Incredibly slow Samsung 830 SSD in Dell XPS8300

[withanee67]

I think I must win the prize for the slowest ever Samsung 830 SSD – see the benches below (<100MB/s!). Can anyone suggest what might be going wrong?

Here’s the problem:

I have just replaced the OS HDD in my XPS8300 with a brand new Samsung 830 SSD 128GB (more details of the setup and how I migrated are given below). The new SSD works, but its performance is WORSE than the HDD in almost all respects:

■Windows startup: the SSD takes 65s vs 30s for the old HDD.
■Program launching is slower, eg 6s to launch Word, and some other programs have taken 15s or more to appear.
■The benchmarks are appalling, and contradictory:

CrystalDiskMark bench is the slowest I’ve seen anywhere; even others complaining of poor performance are getting >200MB/s sustained; I’m getting <100! It’s no better than I had before with the HDD (for comparison below)

Measured with AS SSD (below), the sustained write performance appears to be better (contradicting the CDM mark), but the 4k and access time benchmarks look dreadful to me. In fact the 4k test took so long to run (30 minutes or so) that I didn’t run the 4k-64Thrd test.

Something is clearly badly wrong – not just with the benchmark results but with the real-world performance too. Can anyone suggest what?

System setup

To answer the obvious questions, here’s my setup:
■ Dell XPS8300, about 12 months old running Win 7 Pro 64 bit
■ Core i7 3.4GHz, 8GB Ram
■ Originally had 2 x 1TB Hitachi disks, one as C: (OS), the 2ndas a D: Data drive. I replaced the first disk with the SSD (see below for how I did it).
■Intel Series 6 motherboard (if I have that right), with 2 x SATA-III
■(6Gb/s) ports (0 & 1) and 2 x SATA-II (3Gb/s) ports (2 & 3). The HDD’s were on the 6Gb/s ports, numbers 0 & 1. The SSD is now on port 0 (ie 6Gb/s)
■ The SATA controller reports as “Intel Desktop/Workstation/Server Express Chipset SATA AHCI Controller”.
■ Set to AHCI mode (and always has been)
■ Running Intel Rapid Storage Technology (in SATA mode, I don't have RAID)

Drivers are up-to-date according to Windows, & device manager shows no conflicts. Samsung firmware is the latest.

Migration process

Again to answer any obvious questions in advance, here’s how I did the migration from HDD to SSD:

1. Cleaned up the OS HDD to save space/eliminate bloat (backed-up & deleted the Recovery Partition after moving the bootmgr from there to the C: (OS) partition using Easy BCD; deleted unnecessary recovery points etc).

2. Installed the SSD temporarily on SATA port 3 (SATA-II, 3Gb/s).

3. Cloned the C: partition from the HDD to the SSD using Macrium Reflect 5.0, using “intelligent sector copy” (copies only the used sectors, apparently) and set the SSD as Active

4. There was also a 2GB Dell OEM partition on the HDD which I did NOT clone across to the SSD – after reading up on this partition I concluded it wasn’t necessary…

5. Disconnected the old C: HDD and reconnected the SSD in its place (on port 0, SATA-III 6Gb/s)

6. Checked BIOS set to boot from the Samsung

7. Rebooted OK, but very slow as described

8. Ran WEI, giving disk score of 7.8 (woo-hoo! If only the SSD didn’t actually go at a snail’s pace!)

9. Ran Samsung’s optimisation tools, and set 11.9GB overprovisioning as recommended, all to no effect. Still as slow as ever.

10. On advice from elsewhere, tried to check the drive is running in UDMA5 mode, not PIO. But Device Mgr doesn't list any ATA channels, only the Intel controller mentioned in my system description, and the properties of that don't mention transfer mode.

11. Replaced the Samsung with the old HDD, which is performing as normal.

Some thoughts of mine:

1. Is the Samsung 830 simply faulty?
2. Have I messed up by not cloning the Dell OEM partition to the SSD? Can’t see why this should be as I understood it only contains Dell utilities/tools.
3. Does the difference in the benchmarks using CrystalDisk and AS SSD indicate the source of the problem?
4. How can I check that I'm not running in restricted PIO mode (ref step 10 I tried above)

Grateful for any help that any of you knowledgeable souls can offer, please! Thanks.

 

[RealBeast]

Since you cloned, your SSD alignment could be off, which could cause poor performance. Check it by opening system information (type msinfo32 in the start button run box). Look in components/storage/disk, and for the SSD note the Partition Starting Offset -- divide it by 4096. If the answer is not a whole integer, then you have an alignment problem.

For example my OS SSD has two partitions, the SRP and the C drive and they have Partition Starting Offset values of 1048576 and 105906176, dividing each by 4096 gives me 256 and 25856.

If alignment is your issue, the easiest thing to do is a clean re-installation after a secure erase, but you can also attempt to fix the alignment with parted magic if you are familiar with using it.

 

[bucknutty]

The win 7 install will start a mechanical drive after 63 empty blocks, while SSDs require 64 blocks of data for optimal performance.
As Real beast stated the if by checking msinfo you dont have a whole number then your partition is not in the ideal place for your ssd. I have seen tools to fix this that call for running ubuntu live CD and repositoning the partition but most of them require a repair install of windows. I only tried it once and it failed to boot.

If the alignement is off your best bet is to do a clean insall of windows 7 and at the part where windows asks you what disk to use, hit advaced and delete the partition. Then let the windows installer make you a new partition.

Of course this means loosing all your data and such so you need a good plan for saving all that stuff before you begin.

 

[RetiredChief]

Realbeast, His snapshot of AS SSD indicates partition alignment is OK
Not sure if it just checks the OS partition, but then Do not know what other partitions are on SSD - should have the tiny 100 MB partition).

I only recommend Migrating (Never cloning) a HDD -> SSD in rare cases. If Posible installation should be done as a CLEAN install with ALL other HDDs disconnected.
The SSD should show TWO partitions, a small 100 MB system partition and the 2nd partition as the remainder of the drive (OS). Have a feeling that the cloning which is a sector-by-sector copy. As stated I never recommend cloning a HDD to a SSD. Migration is exceptable if a clean install is NOT an option (ie My wife's system). For migration, recommend EasyUS or EZ-Gig. Have used both. EasyUS is a freebee, EZ Gig (which I prefer) while the software is free, it requires a 10->15 buck USB->Sata ( and NO can not use any old one).

 

[withanee67]

All above, thanks for your responses. I need to check my forum settings as I didn't receive any notifications of your replies...

RetiredChief, yes I had asumed that the AS SSD report indicated no issue with partition alignment, and following Realeast's advice confirms that.

RetiredChief, I only have a single partition on the SSD, the OS partition. I don't have the 100MB system partition; there wasn't one on the source HDD drive. Being a Dell, the primary HDD was configured with a 2GB Dell"OEM Partition", a 12GB Recovery partition and an OS partition taking up the rest of the drive.

As mentioned in my original post, I deleted the Recovery partition from the HDD before cloning, and only cloned the OS partion, leaving the Dell OEM partition behind. I understood this only contains Dell utilities and tools for their tch support, which I have no intentio of using.

Btw, I also undertood from reading elsewhere that the 100MB system partition present in a standard Win7 install is only needed for future implementation of bitlocker, and can be safely done without. Anyway, it's a moot point, as I have never had a system partition on my original primary drive, so can't clone it anyway.

I want to avoid doing a clean install - it's such a pain to reinstall all the drivers and programs and get them configured how they were before.

I think my next step is to do a sector by sector clone of the HDD, including both the Dell OEM partition and the OS partition, ie a complete image of the current HDD which is performing well. If that doesn't work, I'll check out EasyUS.

I'll report back.

 

[RetiredChief]

Try this, if dead set against a "Clean" install.
Down load EasyUS (use a Sata to usb adaptor)and do a Migration. DO NOT think a sector by sector clone is good idea.
The two programs that I've had good luck with is EasyUS and EZ Gig. Both are free, But the EZ Gig (which I prefer) requires a speacial usb->sata Adaptor.

With all the benchmarks that you have run, you may need to do a secure erease.
Repeated benchmarks with out suffient time for trim and CG to restore the drive will result in POOR benchmarks.

EasyUS (use ther Free version): http://www.easeus.com/?gclid=CJPBw-_u77ICFeaDQgod0GIAcw

EZ Gig w/software and adaptor: http://www.newegg.com/Product/Product.aspx?Item=N82E16812161004

 

[withanee67]

Here's an update:

Rather than contnue fiddling around, I thought I would see how the drive performed with a clean install, to rule in/out any issues with cloning or migration. It made no difference; the drive performance just was just as bad as before.

There was a minor improvement in start-up time, but only down to 45 seconds from a minute. (This was after letting Windows settle down with installing updates etc, so was a fair test.) CrystalDiskMark results were almost identical to what I have posted previously.

On RetirdChief's advice, I performed a SecureErase first, before I did the clean install.

I have concluded that the drive is faulty and have RMA'd it.

I'll let you know how I get on with the replacement!

 

[eXistenZ]

I had the same problem when I first got mine.

It seems I never deleted things enough for the TRIM and garbage collection to do its work. Now, I shut off the Anti-V, delete a gratuitous file on C:, and go for a jog, letting the computer sit at idle. And no background downloading either. Checked Resource Monitor>Disk to see that everything is idle.

That seems to improve things, but still I only occasionally get the 30k random writes speeds.

This thing with an access time of 2.+ sec is crazy and I got the same thing -- at first. This resolved the problem for me.

 

[withanee67]

Hi eXistenZ,

Interesting tip, thnks. I don't think it would have helped my situation, when a full SecureErase did nothing to improve speeds, but I'll bear it mind if and when I get my replacment working.

 

[withanee67]

I have success with the replacement drive...

I RMA'd the old slow Samsung 830 SDD and got a fresh one as replacement. Installed exactly the same as before, and again used Macrium Reflect to clone the C: (OS) and Dell OEM partitions across from the old HDD to the new Samsung. I used the "intelligent sector copy" option as before.

Nothing different about the installation from before, but the results couldn't be more different; here's the latest benchmark:

I think that counts as success. Note I didn't have to do anything clever; the partitions aligned themselves properly during cloning, and I just ran Windows Experience Index to allow Windows to detect and adjust itself to the SSD. I ran the "OS Optimisation" tweaks in Samsung Magician too for good measure.

Interestingly, real-world performance is better than with the HDD, but not mind-blowingly better. Applications such as Word etc open instantly, and startup is down to 20s or so after the BIOS does its stuff (a little hard to measure accurately because I have multiple user accounts and so get the select user screen halfway through start-up). Plenty fast enough, but not quite the 7 or 8s you see with eg a new Ultrabook. No doubt they have lots of other OS optimisations and a minimum of drivers and startup processes in order to get that speed.

I'll have a look at what bloat I can eliminate from the startup, but even if there's nothing, I'm satified with the performance I'm getting now.

Thanks again those of you who took the trouble to offer your thoughts along the way. It seems the drive was at fault all along.

 

[RetiredChief]

Quick question: I think with Crystal Diskmark you can select the type of data to use when running the Benchmark, either highly compressable data (simular to ATTO) or Uncompressable data (simular to AS SSD)
I quit using Crystal Diskmark in favor of AS SSD, and NEVER did use ATTO as it is irrelavent except to compare with manuf advertized performance.

On Boot times, Both my 830 (i5-2500k) and my M4 (i5-2410m) take about 12-15 Sec from "start loading OS" to opening first program (Single user so does not stop). Yes The laptop would boot to windows 8 in about 8 Sec - the ONLY thing I liked about windows 8.

On Program load times, Hard to judge as on my OLD i5-750 using an OLD SATA II SSD (pheonix Pro, simulart to Vertex 2) on a SATA II port. If I clicked on a excel spreadsheet recent link, both the program loaded and called up the spreadsheet in a blink of an eye (Spreadsheet was on a 2nd OLD SSD).

Anway, for what it's worth here is a AS SSD screenshot of my M4. While the 830 has better benchmarks than the M4, There is NO real life difference (As indicated have both), But the M4 has lower power consumpution - Have two installed in laptop. NOTE: The new Samsung 840 soulds like it will be an OUSTANDING choice for laptops as they have cut the power consumption considerably while also improving performance (not sure if it will be notable in real life).

M4 laptop (i5-2410m, 840m dGPU, 8 gigs ram, Blu-ray ROM/DVDRW drive 17 in)

NOTE: this was shortly after I bought it (FW = 0009). I have update the firmware to latest. And I believe that iaSTor was the older ver 9.6 which has also been updated to 10.x

 

[withanee67]

Quick question: I think with Crystal Diskmark you can select the type of data to use when running the Benchmark, either highly compressable data (simular to ATTO) or Uncompressable data (simular to AS SSD)

No, I don't think that's true of CDM. You can choose random data, all 0's, or all 1's. I ran the test on random. From wht I have read, I believe the CDM data is uncompressible, so it gives reliable results unlike ATTO which gives optimistic results with Sandforce controllers. Whatever, I take all the benchmark results with a small pinch of salt, and don't expect them to have a huge degree of accuracy; I only look for gross differences in results over time or versus other tests with the same SSD, wouldn't ever try to compare results produced by different tools.

For what it's worth, here are my results with AS SSD, taken just now:

Seems OK to me. It's a little odd perhaps that a couple of the read scores are slower than the equivalent writes, but I'm not going to worry about it - probably a glitch when running the bench, or something else reading in the background.

I saw the reports on the 840 - looks interesting, but I'm done with SSD upgrades for now. I'll wait until this 830 is dying (hopefully a few years away!) and will see what looks best on the market then.

 

[RetiredChief]

Looks good to me, Also Have a 256 gig and a 128 gig 830 in my Desktop i5-2500k and your scores look very close to mine (Back of mind Overall score was around 730 WHICH is No real diff). And totally agree with you on benchmarks - While they may show differences for the various upper end SSDs, There is VERY LITTLE diff in real life. Only real value is if you think the performance has changed, makes a quick check to confirm.

Glad it's solved and ENJOY