M.2 SSD vs Regular SSD

Laskos · Post by **Laskos** » Fri Apr 08, 2016 11:53 am

yurikvelo wrote:
Nordlandia wrote:Current SSD = speed penalty due to 6-Men probing.
PCIe SSD = cancel out the already applied penalty due to faster speed?

With PCIe SSD my speed can be 100% with 6-Men probing compared to 95/90% with my current SSD.
Even zero-latency (imaginary) storage give speed penalty, because tbcore.cpp + tbprobe.cpp code is CPU-costly

I don't understand this SSD craze for TBs. A small test:

Stockfish 7
30 late endgame positions
30 sec each

1/ No TB:
15,032 Knps

2/ 7200RMP SATA HDD 5-men Syzygy:
12,547 Knps

3/ RAM 5-men Syzygy:
13,299 Knps

No SSD can match RAM Syzygy. The improvement here is 6%, but it is stall way off No TB mark. To spend several hundred bucks on some fancy SSD for 6 ELO points?

yurikvelo · Post by **yurikvelo** » Fri Apr 08, 2016 11:59 am

Laskos wrote:No SSD can match RAM Syzygy. The improvement here is 6%, but it is stall way off No TB mark. To spend several hundred bucks on some fancy SSD for 6 ELO points?

By late EG game is already decided

Let check other position: http://tcec.chessdom.com/archive.php?se=8&sf&ga=13
at Move 72, Komodo get eval +2.02 @ D=39 (position is hard, eval will surpass +6.5 threshold only at Move 90, which is 36 ply deep from here), this is kind of position where EGTB matters most

Laskos · Post by **Laskos** » Fri Apr 08, 2016 12:18 pm

yurikvelo wrote:
Laskos wrote:No SSD can match RAM Syzygy. The improvement here is 6%, but it is stall way off No TB mark. To spend several hundred bucks on some fancy SSD for 6 ELO points?
By late EG game is already decided

Let check other position: http://tcec.chessdom.com/archive.php?se=8&sf&ga=13
at Move 72, Komodo get eval +2.02 @ D=39 (position is hard, eval will surpass +6.5 threshold only at Move 90, which is 36 ply deep from here), this is kind of position where EGTB matters most

I was not saying that TBs don't matter, I am saying that SSD gives some very modest improvement compared even to HDD at not too fast time control.

I checked that position with SF 7 to 30 seconds on 4 cores:

No TB 10,200 Knps
HDD Syzygy: 7,900 Knps
RAM Syzygy: 8,600 Knps

The improvement is a bit larger than average endgame, some 9%, but that's all. To pay hundreds of bucks on some better SSD for this sort of detail? Better have more powerful CPU or RAM.

kgburcham · Post by **kgburcham** » Fri Apr 08, 2016 2:42 pm

yurikvelo wrote:

kgburcham wrote:here is an ATTO bench on my Intel SATA SSD where my EGTB are stored.

please run CrystalDiskMark, IOPS in random read mode.
The same question to Jon Fredrik Åsvang

Code: Select all

-----------------------------------------------------------------------
CrystalDiskMark 3.0.3 x64 (C) 2007-2013 hiyohiyo
                           Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]
 
           Sequential Read :   243.968 MB/s
          Sequential Write :   127.766 MB/s
         Random Read 512KB :   223.618 MB/s
        Random Write 512KB :   127.005 MB/s
    Random Read 4KB (QD=1) :    12.803 MB/s [  3125.8 IOPS]
   Random Write 4KB (QD=1) :    28.522 MB/s [  6963.5 IOPS]
   Random Read 4KB (QD=32) :    55.477 MB/s [ 13544.1 IOPS]
  Random Write 4KB (QD=32) :   104.684 MB/s [ 25557.6 IOPS]
 
  Test : 1000 MB [E: 0.1% (0.1/111.7 GB)] (x5)
  Date : 2014/12/19 1:28:15
    OS : Windows 7 Ultimate SP1 [6.1 Build 7601] (x64)

not sure if settings are what you wanted. Of course this is PCIE.
-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 1829.325 MB/s
Sequential Write (Q= 32,T= 1) : 447.114 MB/s
Random Read 4KiB (Q= 32,T= 1) : 273.191 MB/s [ 66697.0 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 401.868 MB/s [ 98112.3 IOPS]
Sequential Read (T= 1) : 782.459 MB/s
Sequential Write (T= 1) : 383.784 MB/s
Random Read 4KiB (Q= 1,T= 1) : 25.541 MB/s [ 6235.6 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 97.861 MB/s [ 23891.8 IOPS]

Test : 1024 MiB [C: 15.7% (34.6/219.8 GiB)] (x5) [Interval=5 sec]
Date : 2016/04/08 6:38:58
OS : Windows 8.1 Pro [6.3 Build 9600] (x64)

and here is SATA, EGTB stored here

-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 542.822 MB/s
Sequential Write (Q= 32,T= 1) : 274.073 MB/s
Random Read 4KiB (Q= 32,T= 1) : 249.298 MB/s [ 60863.8 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 179.867 MB/s [ 43912.8 IOPS]
Sequential Read (T= 1) : 494.711 MB/s
Sequential Write (T= 1) : 243.908 MB/s
Random Read 4KiB (Q= 1,T= 1) : 28.616 MB/s [ 6986.3 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 105.158 MB/s [ 25673.3 IOPS]

Test : 1024 MiB [E: 25.4% (56.6/223.2 GiB)] (x5) [Interval=5 sec]
Date : 2016/04/08 6:51:48
OS : Windows 8.1 Pro [6.3 Build 9600] (x64)

yurikvelo · Post by **yurikvelo** » Fri Apr 08, 2016 3:35 pm

Despite big advantage in Sequential Read, almost no progress in random read. At low query depth old SATA is even faster than new fancy PCIE

+10% increase at QD=32, -12% regress at QD=1

SATA3: Random Read 4KiB (Q= 1,T= 1) : 28.616 MB/s [ 6986.3 IOPS]
PCI-E: Random Read 4KiB (Q= 1,T= 1) : 25.541 MB/s [ 6235.6 IOPS]

SATA3: Random Read 4KiB (Q= 32,T= 1) : 249.298 MB/s [ 60863.8 IOPS]
PCI-E: Random Read 4KiB (Q= 32,T= 1) : 273.191 MB/s [ 66697.0 IOPS]

kgburcham · Post by **kgburcham** » Fri Apr 08, 2016 3:50 pm

I ran your position through this test.

[d] 8/5pk1/6r1/pR6/8/2N3KP/8/8 w - -

EGTB SATA SSD E:Drive
Operating system on PCIE C:Drive
Boot from PCIE C:Drive

Engine: Komodo 9.42 64-bit (8192 MB)
by Don Dailey, Larry Kaufman, Mark Lefler
Threads now set to 8
Found 172 tablebases (up to 6-man).
Probing tablebases up to 6-man.

35.01 1:05 +1.66 72.Kh2 Ra6 73.Rb2 Rc6 (1.407.406.899) 21641 TB:48.110.723
36.01 1:15 +1.66 72.Kh2 Ra6 73.Na4 Ra8 (1.636.216.022) 21779 TB:54.895.925
37.01 1:21 +1.67 72.Kh2 Ra6 73.Na4 Ra8 (1.790.184.424) 21868 TB:60.116.374
38.01 1:35 +1.67 72.Kh2 Ra6 73.Na4 Ra8 (2.087.423.380) 21879 TB:72.408.568
39.01 1:54 +1.68 72.Kh2 Ra6 73.Na4 Ra8 (2.495.255.035) 21816 TB:92.465.025
40.01 3:43 +1.75++ 72.Kh2 Ra6 73.Nd5 (4.989.778.914) 22303 TB:190.410.366

EGTB PCIE SSD C:Drive
Operating system on PCIE C:Drive
Boot from PCIE C:Drive

Engine: Komodo 9.42 64-bit (8192 MB)
by Don Dailey, Larry Kaufman, Mark Lefler
Threads now set to 8
Found 172 tablebases (up to 6-man).
Probing tablebases up to 6-man.

35.01 0:35 +1.66 72.Kh2 Ra6 73.Rb2 Re6 (769.453.124) 21532 TB:17.947.747
36.01 0:50 +1.72 72.Kh2 Ra6 73.Rb2 Re6 (1.109.171.006) 21956 TB:25.591.608
37.01 1:10 +1.73 72.Kh2 Ra6 73.Rb2 Rc6 (1.560.802.033) 22045 TB:37.699.566
38.01 1:24 +1.74 72.Kh2 Ra6 73.Rb2 Rd6 (1.871.578.063) 22201 TB:47.402.010
39.01 1:52 +1.81++ 72.Kh2 Ra6 73.Rb2 (2.527.544.910) 22420 TB:64.812.719
39.01 2:03 +1.82 72.Kh2 Ra6 73.Rb2 Rc6 (2.778.883.523) 22466 TB:70.586.460
40.01 2:12 +1.82 72.Kh2 Ra6 73.Rb2 Rc6 (2.985.562.491) 22476 TB:76.255.372
41.01 2:32 +1.83 72.Kh2 Ra6 73.Rb2 Rc6 (3.443.209.892) 22568 TB:90.833.980
42.01 3:25 +1.84 72.Kh2 Ra6 73.Rg5+ Kh7 (4.661.916.140) 22683 TB:133.306.161

[Event "TCEC Season 8 - Superfinal"]
[Site "http://tcec.chessdom.com"]
[Date "2015.11.09"]
[Round "13"]
[White "Komodo 9.3x"]
[Black "Stockfish 021115"]
[WhiteElo "3223"]
[BlackElo "3224"]
[ECO "B47"]
[Result "1-0"]

1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 Nc6 5. Nc3 Qc7 6. g3
a6 7. Bg2 d6 8. O-O Bd7 9. Nxc6 Bxc6 10. a4 Nf6 11. Be3 Be7
12. a5 O-O 13. Bb6 Qd7 14. Qe2 Bd8 15. Rfd1 Bxb6 16. axb6
e5 17. b3 Rac8 18. Qe3 Qe7 19. Rd2 Bd7 20. Rad1 Rc6 21. Na4
Be6 22. f4 Bg4 23. Rc1 Bd7 24. c4 Ng4 25. Qd3 Nf6 26. Rcd1
Bg4 27. Rf1 Bd7 28. c5 dxc5 29. fxe5 Qxe5 30. Qc3 Qxc3
31. Nxc3 Rxb6 32. e5 Rxb3 33. exf6 Be6 34. fxg7 Kxg7
35. Ne2 a5 36. Bd5 Rb6 37. Bxe6 Rxe6 38. Nc3 Rfe8 39. Rdf2
Rf8 40. Rb2 b6 41. Rfb1 Rb8 42. Rb5 c4 43. Rg5+ Kh6 44. Rg4
Re5 45. Nb5 c3 46. Rc4 Kg7 47. Rxc3 Rd8 48. Rb2 h5 49. Rc6
h4 50. gxh4 Re1+ 51. Kg2 Red1 52. Rcc2 R1d3 53. Nc7 R3d6
54. Kh3 Rd3+ 55. Kg4 R3d4+ 56. Kg5 R4d6 57. h5 Kh8 58. Rc4
Rg8+ 59. Kh4 Rc8 60. Rbc2 Kh7 61. R2c3 Rd7 62. Rc6 Rd2
63. h3 Rd4+ 64. Kg5 Rd7 65. Rh6+ Kg8 66. Nb5 Rd5+ 67. Kf4
Rxc3 68. Nxc3 Kg7 69. Rxb6 Rxh5 70. Kg3 Rh6 71. Rb5 Rg6+
72. Kh2 Ra6 73. Nd5 Kg6 74. Rb2 Rd6 75. Rg2+ Kh5 76. Nf4+
Kh6 77. Ra2 Rd4 78. Ne2 Rd5 79. Ng1 Rf5 80. Ra3 Kh5 81. Kg3
Rg5+ 82. Kf2 Rc5 83. Nf3 Kh6 84. Kg2 Rc2+ 85. Kg1 Rc5
86. Ra2 Rd5 87. Kh2 Rc5 88. Nd2 Kh7 89. Nb3 Rc3 90. Nxa5
Kg6 91. Nb7 Rd3 92. Nc5 Rd5 93. Rg2+ Kf5 94. Rf2+ Ke5
95. Nb3 Ke6 96. Nd2 Ke7 97. Kg3 Rd3+ 98. Nf3 Ra3 99. Kg4 f6
1-0

kgburcham · Post by **kgburcham** » Fri Apr 08, 2016 3:54 pm

yurikvelo wrote:Despite big advantage in Sequential Read, almost no progress in random read. At low query depth old SATA is even faster than new fancy PCIE

+10% increase at QD=32, -12% regress at QD=1

SATA3: Random Read 4KiB (Q= 1,T= 1) : 28.616 MB/s [ 6986.3 IOPS]
PCI-E: Random Read 4KiB (Q= 1,T= 1) : 25.541 MB/s [ 6235.6 IOPS]

SATA3: Random Read 4KiB (Q= 32,T= 1) : 249.298 MB/s [ 60863.8 IOPS]
PCI-E: Random Read 4KiB (Q= 32,T= 1) : 273.191 MB/s [ 66697.0 IOPS]

Yes I noticed that also.

jdart · Post by **jdart** » Fri Apr 08, 2016 6:19 pm

I have been wary about SSD due to reliability issues. I have had drives crap out on me after very little usage.

Apparently now though Intel and Samsung are pretty good. I believe Samsung has a 10-year warranty on their enterprise-class drives (850 Pro). I would still stay away from other vendors.

PCIe is the future for PCs and workstations. I expect the cost to continue to come down, rapidly.

--Jon

uaf · Post by **uaf** » Fri Apr 08, 2016 6:52 pm

jdart wrote:Apparently now though Intel and Samsung are pretty good.

I would include Crucial too.

Nordlandia · Post by **Nordlandia** » Sat Apr 09, 2016 9:30 am

Can someone summarize this thread

Pros/cons

Regular SSD
PCIe SSD
M.2 SSD

M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD

Re: M.2 SSD vs Regular SSD