Assistance - Calling all Samsung EVO owners

[Aurhinius]

no it doesn't look as bad but it's starting to show.

Just to keep everyone up to date I am passing information to abarrass and Samsung in the back ground to try and replicate the issues.

Yeah it's strange they can't replicate it but hopefully they can

 

[Aurhinius]

Had another look at your shots John and it is almost completely identical before and after deleting the 30GB of data but you can quite clearly see that adding the 30GB of data dropped the read performance to approx. 200MB.

The fact that taking it off put everything back almost identical is bizarre that it didn't change anything in terms of the drive layout.

Very interesting you would think taking the data off with a straight delete just deletes the allocation table entry thus not affecting the drive state.

 

[Kotu]

Having similar issues, my samsung 830 256GB has perfect performance (line on graph stays at the max it can reach), whereas my samsung 840 EVO only has 2 peaks where it hits max speed, the rest of the time its below 100MB/s. This only shows on the benchmarks which use the graph, if i run crystal disk mark or AS SSD then it shows quite high results. Ill get a bunch of graphs added in the next few days to compare the 2.

EDIT: Here are the results for me, 830 on the left, 840 EVO on the right, graphs look bad for the 840 but the none graph benchmarks look normal?

 

[fizzy]

As another update to mine - here's the previous last entry after some work were it looked mostly cleaned up after some games copied back :-

Installed a couple more games via steam, and drive has around 140GB free currently. Performance has dropped/major spikes are back :-

 

[wazza300]

has anyone tested with rapid mode enabled? maybe it needs to be on?

EDIT: nope still the same with rapid mode enabled

 

[Aurhinius]

Tried with both rapid on and off.

 

[MichaelAwkward]

Yeah, mine were with RAPID mode on. That's on Windows 8.1 btw.

 

[Toytown]

Given that the EVO line of SSD's all seem to exhibit the same flaw, then surely OCUK should protect its customers by removing these from the online shop?

These drives are not fit for purpose, were not tested enough before they were released and quite frankly are a waste of money until Samsung fix them, either by a firmware update or a replacement product.

Not only would this step give Samsung the kick up the backside it needs, but would stop more complaints about this shoddy product line appearing on this and other forums.

OCUK do the right thing! As right now, you are knowingly selling a defective product to your customers.

 

[fizzy]

Further testing, just to be sure.

Found my MB wasn't quite on latest bios - so updated that (Gigabyte Z77 UP5 TH board, to F12 bios).

To hopefully rule out any alignment issues, I copied all the data off, then used diskpart to clean the disk. Then initialised as MBR (was previously a GPT disk although doubt that would have any effect), and created a new partition.

Double checked alignment was ok :-

Copied all data back (about 320GB) and did another test :-

Pretty much the same as before - lots of spiked performance, just more neatly gathered (were data is).

 

[Tephnos]

Can someone tell me what my results are showing on the previous page? I mean to me it looks kind of horrendous?

I haven't cloned anything. Simply running a default 8.1 install on one drive, and Steam on the other.

 

[Rroff]

EDIT: Here are the results for me, 830 on the left, 840 EVO on the right, graphs look bad for the 840 but the none graph benchmarks look normal?

From casual observation is seems that normal file IO operations work at full speed i.e. if a benchmark uses freespace to create say a 500MB file of random data then reads it back by explicitly opening a handle to that file, but attempts to read data raw directly from the file system seems to be getting hung up somewhere - which hopefully is just a firmware bug and something that can be simply fixed.

 

[ANDARIAL]

Anyone know if it is possible to downgrade the firmware?
I know that I had a near flat line a few firmwares ago

 

[Jokester]

This is my new build with the latest Firmware, so looking healthy enough I think?

Going to keep an eye on things over the coming weeks to see if they change.

 

[Kotu]

wonder why mine does so badly, does it need to be nearly full or something?

 

[Aurhinius]

No at this stage Overclockers and Samsung are doing their best to figure out what the issue is but as most of us have different system, os's drive sizes etc etc it makes life difficult for them.

They are gathering together the components they need to run the tests but I am quite surprised they can't replicate it by simply copying over about 50GB of data to a 250GB drive. At this stage though I am not sure they have tried that.

What I do know is they are taking it very seriously and Abarrass is working hard to help so lets take it easy and keep the information coming. We might stumble on a common theme or figure out where this bug is lurking.

As for not selling the drive - The drive works and it does give the appearance of great performance. I am still using mine and it works fine in most circumstances or at least I wouldn't notice there is this underlying problem until I go looking for it.

Right now though I would buy an MX100 over an EVO until this problem gets resolved but that's my personal opinion and you guys should all make your own choices.

There is a chance here this is something to do with TLC nand and not just the firmware. MX100 uses MLC so hence my choice.

That looks very fresh Jokester is that OS only and what OS is it?


I don't think it would be possible to downgrade firmware.

Tephnos you are seeing the issue most of us are with massive performance degradation on area's of your drive. That shouldn't happen for read performance and that is what we are trying to get to the bottom of. If you are not happy with it and you can do a secure erase and restore your drive then that might be worth a shot. Otherwise you'll need to wait and see what comes of Samsung's investigations. At the moment the drive will still function normally so don't panic and watch this space.

 

[Jokester]

Yeah, Windows 7 64bit on the 120GB drive, nothing installed on the other drive yet. There's a few other bits and bobs installed on the boot drive but nothing major.

I'll start loading up the 1TB drive with games over the next few days and see what happens. The see-saw effect is something to do with allignment I think?

 

[Aurhinius]

Your alignment can be checked using AS SSD but I am sure it is okay as windows 7 was SSD aware so should align okay.

So not entirely sure if that is what is causing the seesaw effect. It's almost like a heartbeat effect but an SSD shouldn't do that and certainly not so regularly.

Love to know what causes it though.

 

[wazza300]

its only specific to the evo aswell,other models 830/840 non pro and 840 pro are fine

 

[ICEx21]

I just ran the benchmark on my drive and the results are horrible, I'll be giving the Evo drives a miss until this is resolved. MX100 for me it seems.

 

[Rroff]

Your alignment can be checked using AS SSD but I am sure it is okay as windows 7 was SSD aware so should align okay.

So not entirely sure if that is what is causing the seesaw effect. It's almost like a heartbeat effect but an SSD shouldn't do that and certainly not so regularly.

Love to know what causes it though.

No alignment problems on my Evo - have a little experience with that as I used to have an older additional SSD in this rig that did have alignment issues.

Its a weird one as any operation that works with a file directly seems to function as expected which covers much of the day to day use of the drive but anything that works with the filesystem wholesale or raw direct disc access seems to trip up - at times its almost like a severely inefficient caching routine.