighted, based on other scores within the 'winsat disk' benchmark. Therefore; it is a sliding scale that we don't know the curve of.

What is even more 'alarming' to me is that the raw score for Responsiveness Overall is 6.9 WEI for both, yet the Hitachi scores 71.16 units and the Intel scores 52.68 units. (I'm using the winsat run with the settings I'm currently using: disk cache enabled and buffer flushing disabled - both Policies checkboxes 'checked').

I realize that as the performance get closer to the 'insane' side of the equation, the scale has to slide ( until all storage solutions are at the same relative level), but it doesn't help us too much right now finding real and significant differences with only SSD's that perform so close to this stratosphere.

But, I guess this is why I'm waiting for Anand's SSD anthology, part IV.

 

davepermen,

I understand you got the updated drivers running, great!

I would be most interested in your comments if you had a chance to run the winsat disk benchmarks and the various settings for disk cache on, disk cache off and disk cache on and disk buffer flushing off and tell us if an Intel G1 gets affected in any noticeable and/or significant way.

I wish I had tried this with the Torqx I had bought (and returned) a while back.

I will definitely be trying this with my clients systems in the new year, so if you don't have the time or inclination to do this, I will know soon enough myself (and post if anything interesting is found). TIA.

 

Hello Phil, your SSD looks fine. Take a look at my screenshot with my G1 I took few weeks from each other:

10/23/08


11/04/08


12/15/08


You see, the dip stabilizes after a while. The drive arrived at my doorsteps at 10/20/08, and the first image was right after being installed with Windows XP. I assume the dip is related to the files that are installed, so you can't really avoid it. The irony of your post I quoted is that it says you feel nothing wrong with it, yet since the benchmark look otherwise you think there's something wrong.

What you should do: Stop benchmarking and ENJOY!!

Your peak will be lower than mine because I use it on the desktop. It looks like SATA2, and well implemented for mobile.

(Regarding power consumption you might want to check something related to Device Initiated Power Management and mobile computers, like here: http://www-307.ibm.com/pc/support/site.wss/MIGR-67811.html

Without proper DIPM it might run at its peak power of ~2W much more than normal)

 

that poor ssd... xp on an ssd, evil

but nice to see how it gets better over time i agree, don't return it, phil

 

If my Intel G2 is capped by SATA I how many MB/sec would I be getting in HDTune?

Yes change critical battery level and update Bluetooth drivers.

If I buy a product that is supposed to do 220 MB/sec and I get 150 MB/sec on the first 10% and 198 MB/sec on the rest, I return it.

I paid a premium for premium performance. Besides that 10% CPU utilization seems wrong too, every time I run the benchmark.

PS. there was someone in NL who had the same problem, he send it back and got a new one which did not have the problem.

 

in AS SSD benchmark, 4k-64thread write test prompts 50% cpu usage for 2.6G C2D, and 4k-64thread read test yields 90% cpu usage in this 160G G2, seem to be much worse than yours.

 

How is my Crystal compared to others?

I was only talking about HDTune which is a simple sequential read test.

For Hdtune 10% is unusually high when there's nothing happening in the background.

 

Bit slow on the write if you have the newest firmware (and I'm assuming the 160GB version) - reads are excellent.

 

you are lucky, i got 20-30% cpu usage for Hdtune sequential read test(block size=64k), for CDM benchmark, here is mine:

 

some complains high cpu usage as well:
"Cons:
3. VERY high CPU overhead/utilization"

http://www.newegg.com/Product/Produ...eoOnlyMark=False&VendorMark=&Page=1&Keywords=

 

I moved my questions about the Hd Tune dip to a separate thread:
http://forum.notebookreview.com/showthread.php?t=446004

I found some info about it, I posted it in that thread.

If anyone has an Intel G2 80GB SSD in a Windows 7 notebook please post a HDTune free shot there.

 

it seems old 02HA firmware has higher 4k write rate in CDM:

http://www.xtremesystems.org/forums/showpost.php?p=4088188&postcount=705

 

Different drivers, controllers and different Windows versions (x32 vs. x64) may also influence results.

 

And potentially IDE mode vs. AHCI... its seems I'm getting about half the speed on 4K writes compared to my original speed in IDE mode after installing SP2...

Which is rather strange... but its write, not read, so no real worry

4K read dropped by about 1MB/s... i.e. nothing as it fluctuates a fair bit too.

 

http://www.xtremesystems.org/forums/showpost.php?p=4064610&postcount=700

 

Hmm... pretty much the same... i.e. inconclusive.
Thanks for the link.

Another forum you frequent?

 

(BTW I use Win 7 now)

You are being downright paranoid. The desktop system I use gets less than 230MB/s. If you bothered to look. G2 is supposed to get little higher in everything at least compared to G1. Your scores are low because its a laptop and SURPRISE, laptops are crippled for lower power consumption.

 

You may be right but that is not an answer to what I was asking. I'll ask it again:

If my Intel G2 is capped by SATA I how many MB/sec would I be getting in HDTune?

 

I don't see laptops crippled with the benchmarks people get...

 

well, afaik, every real world comparison test I have read, IE like this one :
http://techreport.com/articles.x/16979

but, as I said before, new fw, afaik, has solved this problem

 

well, the thing is, no one actually knows how, so called "GC" actually works since it is not explained from manufacturers yet, it has only been guessed how could it work, from anand and others
on the other hand, trim was
GC isn't even official name for the function itself afaik, it was named gc by users and reviewers

 

Post #2420, every question about since your first post about peculiar HDTune should be there.

The very best finely tuned desktop systems would do close to 250MB/s. Average probably in the 220-230MB/s range. Laptops seem to do around 10% lower or so for a good one,

G1 on laptop
http://www.notebookreview.com/default.asp?newsID=4647

192MB/s

G2 on desktop
http://tpholic.com/xe/ibmreport/2358680

232MB/s

The G2's do 14MB/s on laptops with random read 4K and 20-22MB/s on desktops. In first glance, it shouldn't be that random reads are bottlenecked by the SATA, its(random reads) way below 100MB/s anyway. However it is, because its a low power controller.

 

Or its a difference in the driver.
For example LPM is left on on laptops and turned off by the driver on Desktop chipsets. (in theory)

 

Well thanks, that's a lot more helpful.

Thanks to Vostro1400user who helped me here, I managed to get a normal HDTune shot. (1MB block size)

 

I think that is why eye says SSD's are "esoteric"

 

I see your pagefile and MFt is right in the middle of the drive too. I wonder why Windows puts it there?

 

Phil: You are welcome.

I bet the SATA controllers have lower operating frequency in mobile versions just like mobile CPUs and GPUs do. TDPs are lower, its lower for a reason.

We gotta start differentiating between "GC" and "idle GC". Garbage Collection happens in all memory devices, be it hard drives, RAM, SSD. In SSD they only do it at write time to not waste extra write cycles, SSDs with "idle GC" are an exception because they probably have additional algorithms that make it work similar to TRIM. An unofficial TRIM that is. Or maybe the "idle GC" drives don't care and waste write cycles, who knows?

 

Thanks for that, now I understand why you were calling me paranoid. I never even saw that post. This thread is too big to keep track of.

This a good moment to request to everyone: if you have a specific SSD question or issue, please start a separate topic.

This thread is very large with topics running though each other, and very hard to search.

If you start a separate topic it's easy to keep track off and much more easy to find while searching.

 

Actually, in that case I have my own paranoia. Sorry about that.

Here's the G1 with Windows 7, benchmarked November 27, 2009 using CrystalDiskMark 3.0 Beta 1

 

lol...

Yes, manufacturer's do not have to explain everything for the function to be guessed at very, very well (by Anand, not me).

And, correct, GC isn't the official name the official name is "Oh! My God! We have to do something before we can't give these things away!".

The explanation I provided is, again, not mine, but the summation of what I have gathered on this subject from all over the web, Anand putting it most succinctly. You don't have to believe it, but that does not negate the fact that it is the closest we have to the 'truth' about how GC (no matter what we choose to call it) operates relative to TRIM.

Basic facts:
GC = 'dumb' SSD defragging
TRIM = 'smart' high level O/S initiated nand cleaning with the least write amplification factor possible.

Cheers!

 

Well when I first installed the G2 Battery bar went from 5:29 to 4:40 (or something like that)

When I updated the BT drivers and so it went back up to 5:01.

Now after installing that Thinkpad patch, updating my touchpad driver and setting the hard drive turn off period to 1 minute I'm back at 5:20.

 

The CPU utilization in benchmarks will usually show higher with a faster drive. Faster drive makes it a faster system, so CPU utilization goes up. Get it?

I'm not sure how accurate battery meters are. 30 min/5% difference can easily be made up by some inaccuracies in the prediction. What I figure is in most laptops, an SSD will make the results more predictable.

Run down the system and do a battery life benchmark for us?

 

I won't have time to do that but Battery Bar is a lot more accurate than Windows' battery life indication because it estimates battery life based on actual power usage over time instead of current power draw.

 

Strange... the other way round for me...and battery bar always breaks after a little while on my computer...