Samsung 840 EVO 1TB Half performance

M76

M76

[H]F Junkie
Joined
Jun 12, 2012
Messages
14,031
My company recently bought a bunch of these disks and they were installed into random computers around the office.

They have one thing in common. All the drives produce about half the advertised IOPS and Sequential Read/Write speeds, when tested.

What's the problem?

If I enable RAPID mode the drive does perform around the advertised speeds. But that's not real sustainable performance only a software trick to fool the benchmarks. Also some drives are in RAID, so rapid mode can't be used then.

Samsung really included this boost in the specifications as real world performance? OR is something wrong? But with multiple drives in multiple (different) computers?
 
I believe RAPID mode is required to achieve the quoted numbers. Although are any of these systems SATA II or non Intel SATA III ports?
 
Last edited:
As KENNYB said, how did you test? On every Evo drive I've ever installed, (50+) have all been close enough to rated speed and IOPs. And almost 1Gb/sec with Rapid mode enabled. How old are the machines these were installed in? Any SATA2 in the mix? That would definitely explain the R/W discrepancies.
 
M76 said:
My company recently bought a bunch of these disks and they were installed into random computers around the office.

They have one thing in common. All the drives produce about half the advertised IOPS and Sequential Read/Write speeds, when tested.

What's the problem?

If I enable RAPID mode the drive does perform around the advertised speeds. But that's not real sustainable performance only a software trick to fool the benchmarks. Also some drives are in RAID, so rapid mode can't be used then.

Samsung really included this boost in the specifications as real world performance? OR is something wrong? But with multiple drives in multiple (different) computers?
Click to expand...

Got some questions for you:
- What are the actual numbers you are seeing?
- What test were you using to benchmark?
- What interface are you using?

Article on whether you should be using Rapid:
http://techreport.com/review/25282/a-closer-look-at-rapid-dram-caching-on-the-samsung-840-evo-ssd
 
To answer.

I tested with crystaldiskmark 3. 5 pass 1000MB settings. And with Samsung Magician.

Some of the computers indeed have SATAII interfaces. But I'm getting the same numbers on SATAIII ones as well.

What I got is around 250-270Mb/s sequrential read/write. And around 45000. Iops I don't remember the exact exact numbers if it's important I can post them tomorrow.

After enabling RAPID mode it went up to around the advertised numbers that means 540Mb sequential read, and 98000 Iops. On The SATA3 machine. On the SATA II machine only random performance increased, the sequential remained around 280Mb/s.
 
But I'm getting the same numbers on SATAIII ones as well.
Click to expand...

Are these Intel or AMD SATA controllers. Marvell and ASMedia SATA III controllers will actually give you less performance than an Intel SATA II controller in a lot of workloads.

Also it would help some if you have AHCI enabled instead of IDE mode.
 
For the tests determining the IOPS the queue depth is also important. If you test with a lower queue depth than the advertised numbers are based on, you will get lower performance.
If for the sequential performance the amount of data written exceeds the size of the TurboWrite buffer, your performance will significantly decrease.
It is always better to base your buying decisions on actual reviews instead of these best case "up to" numbers.
 
All drives are on Intel controllers. OFC all running in AHCI mode.

Except two that sit on an Adaptec 1430SA in RAID0. But that's also a SATAII controller. Due to raid it does produces over 500MB/s in sequential mode, and gives very good numbers for 4k and QD32 but it still craps out in real world applications. So synthetic benchmarks seem worthless. I replaced two 830 512Gig drives with these. And I get about third or fourth of the performance I used to. When I process data in multiple threads.

Also for the drives sitting on an X58 board, erasing files takes ages. It could take literally an hour to delete a few hundred files from it.

All I can say that I'm really bummed by this I expected the 840 to be a better drive than the 830 series, and not this junker.
 
All I can say that I'm really bummed by this I expected the 840 to be a better drive than the 830 series, and not this junker.
Click to expand...

The 840 Pro is significantly better than 830 the EVO is a little better than the 830 and has the write back cache when the RAPID mode is enabled.
 
drescherjm said:
The 840 Pro is significantly better than 830 the EVO is a little better than the 830 and has the write back cache when the RAPID mode is enabled.
Click to expand...

It might be better for some application. But for me it's much worse than the 830. There is no way around it.
 
M76 said:
All drives are on Intel controllers. OFC all running in AHCI mode.

Except two that sit on an Adaptec 1430SA in RAID0. But that's also a SATAII controller. Due to raid it does produces over 500MB/s in sequential mode, and gives very good numbers for 4k and QD32 but it still craps out in real world applications. So synthetic benchmarks seem worthless. I replaced two 830 512Gig drives with these. And I get about third or fourth of the performance I used to. When I process data in multiple threads.

Also for the drives sitting on an X58 board, erasing files takes ages. It could take literally an hour to delete a few hundred files from it.

All I can say that I'm really bummed by this I expected the 840 to be a better drive than the 830 series, and not this junker.
Click to expand...

If you're using the drive on a X58, it is very likely you're using a Marvell controller at x1 pcie. Not surprised at the performance hit.

Also, prolonged use for the RAID 0 on the Adaptec 1430SA isn't such a good idea just because of TRIM.

Crystal as your only reference isn't such a good idea. Try Iometer as well.
 
M76 said:
It might be better for some application.
Click to expand...

No, not for some application. For all applications. drescherjm is correct that the 840 Pro and EVO are better than the 830.

There must be something wrong in you system.
 
JoeComp said:
No, not for some application. For all applications. drescherjm is correct that the 840 Pro and EVO are better than the 830.

There must be something wrong in you system.
Click to expand...

You mean with all five computers that got the new SSDs installed. Highly unlikely.
 
darkunitzero said:
If you're using the drive on a X58, it is very likely you're using a Marvell controller at x1 pcie. Not surprised at the performance hit.

Also, prolonged use for the RAID 0 on the Adaptec 1430SA isn't such a good idea just because of TRIM.

Crystal as your only reference isn't such a good idea. Try Iometer as well.
Click to expand...


I know the marvell controller is crap, that's why the SSD is connected to the Integrated intel controller, even if it's just SATA300. The marvell controller was crap even with the 830, but the 830 drives worked like a charm with intel rst.
 
JoeComp said:
Actually, 100% certain.
Click to expand...

Let me rephrase that.

Five completely different computers, of which one is a laptop. Has the exact same performance, and you try to blame it on the computers. Nice.
 
M76 said:
Five completely different computers, of which one is a laptop. Has the exact same performance, and you try to blame it on the computers. Nice.
Click to expand...

Not nice. Accurate. And I did not blame it on "the computers".

There is a common denominator other than the SSDs.
 
hey go into the app where you enabled Rapid mode and make sure the drive is set for performance. That same software can make the changes for you if you need. I have he 840 pro 512mb version and it works like a dream. (didn't know the pro had a 1tb version.)

It sounds to me like there is a common thread to the drives that you have installed. And I am not meaning to be rude when I say I think the common denominator is you.
 
I suspect that it's either a common driver issue or a procedural habit that you've picked up and done identically on all of these machines. We all have habits when we load machines. There are too many Sammy Evo's out in the wild w/o complaint for you to claim the entire series is bogus. Did you load Windows from scratch on these or clone from existing installs? Assuming a Clone did you use any of the optimization procedures in the Samsung Magician program?
 
What capacity is are we talking about on these drives?

Keep in mind the performance peak of SSD's shift each generation. An 840 Evo is going to be slightly down on reads and have a lot less performance in sustained writes or with Rapid turned off because it doesn't have as many memory chips on drive compared to the 830 at ~250GB. Think of an SSD as a raid of several memory chips. Remove a couple an performance goes down.

This is why for the M550 and MX100 from Crucial they use older lower density chips for the 256GB and 128GB Drives. It keeps the performance closer to the ~500 and 1TB drives.

If these are 500GB Evo's then ignore the above.
 
You must log in or register to reply here.
Back
Top