HD Tune Results for Samsung EVO 840 250GB and 300GB Velociraptor. Normal?

GlacierFreeze

Golden Member
May 23, 2005
1,125
1
0
Here's some results from HD Tune. The results look kinda funky for my EVO. Anyone know what that could be? Bad cable? Cheap mobo? Or is that normal?

Samsung EVO 840 250GB
2hdtbpe.png



WDVeloci-Raptor 300GB
21ke82p.png


Rest of setup in sig.
 
Last edited:

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
Did you do a performance optimisation in Samsung's Magician before benchmarking?

It can make a difference with my 830 if I haven't done it in a while.
 

GlacierFreeze

Golden Member
May 23, 2005
1,125
1
0
Was the EVO empty when you benched it? What happens if you let the system idle for a good period then re-run the bench?

Nah, not empty. Had Windows installed. Letting it idle doesn't seem to make a difference.

That looks like your computer was busy doing other things while the bench was running.

?

Nah, was idle at the time.

Did you do a performance optimisation in Samsung's Magician before benchmarking?

It can make a difference with my 830 if I haven't done it in a while.

Yeah, think I have it set to Custom with mostly Performance settings.
 

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
Yeah, think I have it set to Custom with mostly Performance settings.

Not OS optimisation, Performance optimisation. It's like Trim, it clears out the unused data so the sectors are ready to write to without a performance hit.
 

GlacierFreeze

Golden Member
May 23, 2005
1,125
1
0
Here's some results from HD Tune. The results look kinda funky for my EVO. Anyone know what that could be? Bad cable? Cheap mobo? Or is that normal?

Samsung EVO 840 250GB
2hdtbpe.png

Decided to try a few things. Quoted is from earlier. And I"m assuming I used the same default 16k settings from HD Tune. I was poking around a bit but it was set to that when I ran it. Looks like a decrease...

Thought I may see if there's any newer Intel Rapid Storage Technology drivers. Uninstalled them, saw there weren't any. Decided to update BIOS to latest one on ASRock site. From 1.50 to 1.90 then reinstalled same IRST driver version. Jumpy lines seemed to be lower overall but AS SSD improved


After BIOS update and Reinstall of IRST drivers.
sndsmu.png



AS SSD before BIOS update and IRST reinstall
10n7sxf.jpg



AS SSD after BIOS update and IRST reinstall
21mguf8.jpg


Biggest thing I notice is a significant increase in 4K writes. About 40MB/s improvement. Also access time dropped in half on write, but some reason can't display the read access time.
 
Last edited:

Tristor

Senior member
Jul 25, 2007
314
0
71
You are also using an older version of HD Tune than the respondent to your post OP. Have you tried using the latest version of the HD Tune utility? They've changed stuff significant since version 4 was released specifically to deal with appropriately measuring performance from SSDs. With version 2.x, I imagine it may account for some of the differences.
 

Groenendiek

Junior Member
Jun 5, 2014
1
0
0
I have roughly the same graph.
It is obviously a windows problem. This is my scenario:
I bought two 840 120GB's. Put them in raid0, 16k slices. Everything went sweet and got speeds up to 1000MBps.
But I knew it wouldnt matter much with a single ssd. Figured I buy a 250G one, clone the raid array to it, I'd be good. Of course I'd had to set ahci back on. And I'd put one 120G ssd in my laptop, and sell the other.

Now after cloning I noticed fan speed fluctuations. CPU was getting hot, turbo got switched on and off. Investigation pointed to windows not being set to deal with ahci properly, and the intel drivers/matrix storage software prolly just eating up resources. So I removed intel matrix storage and rebooted. Still no change. After some more minor tinkering and trying (dont think any helped) and another reboot (yep that must have been it), things are looking up and cpu is running smooth again.

But still, that wonky graph from hdtune. It is still not good. I had ssd's before and hdtuned them, it's not right to have such violent spikes.

Since the only thing really different is how windows gets along with the drive in a nice anhi way versus the raid array, I should be able to find it and fix it. The sure fix would be to reinstall windows. But for me that's a last resort.
So I suppose I'd be looking for driver issues, windows settings for raid that need to be reverted, check put the ssd, optimise it and what else can I try?

I hope TS might recognize something here, and get new ideas to fix his problem.
 

GlacierFreeze

Golden Member
May 23, 2005
1,125
1
0
Yeah I was thinking it maybe a driver issue too. Uninstalled IRST and tried to reinstall/update MSAHCI but no change. Was even slower in AS SSD. Haven't tried reinstalling Windows 7 yet. Hate to have to redownload/install Elder Scrolls Online, takes hours and hours and that's after hours and hours of Win updates. Just putting up with it for now....
 

GlacierFreeze

Golden Member
May 23, 2005
1,125
1
0
Decided to try to reimage it then secure erase. Used Clonezilla and made image onto a secondary HDD. Used Magician's software to create bootable USB stick then secure erased it. Restored image to SSD. Results:

2q1sxav.png
 

TioDrakul

Junior Member
Aug 24, 2014
1
0
0
I have the exact same problem on a Samsung 840 EVO 250GB, very slow read speed where the SSD have data recorded and almost the same graph on HDTune. And what makes me worried is that graph is very similar to what I had when I tested one OCZ Agility 120GB, suggesting that the controller behaves similarly to a Sandforce garbage (Only shows the speed indicated by the manufacturer where they have no user data).

This makes the SDD useless, because what good is a SSD when he reads your data slower than a actual HDD?

kpPlyl1qj
Note: I've done all the pre-checks I should do. The AHCI is enabled, the cable is new, is connected to a SATA III port of Intel chipset, and have the latest Intel drivers installed. It is a problem of the SSD itself, whatever else might can have problems I already checked

The graph: http://imgur.com/N34IGWP (sorry, the shitty imgur do not accept a direct link)
 
Last edited:

redzo

Senior member
Nov 21, 2007
547
5
81
I get the same results on a 250 EVO. It's a system drive. The windows instance is running damn fine: no slowdowns whatsoever. I don't think that the drive is broken. It must be something with the way hdd tune does its benchmarking. Other SSD benchmarking tools do not seem to reflect the same crappy results. I also find it hard to believe that all of our EVO drives are broken.
 

skyhawk21

Junior Member
Sep 13, 2014
7
0
0
Well guys I am having the same dam problem. I have a new Gigabyte Z97-sli motherboard, intel 4690 cpu, using intel driver from gigabytes site. I have 2 samsung evo 840 ssds. One is my boot drive and apps for windows 7 on C:

The other is my games drive which is D drive

None of them are that full to be honest, they are running sata 3 mode with AHCI on. Just upgraded to this setup and on my old MOBO which was intel h55 chipset it did not have these problems but I do think its the SSD. Both of SSDs are showing the same results. The first things installed or sets of data in hdtune and hdtach read anywhere from 50-70mb a sec on graphs. When it gets passed that old data that has been sitting there for awhile, the speeds jump up to normal for uncompressed or free space areas. At the front of the disk the compressed data gets slower and slower with time.

Now someone said it does not hurt performance which is BS. I did a test. I copied BF3 game folder that is being read slow from GAME drive SSD to my main BOOT drive SSD. When copying that folder only which is the slow area, it was writing that folder to my main drive at 50mb a sec which is slow. normal write would be 200-300mb a sec which would be normal fast.

My graphs look exactly like these guys. When you use a benchmark that does sequential read of the compressed used data in the ssd it will show it as slow. using other benchmarks only does sequential read of a empty area on ssd which goes full speed. I have an older ADATA and KINGSTON SSD which do not do this at all. The samsung evo 840 just started showing this problem and I got two which were from best buy on sale now I see why! the area that reads slow in hdtune or hdtach actually copies that slow to or writes that slow.

anyone know of a fix? is it windows 7? intel Z97 chipset drivers? hardware problem in new chipset? because this problem did not happen on H55 intel chipset. intel IST driver?

HELP!

PS SEE ATTACHED pic of hdtune. the area that has been written to at 225 number line area was new data that was written today. the data at 50 line in graph is first data written or old stuff. anything above 225 is blank!

https://picasaweb.google.com/lh/photo/6qJmD3Opsh2Jl1O8v6pS6dMTjNZETYmyPJy0liipFm0?feat=directlink
6qJmD3Opsh2Jl1O8v6pS6dMTjNZETYmyPJy0liipFm0

6qJmD3Opsh2Jl1O8v6pS6dMTjNZETYmyPJy0liipFm0
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
When it gets passed that old data that has been sitting there for awhile, the speeds jump up to normal for uncompressed or free space areas. At the front of the disk the compressed data gets slower and slower with time.
The 840 Evos do not compress data. I'm wondering what this is as much as you are, but it's definitely nothing like that.

Do you have any other SATA 6Gbps PC you can hook it to, using other SATA chipsets, to see if it is the Intel chipset, or firmware?

The graph: http://imgur.com/N34IGWP (sorry, the shitty imgur do not accept a direct link)
Copy the image location, the put that in
.
N34IGWP.jpg
 

skyhawk21

Junior Member
Sep 13, 2014
7
0
0
Yes I did say I tried this before in another system. I did try it on different sata 3 (6) ports and still the same issue in the areas on both ssds I have where I stored data for more than 60 days now. Any new data does not have problem its only when its 60 days or more sitting on ssd and well the TLC chips do not like holding data that long or longer. There are message forums and threads @:

http://www.overclock.net/t/1507897/...ed-drops-on-old-written-data-in-the-drive/200

http://www.reddit.com/r/buildapc/co...e_samsung_840_evo_is_notorious_for/?limit=500

http://forums.overclockers.co.uk/showthread.php?t=18615995
 

skyhawk21

Junior Member
Sep 13, 2014
7
0
0
Unfortunately I remembered I bought the MSATA version of this 840 evo samsung 250 ssd and installed it in my hp split x2 laptop 3 months ago. Lo and behold I do my tests and sequential read where data has been sitting for 90 days or longer reads even slower down to 26mb sequential. This is in Windows 7 and Windows 8 on the laptop. It is something defective in the TLC or the controller on these model of ssds. So now I have 3 I need Samsung to replace with ones that do not exhibit this behavior or they can give me my money back for 3 defective items recently purchased retail!

See pics @ links!

https://picasaweb.google.com/lh/photo/V2aQGXk32YfhjPb6GDpdL9MTjNZETYmyPJy0liipFm0?feat=directlink

https://picasaweb.google.com/lh/photo/HZCfH7vbglxFQaaUPQXDQtMTjNZETYmyPJy0liipFm0?feat=directlink
V2aQGXk32YfhjPb6GDpdL9MTjNZETYmyPJy0liipFm0
 
Last edited:

naukkis

Golden Member
Jun 5, 2002
1,004
844
136
It's TLC-memory. Reading slows down massively when data is stayed in memory chips some time. There should be simple firmware fix that will refresh data regularly to correct read times but that fix also will wear out memory sooner.

Everybody should avoid TLC-memory based drives.
 

Cerb

Elite Member
Aug 26, 2000
17,484
33
86
If it's regularly doing that with stale data, that's not cool...at all. Even assuming it's a problem unrelated to NAND quality/environment (theoretical possibility: it could be spending time looking down unbalanced multiversioned trees, or long lists, if log-structured, with packed random data, and not re-writing them, while new data gets better optimized metadata...which eventually becomes like that old metadata, as it gets cluttered with old references to no longer used data).

If TLC drives are offered for 70% MLC drive prices, I'd probably go for it. At >100%, I've been not getting them since they came out :).
 
Last edited: