Overclock.net › Forums › Components › Hard Drives & Storage › SSD › Samsung 840 EVO read speed drops on old-written data in the drive
New Posts  All Forums:Forum Nav:

Samsung 840 EVO read speed drops on old-written data in the drive - Page 5

post #41 of 3279
Quote:
Originally Posted by gino074 View Post

Hi, that's strange... crystaldiskinfo shows only 2 drives... Which kind of RAID 0 are you running? Hardware or software?
Well when I first discovered the old file problem, I got files reading at ~60MB/sec maybe less. Now that can explain a bottom speed of ~100MB/sec like in your graph for 2 drives in raid 0.
With 4 drives that should be faster but from your benchmarks your top speed is 900MB/sec which is still in line with 2 drives in raid 0... It seems that yours 4 drives are performing like a 2-disk raid 0 confused.gif
Ah of course your PC is still much much faster than with a mechanical drive. Even if read speed drops to that low level, IOPS, 4K, access time are way better than a standard disk. But that does not justify that sequential read on old data sucks very hard... frown.gif

Anyway, If you re-write some of the slow files (by e.g. moving them to an external drive and than back to the EVOs) in parkdale you'll see a "portion" of the drive at top speed which before was slow.

Benchmarks like CrystalDiskinfo, AS SSD benchmark will always show a full-performance drive as they always write a new file for benchmarking. So new file = top speed on EVOs.

@Brainsplatter: IMHO the parkdale block scan is accurate. Maybe you can try to find on your drive other slow files for the disk portions that parkdale shows at 100-200MB/sec (not an easy task, I know...)
Raid0 has 4 drives but only shows as one complete drive; RiVBE bios controlled. The other disk is my RamDisk, I dont have any mechanical drives attached to the computer.
Finding these old files isnt going to be easy and doing this even once a month is a monumental task. Looks like a lot of money was wasted going to Samsung. They must have made a lot of shortcuts and this was a cause that was never been advertised/exposed
post #42 of 3279
If I find some time, I will write a benchmark program for reading existing files. That seems to me the only way to test the existence of this problem on different computers.
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
post #43 of 3279
Ok, I wrote a small tool which will read all files from a given directory (or drive) and below which are larger than 20MB (can be changed) .

On my only 2 week old 1TB EVO which is about 80% full I found no obvious problem. The test included about 5000 files and the average read speed was 430 MB/s. The lowest read speed was 350MB/s.

The tool needs some finetuning and better usability. But I think in a few days I can upload it for everyone to test their SSDs.

Update:
Just did the test on the 1TB EVO again and now I have an average of 477MB/s for all 5860 files. Lowest speed was 450MB/s. Not sure why the result is so different than on first run. Maybe there was background activity on the first run. Will have to investigate.
Edited by BrainSplatter - 8/27/14 at 1:20pm
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
post #44 of 3279
Quote:
Originally Posted by BrainSplatter View Post

Ok, I wrote a small tool which will read all files from a given directory (or drive) and below which are larger than 20MB (can be changed) .

On my only 2 week old 1TB EVO which is about 80% full I found no obvious problem. The test included about 5000 files and the average read speed was 430 MB/s. The lowest read speed was 350MB/s.

The tool needs some finetuning and better usability. But I think in a few days I can upload it for everyone to test their SSDs.

Update:
Just did the test on the 1TB EVO again and now I have an average of 477MB/s for all 5860 files. Lowest speed was 450MB/s. Not sure why the result is so different than on first run. Maybe there was background activity on the first run. Will have to investigate.
Is it writen in C++?, thats my language with MFC
post #45 of 3279
Yeah, C++ just without MFC. Plain Win32 is enough since there isn't much interface.
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
post #46 of 3279
Not directly related to the original problem but still interesting imho.

I figured out where the big difference in average read speed came from when I tested my tool yesterday. It depends a lot on the power saving plan (or whether there is some other process active keeping the CPU busy = 'high performance plan').

With my 830, the difference between Power Saver and High Performance is about 100MB/s.

Here is a screenshot from my benchmark program. While it's running I switch the power plans and you can clearly see the differences in the last column:

Samsung 830 256GB


Samsung 840 EVO 1TB



I didn't expect the power saving plan to make such a big difference.
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
Fun
(15 items)
 
AntiAlias
(11 items)
 
 
CPUMotherboardGraphicsRAM
4770K @ 4.5Ghz @ 1.38v delidded Gigabyte GA-Z87X-OC f9c 2 x Zotac 980Ti Amp Omega @1530, @1530 G.SKILL TridentX 2400 CL10 
Hard DriveHard DriveOptical DriveCooling
Samsung 830 256GB 4 x Samsung EVO 850 1TB RAID0 Samsung Blu Ray Cooler Master Seidon 240M 
OSMonitorKeyboardPower
Windows 7 Professional 64 bit Samsung 27" IPS 2560x1440 Logitech Wireless Seasonic Platinum 860 
CaseMouseAudio
NZXT Phantom 630 white Razer ASUS Xonar U7 external 
CPUMotherboardGraphicsGraphics
7700K@5Ghz ASUS ROG STRIX Z270F Asus 1080Ti FE Gigabyte 1080Ti FE 
RAMHard DriveHard DriveOS
Corsair Vengeance LED 32GB (4x 8GB) DDR4 3400MH... Samsung 960 EVO 500GB 4 x 1TB 850 EVO SSD RAID0 Windows 10 Pro 64 Bit 
MonitorPowerCase
LG 27MU67-B 4K BeQuiet Straight Power 10 800W anidees AI-Crystal black 
  hide details  
Reply
post #47 of 3279
Quote:
Originally Posted by BrainSplatter View Post

Not directly related to the original problem but still interesting imho.

I figured out where the big difference in average read speed came from when I tested my tool yesterday. It depends a lot on the power saving plan (or whether there is some other process active keeping the CPU busy = 'high performance plan').

With my 830, the difference between Power Saver and High Performance is about 100MB/s.

Here is a screenshot from my benchmark program. While it's running I switch the power plans and you can clearly see the differences in the last column:

Samsung 830 256GB


Samsung 840 EVO 1TB



I didn't expect the power saving plan to make such a big difference.

I wonder why would windows decrease power usage on an SSD which allready uses an neglectible amount of power... dafuq windows?
MahPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600 (non k) P8h67-M Pro XFX R9 290 EDBD Kingston 
Hard DriveCoolingOSMonitor
Seagate 7200.11 Coolermaster 412S Windows 8.1 Dell Monicrap 
KeyboardPowerCaseMouse
Dell mechanical keyboard Corsair 850RM unknown Logitech M90 
Audio
Realtek HD integrated audio 
  hide details  
Reply
MahPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600 (non k) P8h67-M Pro XFX R9 290 EDBD Kingston 
Hard DriveCoolingOSMonitor
Seagate 7200.11 Coolermaster 412S Windows 8.1 Dell Monicrap 
KeyboardPowerCaseMouse
Dell mechanical keyboard Corsair 850RM unknown Logitech M90 
Audio
Realtek HD integrated audio 
  hide details  
Reply
post #48 of 3279
Thread Starter 
No, BrainSplatter is talking about CPU power saving working on CPU speed multiplier and CPU voltage. That kind of power saving impacts I/O performance.

I remember that some time ago I read somewhere to disable the power saving on CPU to get more accurate (and repeatable) disk benchmarks. But IMHO if you keep the PC a lot ON, it's better to have the CPU power saving turned on.

On disks, the only power-saving feature I know is putting the disk on sleep if it has not been in use since a configurable time (on mechanical drives this means that the motor is turned off and heads are parked). But as soon as the disk is needed again is turned back on.
post #49 of 3279
I see, your point about the cpu's self downclocking in relation to io makes sense.

Aren't there discs that spin down marginally to eg. 5400rpm instead of normal operation 7200rpm?
MahPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600 (non k) P8h67-M Pro XFX R9 290 EDBD Kingston 
Hard DriveCoolingOSMonitor
Seagate 7200.11 Coolermaster 412S Windows 8.1 Dell Monicrap 
KeyboardPowerCaseMouse
Dell mechanical keyboard Corsair 850RM unknown Logitech M90 
Audio
Realtek HD integrated audio 
  hide details  
Reply
MahPC
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600 (non k) P8h67-M Pro XFX R9 290 EDBD Kingston 
Hard DriveCoolingOSMonitor
Seagate 7200.11 Coolermaster 412S Windows 8.1 Dell Monicrap 
KeyboardPowerCaseMouse
Dell mechanical keyboard Corsair 850RM unknown Logitech M90 
Audio
Realtek HD integrated audio 
  hide details  
Reply
post #50 of 3279
Thread Starter 
Quote:
Originally Posted by ThijsH View Post

Aren't there discs that spin down marginally to eg. 5400rpm instead of normal operation 7200rpm?
Yes, you are right, but before I was talking about power-saving settings that you can configure in windows, like the range of % of minimum-maximum CPU perforformance.

just for curiosity I run again the parkdale scan of my EVO drive disabling the CPU power saving and in this case it's hard to see any relevant difference.
Maybe the scan test is demanding more CPU power, keeping CPU always at max frequency. Or maybe parkdale is averaging or compensating the speed measures...

@BrainSplatter: maybe you can improve your software by implementing a file "rewrite" routine for files reading below a threshold, for example rewrite the file if the read speed is below 200MB/sec whistle.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: SSD
Overclock.net › Forums › Components › Hard Drives & Storage › SSD › Samsung 840 EVO read speed drops on old-written data in the drive