Brand new Force 3 120 GB, broken 4k random write!

Hello everybody.

 

I've just put a 120gb Force 3 in a client's HP laptop (Pavilion dm3-1111sa, dual core AMD CPU, 4gb RAM, latest useless and option-free BIOS). It was to replace a dead spindle.

 

The Win7 product key has rubbed clean off the underside of the laptop so I had to use the recovery discs to reinstall onto this new SSD.

 

So everything's shiny and new, yet benchmarks clearly show that the low-end random write speeds must surely be broken and I can't imagine what I might be able to do to fix it. I'd have preferred a cleaner install, but fear I can't achieve it without a Win7 product key. But surely an HP recovery process can't be solely responsible for degraded random write performance like this!

 

I'd appreciate any timely help anyone can offer, so I can return the machine to it's owner and get a suitable smiley face in return. Right now the specs suggest it'll still be significantly better than what they had before, but with uber-slow, short random writes I can't help thinking it's not running as smoothly as it ought to be...

 

:confused:

 

[EDIT:] Oh, and the SSD seems to have shipped with the latest firmware, 5.02. I believe everything is as up to date as it can possibly be.

here is a good link to free software that will identify your product key for a clean install.

http://pcsupport.about.com/od/productkeysactivation/tp/topkeyfinder.htm

 

You'll probably want to a secure erase of the drive before you try a clean install.

Unfortunately the key isn't much use - it'll be some OEM key that will only work if I install from the correct media, which I don't believe I have.

 

I've used magic jellybean to discover the current key, but I doubt I could make it work.

 

It doesn't seem likely this is performance 'degradation', as such, rather some problem due to a hardware incompatibility, or a fault.

 

I note that it's only QD32 writes which are slow - might that be significant?

 

[EDIT:] I just tried installing some fix pack from the HP site, hoping it might not have been used. Can't tell if it did any good, but I did another 4k Qd32 test and it shot up to 36mb/s! Rebooted and it's back down to < 1.

CHECK YOUR PARTITION ALIGNMENT. It's possible that the method for recovering / restoring the OS did not align the partition correctly for an SSD.

Will do, once my reinstall finishes. I'm trying an unactivated install of 7 x64 to see if the problem persists before I reinstall from the recovery disks.

 

Incidentally, I'm concerned that when I tried to use Parted Magic to secure erase the SSD, it finished instantaneously, which can't be right.

 

I'm having no luck at all. It's a wonder if works at all - it even boots pretty quick. If I hadn't run the benchies I might not have realised anything was wrong!

 

I'll report back when I can confirm that the partition I'm testing is correctly aligned.

Incidentally, I'm concerned that when I tried to use Parted Magic to secure erase the SSD, it finished instantaneously, which can't be right.

 

Actually yes, this is correct. The internal secure erase of an SSD is completely different from standard HDDs. It finishes typically in under 5 seconds for most users. ::pirate::

If it's writing 1s to the whole drive, it can't possibly be that quick.

 

At the end of part C here I quote:

 

"Give it time to finish, once it is done it will disapear."

 

This is for an SSD, and implies it requires time as I'd expect. Mine wasn't even one second. It was as if it didn't actually bother doing anything.

 

Anyway, I've now reinstalled 7 x64 SP1 clean. It's picked up all the device drivers, at least basic versions - nothing flagged in device manager. I've run crystal mark again and I'm gettign essentially the same results although all the numbers are a fraction higher. This time the fastest of the 5 runs for 4k QD32 random write is 1.946mb/s, and other runs were as low as 0.2.

 

I've run msinfo32 and the offset is divisible by 4096, implying the partition in question is sufficiently aligned.

 

I'm not convinced it's been properly erased, so I might try the drive in my own machine and see what happens.

If it's writing 1s to the whole drive, it can't possibly be that quick..

 

 

Actually yes, this is correct. The internal secure erase of an SSD is completely different from standard HDDs. It finishes typically in under 5 seconds for most users. ::pirate::

 

You may want to do some Googling if this is a concern for you. As I said, the internal SE of an SSD is COMPLETELY different from how it is done on an HDD.

 

In short, SSDs are secure erased by a controlled voltage spike from the controller. Writing 0s or 1s to any disk is part of sanitizing but the INTERNAL SE of an SSD does not do any writing to the NAND at all. It zaps it. ::pirate::

If it's writing 1s to the whole drive, it can't possibly be that quick.

 

At the end of part C here I quote:

 

"Give it time to finish, once it is done it will disapear."

 

This is for an SSD, and implies it requires time as I'd expect. Mine wasn't even one second. It was as if it didn't actually bother doing anything.

 

Anyway, I've now reinstalled 7 x64 SP1 clean. It's picked up all the device drivers, at least basic versions - nothing flagged in device manager. I've run crystal mark again and I'm gettign essentially the same results although all the numbers are a fraction higher. This time the fastest of the 5 runs for 4k QD32 random write is 1.946mb/s, and other runs were as low as 0.2.

 

I've run msinfo32 and the offset is divisible by 4096, implying the partition in question is sufficiently aligned.

 

I'm not convinced it's been properly erased, so I might try the drive in my own machine and see what happens.

 

Do you know if that laptop is operating in AHCI mode? Given that it is only the performance at high (or any) queue depth ("QD 32" in the benchmark table) that tends to indicate the SATA mode is IDE, so no NCQ is working.

 

You should search for "AS SSD", download (free) and run it. That will verify the SATA mode and alignment.

 

Laptops and their, as you correctly put it, "latest useless and option-free BIOS", are the cause of most of the issues laptop users have with SSDs. Can you find out what SATA chipset and SATA driver is installed on that laptop?

 

If the SSD you installed is new, there would not be a need to SE it.

 

A SSD is not secure erased by writing to it, as has been said in this thread. The SE tool provided by a different SSD manufacture that I've used completes in no more than a few seconds, so the time span of a SE is not any indication that it did not occur.

 

If that SSD performed better in your PC, that would obviously point to the laptop.

Ah, I see now how the short SE time could be genuine, thanks for the explanation.

 

I've run the drive in my main machine and the 4kQD32 write speed jumped all the way back up to about 180mb/s or something. clearly a laptop issue.

 

I've run AS SSD and it seems to be concurring with msinfo32, that the partition is appropriately aligned, AND that "MSAHCI" is "ok" which I presume means AHCI is indeed enabled.

 

Also, Device Manager lists the IDE/SATA controller as a "Standard AHCI 1.0 SATA Controller", so I guess it SHOULD be able to cope. I'm not getting any indications that there's a better driver available, so I'm about to embark on some hardware ID research...

 

[EDIT:] An additional question, while I'm here: I noticed upon unboxing the SSD that, in contrast to others I've used in the past (including three of my own from Intel and Kingston), this one is floaty light by comparison. I always wondered how much weight be might saved by swapping a laptop spindle for an SSD, but my current SSDs are all almost identical in weight to their counterparts. This one, however, feels light and plastic - no bad thing so long as I can understand the reason and know that it's not a significant build quality differentiation. I see no reason to pack an SSD in heavy protective material, so if that's the only difference then I'm in favour of the 'cheaper' consitency. I would have expected them all to be this light.

Ok dudes, problem solved.

 

Everything seemed to be in order, but of course I proved the SSD worked fine in another machine.

 

I did my research into the rather generic-sounding "Standard AHCI SATA Controller" and discovered it's hardware ID was that of the AMD SATA Controller.

 

I looked far and wide for an obscure HP driver pack which happened to include a .inf with the correct ID. It failed to install, but that was just because the installer was looking for other hardware I didn't have because it was meant for a different laptop, presumably.

 

I manually installed that single inf from the pack and after a reboot the system had re-identified the controller as a Storage Controller rather than an IDE/SATA controller, implying the driver/chipset supports RAID functinality that wouldn't ever be used.

 

With this replacement AHCI driver, which appeared differently in AS SSD (ahcix64 as oposed to msahci) I ran AS SSD and CrystalMark and now the 4kQD32 random writes are back up to 60+, from memory.

 

I might be mistaken but the system seems less glitchy, now. I'm happy, it's happy, but for some reason the HP recovery disks don't install a proper driver for the sata controller, which leads me to wonder if their laptops EVER perform as AMD would have intended...

Hello. I own a Dell M5010 laptop, specs in my profile. I recently installed a Corsair Force 3 120GB SSD and my 4k QD32 write speed is extremely low, just like in the OP post, my speed is under 1 MB (CrystalDiskMark score). Everything else is fine.

 

The alignment reported in msinfo32 is 106,901,504, which should be correct. AS SSD reports that the "amd_sata" or "AMD Sata Controller" is ok. Should this be reported as AHCI? I have enable AHCI mode in the BIOS of my laptop.

 

What should I do? Thank you in advance.

Run a ATTO benchmark and post a screenshot.

Here you go. I'd also like to mention that I get some 30s-1min freezes randomly, once or twice per hour. Also, during the ATTO benchmark test, the SSD seemed like it froze (system became unresponsive) during the first tests, the ones with low speeds.

I restored a system and file backup on the SSD using EasyUS Todo Backup and then I repaired the bootmgr with the Windows 7 CD. I'm running Windows 7 x64 SP1.

Unfortunately it is the only thing that I cannot try. I cannot install a fresh copy of the OS because I lost my windows key and I cannot activate my system afterwards. I tried to find my key using magic bean finder, produkey finder and a couple of other utilities but none of them recognize my key or they just show a generic BBBBBBBB key.

 

You can install and use the O. S. with out entering the PID# for a few days just to test the system.

What do you mean by that? Without? Any other suggestions, please?

 

The drive is new, the only thing that I did was to restore a system image onto it.

This might work.

http://www.joshcellsoftwares.com/2011/09/AdvancedTokensManager.html

So I secure-erased the SSD and I installed Win7 x64 from scratch. These are the new ATTO results. Improved but still not good, for the first values.

 

The SSD is connected to a SATA II interface, as you might have guessed from the results.

 

I want to mention that the SSD froze once during the first 5 minutes of booting the OS for the first time. What can I do to fix the SSD?

Plug the SSD into a different computer and run a ATTO benchmark.

I dont have any other computer..

Ask your friends and see if they will let you test your drive.

I live in a foreign country and my options are limited. None of my friends are willing to open their laptops because they will lose the warranty. Could the interface be the problem? I have an eSATA/USB external enclosure, should I run an ATTO test with the SSD connected to eSATA or USB?

Ok so I plugged the SSD in the esata interface (sata I i believe). These are the new results: first values are significantly improved but the limitations brought by the sata I interface are noticeable. I'm sorry but I couldnt use paint cause my laptop is stripped down of everything, just minimum functionality, so I saved in a word doc.

 

What should I do? This means that my internal sata II interface is faulty? If so, should I keep the SSD on the eSATA interface? Are the lower maximum speeds a serious disadvantage over the speeds that i would get from keeping the ssd attached to the sata II interface?

 

 

LE: here's the screenshot of the ATTO benchmark with the SSD connected to my eSATA I interface

 

LE2: I disabled the Direct I/O and Overlapped I/O features from ATTO and the results are tremendously improved. What does this mean? Where does the problem lie?