My Patriot Inferno SSD performs bad.

The problem? The DuraWrite technology slows down the writes to prolong the life of the nand chips. If you write more than the HD 'thinks' is 'normal' for/in a certain period of time, it will artificially limit the write speed to protect the nand chips - the recovery might take days for DuraWrite to call itself off - but in the meantime, if you're continuing to do editing with large files (or, many smaller ones, like still images), the performance is worse than a 2-3 year old 5400 RPM HD (i.e. less than 60MB/s).

Just something to look out for - or, at least be aware of.

Edit: Sorry, I think I was mixing up DuraWrite with 'Lifetime Throttling'.

Edit #2: Also, with the size of SSD you're looking at (256GB), it may be less of an issue than what I experienced with my 100GB Inferno.

 

Source: AnandTech :: Your Source for Hardware Analysis and News

 

Phil, I have no trouble believing Anand's testing methods, nor in trusting his findings. However, nobody buys an SSD and uses it solely like how Anand tests it.

Source:
General Discussion Secure Erase, TRIM, and anything else Sandforce


(See Post 513 in above link for context of the quote above).

Here is my Inferno SandForce based 100GB SSD that was installed a mere 6 weeks ago:

 

He and Techreport simulated worst case scenarios and never saw any of the problems that you saw.

The post you are linking to is running in RAID. RAID means TRIM is not working. For people with TRIM the performance should bounce back quickly.

Here's what a normal functioning Sandforce drive should perform:


Your whole performance looks very bad, not just your write speeds.

Could be there's something wrong with your SSD, notebook, software or drivers.

I had similar performance when I had write caching disabled. Try that.

PS. I've moved your posts to this thread. Feel free to edit the first post in the thread so that it's more descriptive of your problem.

 

Phil,

Thanks for suggesting to check the write caching on the drive. It was enabled, but trying it both ways did not improve the drive's performance. I even tried the 'Turn off Windows write back caching on device' option and again - no difference.

The post I quoted may have been using a RAID setup - but even that author has tried a single drive (search that post) and he, along with others has found the same performance hit - TRIM effectively does not work.

 

It remains strange that neither Anandtech nor Techreport (or any techsite) found something similar.

Looking at your 4K read performance, your SSD doesn't look like it's performing normally in general. I doubt this has anything to do with performance degradation.

PS. Please tell me how I could get the same performance degradation as you, I might be testing a Patriot Inferno SSD soon.

 

Phil,

Six weeks ago (almost to the minute), I installed Win7 x64 and my standard programs (not including the 'user' folder ~55GB including the Windows directory). I then proceeded to edit/convert about 4GB of images a day (albeit in 4 or more different programs so equivalent to 16GB of data).

I also run an 8GB outlook (pst) file. And except for the last few days, MSE too.

With this usage, I could see the drive getting slower and slower.

I will be very interested in your findings with my specific SSD.

 

And if you let it idle overnight, does that change anything to the AS-SSD result?

 

I've left it idling for almost a week (7 days) with no usage and the only change is the 'score' went from 323 to 318.

 

Well that makes a lot of sense because idling should only increase performance if you don't have TRIM.

The 4K random read is 40% lower than it should, 4K write is 55% lower.

Did you check partition alignment?

What does AS-SSD look like when it's in safe mode?

 

The alignment shows correct as per AS SSD screenshots.

In safe mode - hahahaha... the mouse doesn't work!

But, I still got a screen shot see below - no notable difference (score=331).

The second screen shot is with the MS-AHCI driver installed (not IRST) and that shows the best 'score' so far - note this is in 'normal mode', not safe mode.

The drive looks very bad to me - it is not the system - it's the SSD.

I will try a secure erase and install it on my desktop to see if it performs the same later this week or early next week (pretty busy right now).

Thanks for the suggestions.

 

You've probably done most of these tips but I though I post it here anyway. These are tips to enhance SSD performance.

1. Make sure you did a clean install, from the ground up (removing all partitions before you start by choosing custom setup)

2. Enable AHCI in BIOS.

3. Install the latest Intel Rapid Storage driver (Download Intel® RST Driver Files for F6 Install from Intel.com. Go to Device Manager, select under "IDE ATA/ATAPI controllers" the "Standard AHCI 1.0 Serial ATA Controller". Then right click, update driver software. Browse, browse. Now browse to where you saved the Intel Rapid Storage driver. Select the folder and click OK. Reboot required.)

4. Enable write caching (Device manager → Disk → Double-click on you disk → Strategy tab → check BOTH)

5. Check that TRIM is enabled (Fsutil command)

6. Disable scheduled defragmentation

7. Disable pagefile

8. Disable hibernation

9. Disable indexing service

10. Disable prefetch

11. Disable superfetch

12. Check partition alignment (shouldn't be necessary if you did a clean install with Windows 7)

 

this really looks like a chipset issue.. G73 is also getting a similar problem and has to do with power saving issues with chipset... have u tried taking the benchmark under full load of CPU? It seemed in the G73 then the SSD's performed properly then. Try this tiller.. if it works , i think we can confirm that u have the same problem as G73! Only thing i can think off...

 

Les, I have always disabled system restore - since day two (and three installs later) of Windows XP.

I also ran AS SSD in safe mode and it showed a 3 or 4% improvement (from a score of 318 to 331).

Yes, I have had less than stellar experiences with SSD's - but it is not the benchmarks that make me realize this - it is the platforms from which I'm coming from.

I've always had the fastest HD's, with appropriate partitioning (some call it short stroking) and a program to defrag them optimally (the last few years PerfectDisk is untouchable). When I try/test a new SSD - I install all my programs (or as much of them as I can, considering their smaller capacities) and just use them as I do my HD's.

With the SSD's, do I see glimpses of speed? No doubt! But in a sustained workload situation of daily converting RAW images to finished jpegs, tiffs and PSD files, I have yet to see an SSD maintain a speed advantage over mechanical HD's.

Is the Inferno I have defective as Phil suggests? Maybe, but the OCZ forums are stating that this 'defect' is by design (DuraWrite).

I may be the only person on this board that does not see SSD's as the holy grail of fast storage - but I think that is because I come from a postion of knowing what fast storage really is.

Not to mention, that other internet sites have reported my exact same issues with SandForce drives long before I even bought one (wish I knew to search for 'Lifetime Throttling'.

Like I've mentioned to Phil before, I trust Anand up to the level of testing he does - and that testing in no way mimics on how I and many others use or want to use these drives.

I also respectfully disagree that a degraded SSD is still superior to a 7200 RPM HD - it depends of course on how far degraded the SSD is.

 

sean473,

that's a great idea!

Do you have any specific program to stress the CPU with?

I'll try with Intel Burn test and edit this post soon.

(rubbing hands together....)


Edit:

sean473 you're amazing! I'll be posting benchmarks soon, but you're right - unless the CPU is cooking - the SSD is sleeping.

Wow!

Now, a spoiler: Even though the SSD benched so well - I tried launching PS CS5 while Intel Burn was running on high (4GB RAM Stress test) but got worse times: 19 seconds (vs. about 7 or 8 normally with this SSD vs. 4 seconds with XT on a slower platform).

So, how do we 'turn the chipset on' permanently?

I'll post later with the benchmarks - have to go shoot some dancers now.

 

Phil,

please understand that I do not have a crusade against SSD's. I thought I was engaged in an ongoing conversation and am simply offering my direct observations about them. (Observations which keep evolving, btw).

I will keep my comments constrained to the thread above - sorry to be so aggravating to you and the other mods that you have to delete my posts - please believe that that is not my intentions (to aggravate anyone).

I am simply trying to get the best performance possible for my hard earned dollars. Same as everyone else here, I suppose.

Thanks for your patience with me.

And again, my observations are based on reality - the only difference is that my reality with SSD's so far has been quite miserable - unfortunately.

 

So, just like sean473 suggested, I tried running the benchmarks with the cpu loaded.

I used Intel Burn Test 2.50 at very high settings (using 4GB or RAM - I have 8GB installed) and once this stress test was started I ran crystaldiskmark and AS SSD benchmarks to see if the 'scores' would improve.

To my surprise, they did.

Since I can't figure out how to include images linked to certain text, I'll be making multiple posts to show the results I got.

The first benchmarks are with IBT (Intel Burn Test) and without IBT for CrystalDiskMark.

Quite an improvement in the 4K reads and writes!

 

no problem dude.. so like i said , its got to do with chipset and power saving issues... according to hell cry , under load when copying files and all , the SSD will perform ok.. in idle mode , shouldn't have any effect... i don't know how to pernmanetly stress CPU... it seems u need a BIOS update.. Inform Gary Key(Asus rep here ). that ur having same problems as G73 with SSD... and see if he can do anything after he fixes all G73 issues

 

This might be interesting

 

We have recently been having trouble with our exchange server and people complaining that Outlook is really slow which turned out not to be surprising given how few people were auto archiving and deleting really old messages with really large attachments. The absolute worst offender was the reception computer which had a mailbox size of about 2.3gb. I'd like to add that this computer was retaining emails and calendars from as far back as February 2005.

There is absolutely no reason why you cannot archive 7.8gb and move that archive.pst file off your system disk.

Secondly, your as-ssd benchmark results are really not bad at all when you consider the following:

1) You are using as-ssd which uses highly incompressible test data and this gives Sandforce drives their least favourable results. Try using ATTO which by contrast uses highly compressible test data.

2) the apparently huge amount of writes you are hammering a client class drive with. If drive life and the intervention of Duraclass bothers you, then you should be using an enterprise class Sandforce drive as mentioned in the ssd benchmarks thread.

3) Your benchmark scores aren't even that bad. Your 4k reads/writes are similar to those of my Vertex 2 at idle states which the above 2 posters noted on pm55/hm55 chipsets since they have fairly aggressive power saving features. Your sequential writes are bad but honestly, how often are you writing sequentially to your disk? Mechanical hard drives are great at sequential writes because the read/write head doesn't have to seek around the platter. If you are writing sequentially alot you can do this to a secondary hdd and really not notice any real loss in speed. Random read/writes, especially small ones are what ssds are really good at than hdds aren't so things like caching and page filing you want to do on your ssd.

Whether your low seq writes are down to Duraclass or a misaligned system partition or something else I don't know. Nobody but Sandforce knows exactly how Duraclass works anyway. You could always secure erase and remount an image of your system disk which should get you day 0 read/write speeds. You don't want to be making a habit of defeating Durawrite like this however.

 

Hayte, it turned out his low benchmark scores were caused by the PM55 chipset problem. Since he applied some tweaks his benchmark results look fine.

 

If you install IRST via Device Manager, you will be missing the service IAStorDataMgrSvc which is only installed with Setup. I experienced better scores when this Intel Rapid Storage service is installed and running.

 

Interesting, I will have to try that.