Why is my SSD so slow?

I too disable system restore as soon as a clean install of Win 7 boot up for the first time. However, I don't believe disabling indexing is an advantage for most users, nor do I think it will affect the benchmarks of an SSD.

I would be curious to see if Zaz finds disabling indexing of any value in his use.

 

If you're referring to disabling indexing on the drive, yes I did that, but I did not disable the indexing service.

At any rate, I finished my clean install and got everything the way I like it. I then ran the optimizer. My resulting score was 228 on the AS SSD benchmark and honestly can't really tell much of a difference between it and a platter based drive in my boot times or how fast seem to open. Now this is just my perceptions, but if I don't perceive it to be any faster, why spend the money or give up the space.

So screw it. I ordered the 500GB XT. I just hope it ain't too noisy.

 

ZaZ,

I have a 'score' of 320 on my Inferno and when its fast - its nice. But when it decides to have fits the XT in my other notebook run circles around it.

I hope it is quiet enough for you - I don't notice the XT being louder than a normal notebook HD, but it is definitely louder than any SSD.

Remember that with the XT, you want to use it 'normally' for a few days before it hits its stride - this may also affect how loud it is for you as the data will be cached to the nand memory and won't need to position the heads around the platter.

Good luck.

 

Well then, I await the perfect solution.

 

After reading all this stuff, now can some of you SSD gurus recommend all settings to enable, disable, verify to summarize?

- AHCI enabled in BIOS
- install Intel Rapid Storage 'F6' drivers during Windows install (for intel chipsets only obviously)
- verify alignment
- verify TRIM enabled (fsutil command)
- drive indexing off
- write cache off
- turn off paging file / virtual memory (if user's habits don't dictate its necessity)

Anything else?

 

Id' turn off system restore too.

Stamatisx recommends a 4K cluster size for optimal performance.

 

Will system restore affect performance though, or just concern about free space and/or unnecessary writes?

For 4K cluster size, are you referring to setting the 'allocation unit size' to 4096 bytes when formatting it? NTFS defaults to 4K anyhow doesn't it?

 

I'm not sure how much it affects performance but these files have to be written some time, so there should be at least some impact on performance.

Oops I meant 64K cluster size. It works with "format fs=ntfs quick unit=64k". Like described here:
http://forum.notebookreview.com/win...-boot-time-tweaks-allowed-37.html#post6518062

 

Interesting, thanks. Does it really make much difference for day to day use though? Any detrimental effects? I may try this with my next format.

 

I haven't tried to measure it. Maybe stamatisx can shed some light on it.

 

It does make a difference, it gives faster access times, I get better benchmark results and less fragmentation on the expense of free space (which is not that much, but slightly less compared to 4K)

 

Thank you. How much less space are we talking about? I'm assuming that since it would consume 64KB for every block even if it it's only 1 byte, then if you have lots of small files it probably wouldn't make sense? Otherwise probably not a huge difference?

Are there any down sides besides reduced free space?

HP is sending me a replacement SSD, supposedly one that supports TRIM, since their first run Samsungs didn't support it, even though they told me thy would before I bought the machine. Should be this week sometime. I will try the 64K then.

 

That's correct and so far in my installation it has almost no difference.

I am not aware of any and haven't seen any down side after all those months.
Even MS suggests the use of 64K on HDDs.

It won't be a dramatic increase, but even a small increase is always better than nothing

 

That's what she said...

Well with the Samsung any amount of improvement is welcome. All around read speeds are pretty good, but 4k writes are abysmal.

 

I really wish I would have found this thread sooner.
I tested my Intel 80GB SSD quite a bit, and through some reading I also found that if you install Vista using a disc that is pre-SP1, it will NOT align the disc properly. Mine are RTM and it took me a little while to figure it out, I was also getting some low numbers in crystaldiskmark.
Upon booting up with the vista install disk, get to the terminal, and I just opened up diskpart, and manually aligned it, and installed from that partition and my SSD was much faster after that.
You should keep in mind that if you restore from an image you made, it will wipe out any partition you made.
In diskpart I just used

and it was fast after that.

Here is a good article from Microsoft that's worth reading.

 

So if you do an install with a pre-sp1 Vista disc, it will not align properly?

 

No it wont. You have to do it manually.

 

According to the alignment calculator on page five, my SSD is in alignment and I installed SP2 manually.

 

OK, I think I'm onto something here. I noticed when I had the power management software set to maximum performance, the SSD was much more snappy. I ran the AS SSD Benchmark software. It score between 340-360, which is close to the clean install score. When I'm in lower mode, it scores between 210-230. I set the CPU to adaptive performance in power management software. It scored between 300-320 and it seemed quicker, but I'm concerned about the fan being more noisy. One of the things I really like about my SSD is when the CPU is parked in low power mode it runs completely silent. If I put into adaptive mode is the fan going to come on more often? I know it does in high performance mode. I am also somewhat concerned about the loss of battery life if I'm in adaptive mode.

 

I've read some laptops throttle back to SATA I in power saving mode. Maybe this is the case with your X200...

 

So the question becomes if I order the XT, will that be capped too? If do does any SSD or hybrid like the XT make sense.

 

ZaZ,

The subtle nuances and dependancies of SSD's, chipsets, drivers, Windows power saving settings and other inter-related settings/tweaks/etc. (disable system restore, disable pagefile, etc.) have not been discovered yet.

The best answer is that you will have to answer that question in your own system.

My guess is that the SSD is taking a huge hit because of the power savings setting used but a mechanical HD will be able to disguise any High Performance vs. Balanced power setting comparisons much better.

Why? Just a gut feeling I have. But until or unless you try it in your system, you won't know for sure no matter what anyone else may guess at.

As an aside, the Inferno in my system cannot be suffering from the same thing as your Intel because I keep my power savings mode in 'High Performance' all the time (and still get great battery life too - ~8 hrs).

A final point: an hybrid drive is not seen at all by the O/S as an SSD. This fact alone will make it behave differently than the Intel SSD you're now experiencing these issues with.

With the XT, when data is asked for (and its in the cache) it is given as fast as an SSD would give it - but without the overhead that a full fledged SSD has to deal with like garbage collection, TRIM, and Wear Leveling.

If you're able to buy an XT with a return policy, just make sure you'll be able to test it before the return period is up.

Good luck.

 

Not too drag up an old topic, but I've been limping along with my Intel G2 for the past few months. Last week I bought a Agility 2 drive to use in my desktop, but I thought maybe I'd test it out in X200 just to see how it performed as compared to the Intel drive.

The results I got were interesting. Just to recap my Intel results I usually scored around 220 in low power mode using the AS SSD benchmark, which to me made it not much better than a platter based drive performance wise. With the CPU in adaptive mode, the scores were around 300. At max power I came in came in about 350. This was using Vista on a clean install using the Intel Rapid drivers.

Flash forward to now using the Agility 2 doing the same clean install using the Rapid drivers, I scored about 300, regardless of what the power state of the machine was. I didn't really test it out much in terms of how fast the felt in comparison to the old one, but I did find the results interesting. It scored higher than the Intel in low power, but lower than the Intel at max power. I would think if there were something wrong with the machine, it would preform the same, but that's not the case. Anyone care to take a crack at analyzing this one?

Another interesting tidbit is the OCZ drive only hit about 150MBs while reading, but the Intel drive was usually around 250MBs. From what I saw and I wasn't paying real close attention, the OCZ drive was supposed to be slightly better. Thanks for any help.

 

I wouldn't assign to much value to AS-SSD scores. AS-SSD uses compressed data which makes the Agility 2 look less good. If you'd do ATTO the Agility 2 would look much beter.

But who cares about all these synthetic scores. In real life the difference X-25M and an Agility 2 is (/should be) negligible, except for large files copies because the Intel has lower sequential write speeds.

 

Yeah I think theres a tendency to go overboard on the synthetic benchmarks, chasing high scores that due to power management features of most mobile chipsets are only achievable under stress test conditions. Theres a tendency (and not just in this forum) to tweak and tweak and slam your drive with benchmarks which are just gigabytes of pointless writes. If you are running Windows 7 and did a fresh install on your SSD you probably don't need to do anything so enjoy the speed. Stuff like write caching will be enabled if necessary, defrag and superfetch will be disabled, partitions will be automatically and correctly aligned etc. (Vista and XP folks: hop on board the SS Windows 7 as soon as you are able because it rocks.)

There is also a significant disparity in test results depending on chipset (both mobile and desktop), power profiles, operating systems and the compressability of the test data. So it pains me to see people repeatedly running CrystalDiskMark 3, queuing up about 8gb of write tests, not quite getting the 4k result they wanted so they go secure erase, reinstall (another 30gb+ of pointless writes) and start benching again. Its not that you are writing to the drive alot that erks me. Its that its all pointless. (edit: I don't want to disseminate the idea of aggressively limiting writes to your SSD because theres a tendency for people to overboard on that too, up to and including disabling system page file (!), system restore (!) and moving their browser cache to an external hard drive even though browser cache read/write patterns are exactly the kind of thing that SSDs are good at and HDDs aren't).

In the end, you bought an SSD for the speed and it came with a 3 to 5 year warranty. Provided you aren't doing something crazy like running synthetic benchmarks every day or running massive databases or file servers you aren't going to wear it out any time soon. Very few people in this thread are showing results which I would describe as aberrations. Mostly its paranoia over whether or not their benchmark scores are an aberration. Somebody a while back seemed worried about the speed of their SSD by citing 4kqd32 scores of 80mb/s when they should be getting 130mb/s (caveat: under conditions where power consumption and heat are not issues, on the right chipset). I mean, thats still like 8 times higher than the old Indilinx drives which was higher than any mechanical hard drive anyway.

I guess this is just what happens when industry folks differentiate very similar products by speed and market them accordingly. It leads consumers to believe that an inch is closer to a mile than it is to a foot.

 

I don't know that I'm chasing high scores, but I do know my Intel drive is significantly slower in low power mode. I didn't really test the OCZ drive much for app load times or boot times, but just mainly used the benchmark to note the differences.

 

Pretty much everyone's SSD is slower in low power states unless they have been deliberately disabled. Thats what low power states are for - things get turned down or switched off when not in demand. Furthermore the Sandforce drives will also slow down the more you fill them up because they'll use all unwritten or TRIMed space on the drive for over provisioning if its available which is a good thing. It does not however stop the mob over at the OCZ forums from panicking that their drive performance is 'degrading' the more they fill it up. Thats like complaining that your engine is getting louder when you drive faster.

Aside from the SSD, during various idle states your cpu core voltage and clock speed will also step down. Your mobile gpu will do it too. My desktop PC has a Radeon HD4670 that drops its core and memory clocks to 300mhz/570mhz from 750mhz/750mhz respectively when my computer is idling. This is completely normal and a good thing and sure enough, when you fire up Mass Effect 2 with CCC open you see the clocks switch back to stock settings so no harm done eh?

I don't know what else to say. Your benchmark numbers were very good. Heres my Vertex 2E 120gb for comparison:



Which makes sense at idle and given the data is not very compressible.

I knocked these up real quick to show you how my 4k writes more than double in CrystalDiskMark if I run a Prime95 torture test in the background as opposed to benching it during an idle state.

Idle vs Stress Tested:

 

I spoke with Intel tech support. They told me the drive shouldn't be slowing that much using low power mode. Was your Vertex score at high or low power?

 

AS SSD was run during idle. If you want I'll run it again with Prime95 torturing my computer in the background but ehhh. It doesn't really matter and I'm talking about the drive I own here.

 

I'm taking about the power state. When my CPU is set to low power, the SSD slows significantly. It's pretty peppy at maximum power, but then I've got fan noise issues.

 

Hayte, if you change thise line while running in high performance mode your SSD performance will be restored to normal:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Intelppm

Change on "Start" from 3 to 4.

 

Wait, why would I want to disable that? Intelppm is an important process that controls cpu stuff like turbo boost?

edit: yeah it looks like it. I wouldn't disable cpu idle states via the registry either because it will cause your cpu fan to spin up and will make your computer sound like a dustbuster.

Basically what these tweaks are doing is defeating all of the things that you bought a laptop for: low heat production (enabling it to be fitted into a smaller chasis with lower heat dissipation), low noise, low power consumption and mobility on battery range. If you are going to defeat all of those things to get higher numbers in a synthetic benchmark you may as well just buy a desktop computer where power efficiency and mobility is of no concern.

 

I did not know you have turbo boost. The tweak that keeps turbo boost is a bit more complicated, you can find it here.

After applying the tweak on my notebook installation performance went up with as much as 50% (sometimes even more). I don't care about synthetic numbers.

No, if you only apply the tweak to the high performance profile you'll have installation performance back to normal when you need it.

 

But I already posted screenshots showing that under heavy load conditions my 4k writes more than doubled without having to disable any cpu related drivers/processes.

 

Yes under heavy load, but installing a program isn't heavy load.

Here's what the tweak did to my Acer 1830T:
Installing Photoshop CS4 - without tweak 9:30 min. With tweak: 4:30 min.
Installing MS Office 2007 - without tweak: 8:32 min. With tweak 5:18 min.

Anyway I'm glad that I can have maximum SSD performance when I need it.

 

My own experience with that particular tweak was that my 4k numbers almost doubled, but at the cost of hi def video playback from file. I had lost turboboost as a result of the tweak and my cores were going haywire. Things were fast, but they couldn't smoothly handle the playback of .m2ts files. It would freeze/unfreeze rhythmically.

Upon system restore, but core activity normalized and playback was much smoother. It seems we experience this tweak quite differently.

 

Well I have no turbo boost. By the way do you mean you got all those problems from only disabling Intelppm?

How did the JJB tweak work for you?

 

Only one comment from my experience so far, neither mine nor JJB's tweak could give me the maximum SSD performance (I am talking about the same SSD).

On a previous generation chipset (MCP79 from nvidia) with a C2D CPU, the same SSD was performing much better than it is now on an i7 940XM overclocked with the Intel PM55 chipset, even if I run IntelBurnTest or wprime in the background.
My opinion is that the SSD performance shouldn't be connected with the CPU load (the more CPU load, the better the SSD performance) and I don't know if Intel did that on purpose or by accident but it makes no sense. Wouldn't you like to have an SSD with 0.05ms access times with or without CPU load (like on the nvidia chipset)?

 

I take back what I said. I blended my experiences. The intelppm tweak disabled my turboboost, but JJB's tweak gave me "crazy cores".

 

i think i am in the same boat as anseio
just hope intel fixes it properly...
while we all appreciate these fixes, its like a "band-aid" fix atm...lose cpu power(or boost) or go with better ssd performance for some users...