Intel 80GB SSD G2 performance problem?

[Apollo 11]

I have read many threads here about the new Intel SSD G2 and people are getting about 22-25MB/s read and about 60MB/s write in 4k random tests in CDM (CrystalDiskMark 2.2).

However in CDM (100MB test) I get only about 17,5MB/s in read and about 33MB/s in write, the sequential transfer rates are inline with what others are getting. In AS SSD Benchmark 1.2 I get 16/28 read/write in 4K and 121/54 read/write in 4K-64Thrd.

I changed the hard drive only couple of days ago, made an image of the old drive and restored that to the new Intel SSD. Since then I have installed only couple of programs, and not done any heavy benchmarking. I have quite optimized system (some services turned off, defrag off, boot time defrag off, no pagefile etc.) and almost 60GB of free space still on the drive.

I have tested with SuperFetch on/off, and Avira Antivir Guard on/off with similar results. I even aligned the partition to 64KB (it was at 1024KB) and restored the image again, and of course this had no effect .

So my CDM scores are about 25%/45% read/write lower then what others are getting. Any ideas why such a big difference? Defective SSD?

System specs:
HP EliteBook 8730W, 4GB, Vista Business 64-bit SP2.
Intel Mobile PM45 Express Chipset ICH9M-Enhanced.
AHCI set in BIOS, latest Intel Chipset drivers, latest Matrix Storage Manager.
Write caching and advanced performance enabled in Windows.

 

[jimhsu]

The Random read is a bit low, but the write score is consistent with my steady state benchmarks:

"New":

"Steady state":

PCPer already mentioned that some fragmentation would be result of more aggressive write combining algorithms (that's why random writes got improved so much). TRIM should fix it when it comes out.

I may also suspect your notebook chipset. You mentioned you had updated drivers, though.

 

[astrallite]

Intel Mobile PM45 Express Chipset ICH9M-Enhanced.

You need ICHR10 (X58 chipset IIRC) in order to get full SATA II output.

Jimhsu -- looks like you G2 owners really need the TRIM update because your steady state performance drops off quite a bit. The G1 drives with the new firmware keep ~90% of their "new-drive" write performance at steady state

 

[GotNoRice]

You need ICHR10 (X58 chipset IIRC) in order to get full SATA II output.

Jimhsu -- looks like you G2 owners really need the TRIM update because your steady state performance drops off quite a bit. The G1 drives with the new firmware keep ~90% of their "new-drive" write performance at steady state

I checked and I had 34.6 gigs free after doing a fresh Windows 7 install a few days ago. I copied 30 gigs of media files onto the drive, deleted them, then copied another 30 gigs of different media files onto the drive. I'm assuming at that point every bit of flash on the drive has been filled at least once. This is also on an ICH10R, on my Asus P5Q-Pro (P45).

Diskmark right after fresh Windows7 Install:

After deliberately filling up the free space on my X25-M G2 multiple times:

 

[Apollo 11]

You need ICHR10 (X58 chipset IIRC) in order to get full SATA II output.

Really? The ICH9M-E is quite new, and Intel says it has SATA 3Gb/s support. And as I wrote before, the sequential transfer rates are inline with what others are getting, aroung 250/85MB/s.

Jimhsu -- looks like you G2 owners really need the TRIM update because your steady state performance drops off quite a bit. The G1 drives with the new firmware keep ~90% of their "new-drive" write performance at steady state

Actually I have read that the Intel SSD G2 drives' performance degradation should be quite low, although this site only tested sequential performance.

 

[jimhsu]

Really? The ICH9M-E is quite new, and Intel says it has SATA 3Gb/s support. And as I wrote before, the sequential transfer rates are inline with what others are getting, aroung 250/85MB/s.



Actually I have read that the Intel SSD G2 drives' performance degradation should be quite low, although this site only tested sequential performance.

They tested only seq write degradation. I am searching for 4K random write stats but can't find any.

I have ICH9R, on an Asus P5E-VM MicroATX board. I'll take a look at chipset drivers to see if I can lift the random write also.

 

[astrallite]

At least with Indilinix drives, random read/write is typically about ~1/3 lower on ICHR9 vs ICHR10. This is pretty consistent with multiple user experiences on the OCZ forum.

 

[Apollo 11]

The Random read is a bit low, but the write score is consistent with my steady state benchmarks:

"New":

"Steady state":

Yes, but I have used the drive only couple of days. Only normal use, not much bechmarking or heavy writing. So really shouldn't be at "steady state" yet .

Could you give me your AS SSD Bechmark results for 4K and 4K-64Thrd in IOPS?

I have about 4200/6700 in 4K and about 34000/14000 in 4K-64Thrd. These results look as should according to Intel.

 

[Apollo 11]

At least with Indilinix drives, random read/write is typically about ~1/3 lower on ICHR9 vs ICHR10. This is pretty consistent with multiple user experiences on the OCZ forum.

Ok, that would explain the difference almost completely then. Thanks for the info!

 

[jimhsu]

Interesting that SSDs finally expose on-board controller inadequacies. I never thought of the controller before and only assumed it was important when considering RAID setups. But now storage has been turned on its head and the controller is to blame ... interesting times.

 

[Apollo 11]

Interesting that SSDs finally expose on-board controller inadequacies. I never thought of the controller before and only assumed it was important when considering RAID setups. But now storage has been turned on its head and the controller is to blame ... interesting times.

Out of curiosity I decided to bechmark in SafeMode. To my suprise CDM scored 22,6 / 60,7 MB/s read / write and AS SSD 22,9 / 41,8 MB/s in 4K and 144,1 / 54,7 MB/s in 4K-64Thrd. So these results are inline what everyone else is getting, be it ICH9 or ICH10.

It seems, at least in my case, that maybe the ICH9 is not to blame. There seems to be some software, driver etc, slowdown.

Has anyone any ideas? I would like to get the random operations performance to "normal" levels.

 

[jimhsu]

Very interesting... I noticed this also. Something I installed caused write speeds to suddenly drop. I thought it was steady state but it might not yet be.

 

[Apollo 11]

I updated my laptop BIOS to latest version, and now Safe Mode vs. normal mode bechmarks are quite close.

CDM and AS SSD results in 4K random reads / writes are still about 10-15% slower then "optimal new state" but I think that could just be "steady state" setting in because I have been transferring lots of big files to the drive.

As soon as the new Intel firmware is released, I will flash my drive and test again. That firmware update shouldn't have anything to do with performance, but you never know .

 

[Apollo 11]

The new firmware for Intel SSD G2 drives is out.

I installed the new update and obviously it didn't have any effect on the performance.

 

[bunta]

Im obviously in desperate need of help - posting for help everywhere!!. Im actually happy i found another person with the SAME EXACT problem as i do.

Apollo 11, did you ever find a solution??

 

[Burner27]

Deleted

 

[PHiZ]

Any idea what is wrong with my drive? i am using a Patriot Torqx 128GB with fw1571

Well, judging by your competency in placing your post in a relevant thread, I'd say your SSD problem is that you probably have installed it into your toaster.

 

[Burner27]

Well, judging by your competency in placing your post in a relevant thread, I'd say your SSD problem is that you probably have installed it into your toaster.

Your sarcasm is noted, but not appreciated. I was just asking if I have an issue with my drive.

 

[C-rizzle]

My X25-M 80Gb G2 is on an ICH9R, and I've got it in IDE mode and my #'s look very close to people's scores on an ICH10 with AHCI.

see: http://www.crowdcontrolusa.com/overclock/80gb-ssd.gif

 

[zeroyon]

I just got a new 80GB X25-M G2, and I realized that at the "10% position" on the media, there is a MASSIVE performance dropoff. I've only installed windows 7 and a few apps on it so far.



Is this normal? Can anyone else with the same drive please check to see if it is similar?
Thanks.

 

[PHiZ]

I just got a new 80GB X25-M G2, and I realized that at the "10% position" on the media, there is a MASSIVE performance dropoff. I've only installed windows 7 and a few apps on it so far.

Is this normal? Can anyone else with the same drive please check to see if it is similar?
Thanks.

What is that supposed to be measuring??? 10% position??? If that test was designed for rotating HDDs, then the results may not even be applicable to SSD...

 

[TehQuick]

probably meant "10% filled"...

 

[Apollo 11]

Im obviously in desperate need of help - posting for help everywhere!!. Im actually happy i found another person with the SAME EXACT problem as i do.

Apollo 11, did you ever find a solution??

Unfortunately not. I have been using Windows 7 RTM now six weeks and don't use Vista anymore. I re-tested with AS Disk Benchmark, and my random 4K and 4K64-Thrd results are down quite a bit.

Six weeks ago / today (in MB/s):

4K: read 20,7 / 16,4 and write 30,2 / 22,4
4K64-Thrd: read 138,2 / 127,5 and write 42,5 / 37,3

I have read that the new G2 really shoudn't drop so much in performance (and the starting results after Win 7 install were already "low"ish). Actually G2 should keep it's performance level quite near new levels.

But the good news is that the drive feels as fast as new, and I am not experiencing any performance problems. Granted, I use my computer mainly for light workloads. But I hope that the upcoming firmware update for G2 that enables TRIM and new Intel Matrix Storage drivers will eventually fix this problem.

(BTW, I have three partitions on my G2, one for Win7, one for Vista and one data partition. I always test the OS-partition. I have latest Intel beta Matrix Storage drivers installed and I am currently using Windows 7 Pro 64 bit).

 

[bunta]

My X25-M 80Gb G2 is on an ICH9R, and I've got it in IDE mode and my #'s look very close to people's scores on an ICH10 with AHCI.

see: http://www.crowdcontrolusa.com/overclock/80gb-ssd.gif

C-rizzle, Im assuming this is your desktop correct?

How come youre using IDE mode rather than AHCI?

I was just told that the problem could just be on "mobile chipsets" which attains to notebooks/laptops...

 

[bunta]

You need ICHR10 (X58 chipset IIRC) in order to get full SATA II output.

Jimhsu -- looks like you G2 owners really need the TRIM update because your steady state performance drops off quite a bit. The G1 drives with the new firmware keep ~90% of their "new-drive" write performance at steady state

Astrallite, what are you laptop's benchmarks like?? Have you tried using the X-25M in one of the laptops and see what the benchmarks are like?

 

[zeroyon]

What is that supposed to be measuring??? 10% position??? If that test was designed for rotating HDDs, then the results may not even be applicable to SSD...

I'm not sure exactly what it means, maybe it means that certain block location on the SSD? Considering that there are 10 NAND flash chips on the SSD, and that my drive performance is poor right up until 10%, maybe one of the NAND flash chips in my SSD is bad?

I ran HDTune and it showed similar performance as to what I got with sisoft sandra:


Compare it to the HDTune test on this site:
http://www.legitreviews.com/article/1022/5/

Should I return my SSD? Or is this expected of them after you start writing data onto it?

 

[bunta]

^^^^

To tell you the truth, i think thats normal. Ive seen plenty of HDTune benchmarks with that dip.. The one on legitreviews is also a 160gb version which will perform a little better than your 80gb - which has nothing to do with the dip but im jus sayin... Google Hdtune images and X-25m, youll see what i mean.

Now do me a favor, and download

http://release.crystaldew.info/CrystalDiskMarkSetup

and bench it.

 

[zeroyon]

^^^
Well if it is normal that is good to know, I just found it odd that performance improved drastically exactly when it reached "10%", which I assume means it started reading from another flash chip.

Here are my CrystalDisk scores:



I seem to have an oddball SSD. My write speeds are better than most, my 512K read speeds are horrid, and that low speed during the first "10 percent" still worries me....

 

[bunta]

^^^
Well if it is normal that is good to know, I just found it odd that performance improved drastically exactly when it reached "10%", which I assume means it started reading from another flash chip.

Here are my CrystalDisk scores:



I seem to have an oddball SSD. My write speeds are better than most, my 512K read speeds are horrid, and that low speed during the first "10 percent" still worries me....

Zeroyon,
thank you for the favor.
One more question.. Is the CrystalDisk benchmark from a desktop or laptop? If its a laptop, what brand/model?

Your write speeds are in fact better than most...Its the highest ive seen for the 80gb version without RAID.

 

[zeroyon]

^^^
On second thought, after looking through a lot of benches, I'm not so sure that the performance of my drive is "normal". Most benches I have seen have steady performance across the drive (such as http://news.driversdown.com/news/Hardware/200907/28-13928_7.html ).

I guess it is good that my drive is faster than most in the write category (or so it seems so far), but having something that is drastically different than others usually means it isn't working as it is supposed to, and therefore (probably) faulty...

The test is from a desktop. Here are some quick specs:
CPU: Core i7-920 @ stock
MOBO: Asus P6T
RAM: 3x2GB OCZ Gold 1600MHz
GPU: Nvidia 260GTX 216

To add, here is my bench at 50MB instead of 100MB:

and 500MB:

 

[a123456]

^^^
On second thought, after looking through a lot of benches, I'm not so sure that the performance of my drive is "normal". Most benches I have seen have steady performance across the drive (such as http://news.driversdown.com/news/Hardware/200907/28-13928_7.html ).

I guess it is good that my drive is faster than most in the write category (or so it seems so far), but having something that is drastically different than others usually means it isn't working as it is supposed to, and therefore faulty...

The test is from a desktop. Here are some quick specs:
CPU: Core i7-920 @ stock
MOBO: Asus P6T
RAM: 3x2GB OCZ Gold 1600MHz
GPU: Nvidia 260GTX 216

Did you try sticking your SSD in another computer so it's not the C drive, which would be the Windows system drive and possibly the page file drive that would have massive interrupts? Granted, it probably shouldn't be stuttering at the same spot each time but who knows what HD Tune does?

I get the same general speeds as the OP with an Intel G2 with 4k read/write. As in 20-30M/sec. It's with an X58 and Win 7. I wouldn't be surprised if one of these is causing a problem but I'm too lazy to move the SSD to another computer to perform better tests since I'm happy with my new real-world performance and don't feel the need to do other fixes like flash my BIOS for what would essentially be benchmarking gains. The Indilinx drives had pretty much the exact same load times as the Intel drives for apps even though their random reads were only in the 30M/s region. However, if you feel your benchmarks aren't up to par, that's what the 3 year warranty is for.

The full format or erasing of the SSD that I did might kill some performance since TRIM is not implemented on the Intel SSDs yet. Anandtech's review showed anywhere from .2% to 15% losses.

 

[zeroyon]

^^^
I don't have another computer to test it on, but I booted vista using a WD1TB drive on the same computer, and I get almost the exact same results there (numbers are a bit different but it still takes a massive hit during the first "10 percent"). I also disabled the pagefile.sys and hyberfile.sys, and I get the same results either way.

How do you actually know that windows 7 has detected your SSD as an SSD? Just simply by your drive working? Also, how will you actually know if it is using TRIM (once it is implemented). Is there something that displays or shows that in the OS?
These things have probably been covered somewhere before, sorry...

Also to note: the version of windows 7 I am using is not the RC, it is the "Windows 7 Professional Full" 64bit version (build 7600). I got it for free through the Microsoft ELMS and my university. That might be why I am seeing higher write speeds, did MS improve SSD performance in this version of windows 7 over the RC versions?

 

[C-rizzle]

C-rizzle, Im assuming this is your desktop correct?

How come youre using IDE mode rather than AHCI?

I was just told that the problem could just be on "mobile chipsets" which attains to notebooks/laptops...

Yeah, desktop. P5K-E.

I couldn't get Windows to install in AHCI mode. I don't think my SATA DVD drive likes AHCI mode ? Either that or maybe b/c the other secondary (data) drive in the system was formatted in IDE mode?

 

[a123456]

^^^
I don't have another computer to test it on, but I booted vista using a WD1TB drive on the same computer, and I get almost the exact same results there (numbers are a bit different but it still takes a massive hit during the first "10 percent"). I also disabled the pagefile.sys and hyberfile.sys, and I get the same results either way.

How do you actually know that windows 7 has detected your SSD as an SSD? Just simply by your drive working? Also, how will you actually know if it is using TRIM (once it is implemented). Is there something that displays or shows that in the OS?
These things have probably been covered somewhere before, sorry...

Also to note: the version of windows 7 I am using is not the RC, it is the "Windows 7 Professional Full" 64bit version (build 7600). I got it for free through the Microsoft ELMS and my university. That might be why I am seeing higher write speeds, did MS improve SSD performance in this version of windows 7 over the RC versions?

I don't know. If you think your drive is broken, then RMA it. It's not that big a deal. It is your drive and your customer satisfaction after all. Intel has customer support numbers that you can call up, although I doubt the tech support guy knows anything about computers. They usually just have a checklist of stuff to ask you (if at all) before starting an RMA.

This reminds me of the monitor section where people return stuff over slight uniformity issues while others are happy as a pig in mud over cheap TN panels.

I have no idea how Intel will finally implement TRIM. On other drives, you can actually manually run it as a utility in addition to what the OS normally runs in its day-to-day operations.

I'm running the RTM also, so we have the same version.

 

[Old Hippie]

I couldn't get Windows to install in AHCI mode.

Some of those P35 chipsets didn't like SATA optical drives on a fresh install.

My Abit IP35-Pro was the same way and a clean install required an IDE drive.

 

[jimhsu]

It shouldn't have anything to do with physical position. If it's getting stuck at a particular percentage, it may have to do with time. Intel SSDs dynamically adjust read/write workloads to accommodate changes between a heavy seq write workload and a heavy random write workload. It needs to do this because it uses write-combining, which increases random writes a lot (why Intel leads this category) at the expense of fragmentation. Therefore, the "dip" at the beginning is (most likely) the SSD switching from a random-optimized configuration to a seq-optimized configuration.

I have a huge amount of information on this, so I'll not post it all. Here's some for starters:

SSDs all have what is known as an “Indirection System” – aka an LBA allocation table (similar to an OS file allocation table). LBAs are not typically stored in the same physical location each time they are written. If you write LBA 0, it may go to physical location 0, but if you write it again later, it may go to physical location 50, or 8.567 million, or wherever. Because of this, all SSDs performance will vary over time and settle to some steady state value. Our SSD dynamically adjusts to the incoming workload to get the optimum performance for the workload. This takes time. Other lower performing SSDs take less time as they have less complicated systems. HDDs take no time at all because their systems are fixed logical to physical systems, so their performance is immediately deterministic for any workload IOMeter throws at them.

The Intel ® Performance MLC SSD is architected to provide the optimal user experience for client PC applications, however, the performance SSD will adapt and optimize the SSD’s data location tables to obtain the best performance for any specific workload. This is done to provide the ultimate in a user experience, however provides occasional challenges in obtaining consistent benchmark testing results when changing from one specific benchmark to another, or in benchmark tests not running with sufficient time to allow stabilization. If any benchmark is run for sufficient time, the benchmark scores will eventually approach a steady state value, however, the time to reach such a steady state is heavily dependant on the previous usage case. Specifically, highly random heavy write workloads or periodic hot spot heavy write workloads (which appear random to the SSD) will condition the SSD into a state which is uncharacteristic of a client PC usage, and require longer usages in characteristic workloads before adapting to provide the expected performance.

When following a benchmark test or IOMeter workload that has put the drive into this state which is uncharacteristic of client usage, it will take significant usage time under the new workload conditions for the drive to adapt to the new workload, and therefore provide inconsistent (and likely low) benchmark results for that and possibly subsequent tests, and can occasionally cause extremely long latencies. The old HDD concept of defragmentation applies but in new ways. Standard windows defragmentation tools will not work.

SSD devices are not aware of the files written within, but are rather only aware of the Logical Block Addresses (LBAs) which contain valid data. Once data is written to a Logical Block Address (LBA), the SSD must now treat that data as valid user content and never throw it away, even after the host “deletes” the associated file. Today, there is no ATA protocol available to tell the SSDs that the LBAs from deleted files are no longer valid data. This fact, coupled with highly random write testing, leaves the drive in an extremely fragmented state which is optimized to provide the best performance possible for that random workload. Unfortunately, this state will not immediately result in characteristic user performance in client benchmarks such as PCMark Vantage, etc. without significant usage (writing) in typical client applications allowing the drive to adapt (defragment) back to a typical client usage condition.

SSDs have a performance profile that is difficult to assess with traditional benchmarking tools. In essence, it is time-variant. Hence, benchmarks designed for hard drives will often give sub-optimal, confusing, or inconsistent results. This is normal. You can also see the problem in running the SSD on an OS partition that is currently booted -- while the drive tries to adapt to the huge concurrent random workload that something like CDM puts out, it is also being interrupted with hundreds of other requests per second -- Windows, background apps, internet downloads, even tracking where the mouse cursor is.
The "best" benchmark really is to run a prolonged workload on your SSD with the app that you care about. E.g. loading times, rendering/exporting, database, etc.

In case anyone cares, I keep an Evernote public notebook with this kind of stuff:
http://www.evernote.com/pub/jimhsu/StorageSSDs

 

[jimhsu]

If you need a technical reference, please see here:

http://www.cs.wisc.edu/adsl/Publications/ssd-usenix08.pdf