Intel Optane P4800x 375GB $800

ServerSemi

Member
Jan 12, 2017
107
20
18
37
Hard to justify when the 905p 480gb is a little bit cheaper. Yes I know the p4800x has few extra features but not many people use them to justify the extra cost.
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,008
1,569
113
CA
Got the alert for this this AM too, hard to justify with 905 out now, I'd expect the next-gen Enterprise to come out soon.

Unless they only released the 905P to take away PLP from the 900.... :D
 

oddball

Active Member
May 18, 2018
166
53
28
39
What's the advantage of this over a Samsung 960 Pro? The 960 gets 3,500 MB/s sequential vs 2,500 MB/s for the Intel.

We run some of the 960s in production. Use little daughter boards to seat them and then software RAID. Performance is incredible out of the box and they're a lot cheaper.

I'm struggling with what the attraction is for these things. Maybe they can't sustain as many TB written, but you can buy three 960s for the same price. You need to power down a server to swap both..
 

i386

Well-Known Member
Mar 18, 2016
1,936
505
113
31
Germany
I'm struggling with what the attraction is for these things
Optane drives are awesome cache devices and awesome storage devices where performance, especially small random io, is crucial.

can you run this benchmark on the 960 pro?
Code:
diskspd -b4K -c20G -d120 -L -o8 -r -Sh -t4 -w20 testfile.dat
result for the optane 900p:
Code:
Command Line: diskspd -b4K -c20G -d120 -L -o8 -r -Sh -t4 -w20 testfile.dat

Input parameters:

   timespan:   1
   -------------
   duration: 120s
   warm up time: 5s
   cool down time: 0s
   measuring latency
   random seed: 0
   path: 'testfile.dat'
       think time: 0ms
       burst size: 0
       software cache disabled
       hardware write cache disabled, writethrough on
       performing mix test (read/write ratio: 80/20)
       block size: 4096
       using random I/O (alignment: 4096)
       number of outstanding I/O operations: 8
       thread stride size: 0
       threads per file: 4
       using I/O Completion Ports
       IO priority: normal



Results for timespan 1:
*******************************************************************************

actual test time:   120.00s
thread count:       4
proc count:       8

CPU |  Usage |  User  |  Kernel |  Idle
-------------------------------------------
   0|  99.86%|   6.13%|   93.72%|   0.14%
   1|  99.92%|   6.30%|   93.62%|   0.08%
   2|  99.96%|   6.65%|   93.31%|   0.04%
   3|  99.93%|   6.73%|   93.20%|   0.07%
   4|  14.32%|   8.59%|    5.73%|  85.68%
   5|  12.43%|   8.49%|    3.95%|  87.56%
   6|  12.17%|   8.53%|    3.65%|  87.83%
   7|  16.07%|  11.12%|    4.95%|  83.93%
-------------------------------------------
avg.|  56.83%|   7.82%|   49.02%|  43.17%

Total IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |     15431794688 |      3767528 |     122.64 |   31396.02 |    0.254 |     0.185 | testfile.dat (20GiB)
     1 |     15930585088 |      3889303 |     126.60 |   32410.81 |    0.246 |     0.167 | testfile.dat (20GiB)
     2 |     16548052992 |      4040052 |     131.51 |   33667.05 |    0.237 |     0.154 | testfile.dat (20GiB)
     3 |     16710189056 |      4079636 |     132.80 |   33996.92 |    0.235 |     0.134 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:       64620621824 |     15776519 |     513.56 |  131470.79 |    0.243 |     0.161

Read IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |     12345786368 |      3014108 |      98.12 |   25117.53 |    0.253 |     0.188 | testfile.dat (20GiB)
     1 |     12746182656 |      3111861 |     101.30 |   25932.14 |    0.245 |     0.166 | testfile.dat (20GiB)
     2 |     13237805056 |      3231886 |     105.20 |   26932.34 |    0.236 |     0.157 | testfile.dat (20GiB)
     3 |     13365645312 |      3263097 |     106.22 |   27192.43 |    0.234 |     0.133 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:       51695419392 |     12620952 |     410.84 |  105174.44 |    0.242 |     0.162

Write IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |      3086008320 |       753420 |      24.53 |    6278.49 |    0.256 |     0.177 | testfile.dat (20GiB)
     1 |      3184402432 |       777442 |      25.31 |    6478.67 |    0.249 |     0.169 | testfile.dat (20GiB)
     2 |      3310247936 |       808166 |      26.31 |    6734.71 |    0.239 |     0.144 | testfile.dat (20GiB)
     3 |      3344543744 |       816539 |      26.58 |    6804.48 |    0.237 |     0.140 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:       12925202432 |      3155567 |     102.72 |   26296.35 |    0.245 |     0.158



testfile.dat
  %-ile |  Read (ms) | Write (ms) | Total (ms)
----------------------------------------------
    min |      0.031 |      0.050 |      0.031
   25th |      0.223 |      0.225 |      0.223
   50th |      0.231 |      0.233 |      0.231
   75th |      0.241 |      0.244 |      0.242
   90th |      0.269 |      0.273 |      0.270
   95th |      0.293 |      0.296 |      0.293
   99th |      0.355 |      0.360 |      0.356
3-nines |      1.107 |      1.195 |      1.123
4-nines |      6.970 |      7.419 |      7.129
5-nines |     20.088 |     17.419 |     17.697
6-nines |     37.355 |     32.081 |     37.292
7-nines |     38.610 |     38.613 |     38.612
8-nines |     38.612 |     38.613 |     38.613
9-nines |     38.612 |     38.613 |     38.613
    max |     38.612 |     38.613 |     38.613

total:
  %-ile |  Read (ms) | Write (ms) | Total (ms)
----------------------------------------------
    min |      0.031 |      0.050 |      0.031
   25th |      0.223 |      0.225 |      0.223
   50th |      0.231 |      0.233 |      0.231
   75th |      0.241 |      0.244 |      0.242
   90th |      0.269 |      0.273 |      0.270
   95th |      0.293 |      0.296 |      0.293
   99th |      0.355 |      0.360 |      0.356
3-nines |      1.107 |      1.195 |      1.123
4-nines |      6.970 |      7.419 |      7.129
5-nines |     20.088 |     17.419 |     17.697
6-nines |     37.355 |     32.081 |     37.292
7-nines |     38.610 |     38.613 |     38.612
8-nines |     38.612 |     38.613 |     38.613
9-nines |     38.612 |     38.613 |     38.613
    max |     38.612 |     38.613 |     38.613
 

oddball

Active Member
May 18, 2018
166
53
28
39
Trying to get diskspd to work, in the meantime here are the CrystalDiskMark scores for the 960 RAID1

-----------------------------------------------------------------------
CrystalDiskMark 6.0.0 x64 (C) 2007-2017 hiyohiyo
Crystal Dew World : Crystal Dew World
-----------------------------------------------------------------------
* 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) : 3883.319 MB/s
Sequential Write (Q= 32,T= 1) : 1892.487 MB/s
Random Read 4KiB (Q= 8,T= 8) : 1640.478 MB/s [ 400507.3 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 931.910 MB/s [ 227517.1 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 530.267 MB/s [ 129459.7 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 235.239 MB/s [ 57431.4 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 36.504 MB/s [ 8912.1 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 108.280 MB/s [ 26435.5 IOPS]
Test : 1024 MiB [D: 41.9% (195.0/465.8 GiB)] (x5) [Interval=5 sec]
Date : 2018/03/02 16:29:10
OS : Windows Server 2016 Server Standard (full installation) [10.0 Build 14393] (x64)
 

oddball

Active Member
May 18, 2018
166
53
28
39
Code:
Input parameters:

        timespan:   1
        -------------
        duration: 120s
        warm up time: 5s
        cool down time: 0s
        measuring latency
        random seed: 0
        path: 'testfile.dat'
                think time: 0ms
                burst size: 0
                software cache disabled
                hardware write cache disabled, writethrough on
                performing mix test (read/write ratio: 80/20)
                block size: 4096
                using random I/O (alignment: 4096)
                number of outstanding I/O operations: 8
                thread stride size: 0
                threads per file: 4
                using I/O Completion Ports
                IO priority: normal



Results for timespan 1:
*******************************************************************************

actual test time:       120.01s
thread count:           4
proc count:             20

CPU |  Usage |  User  |  Kernel |  Idle
-------------------------------------------
   0|  26.50%|   1.47%|   25.03%|  73.50%
   1|  28.72%|   2.50%|   26.22%|  71.28%
   2|  25.56%|   1.93%|   23.63%|  74.44%
   3|  25.16%|   1.95%|   23.20%|  74.84%
   4|   7.29%|   0.73%|    6.56%|  92.71%
   5|   1.94%|   1.02%|    0.92%|  98.06%
   6|   2.10%|   0.73%|    1.37%|  97.90%
   7|   1.74%|   0.44%|    1.30%|  98.26%
   8|   1.13%|   0.43%|    0.70%|  98.87%
   9|   0.29%|   0.17%|    0.12%|  99.71%
  10|   0.61%|   0.40%|    0.21%|  99.39%
  11|   0.35%|   0.20%|    0.16%|  99.65%
  12|   1.99%|   1.05%|    0.94%|  98.01%
  13|   1.26%|   0.44%|    0.82%|  98.74%
  14|   1.11%|   0.43%|    0.68%|  98.89%
  15|   0.69%|   0.30%|    0.39%|  99.31%
  16|   7.21%|   2.66%|    4.56%|  92.79%
  17|   0.23%|   0.23%|    0.00%|  99.77%
  18|   0.73%|   0.39%|    0.34%|  99.27%
  19|   1.26%|   0.35%|    0.91%|  98.74%
-------------------------------------------
avg.|   6.79%|   0.89%|    5.90%|  93.21%

Total IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |      3420393472 |       835057 |      27.18 |    6958.38 |    1.149 |     2.050 | testfile.dat (20GiB)
     1 |      3344846848 |       816613 |      26.58 |    6804.69 |    1.175 |     2.712 | testfile.dat (20GiB)
     2 |      3413852160 |       833460 |      27.13 |    6945.07 |    1.151 |     2.112 | testfile.dat (20GiB)
     3 |      3387817984 |       827104 |      26.92 |    6892.11 |    1.160 |     2.283 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:       13566910464 |      3312234 |     107.81 |   27600.26 |    1.158 |     2.302

Read IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |      2736201728 |       668018 |      21.74 |    5566.48 |    0.243 |     0.364 | testfile.dat (20GiB)
     1 |      2676518912 |       653447 |      21.27 |    5445.06 |    0.246 |     0.549 | testfile.dat (20GiB)
     2 |      2731417600 |       666850 |      21.71 |    5556.74 |    0.241 |     0.421 | testfile.dat (20GiB)
     3 |      2707939328 |       661118 |      21.52 |    5508.98 |    0.243 |     0.531 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:       10852077568 |      2649433 |      86.24 |   22077.26 |    0.243 |     0.472

Write IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |       684191744 |       167039 |       5.44 |    1391.91 |    4.772 |     2.015 | testfile.dat (20GiB)
     1 |       668327936 |       163166 |       5.31 |    1359.63 |    4.895 |     4.279 | testfile.dat (20GiB)
     2 |       682434560 |       166610 |       5.42 |    1388.33 |    4.793 |     2.243 | testfile.dat (20GiB)
     3 |       679878656 |       165986 |       5.40 |    1383.13 |    4.812 |     2.858 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:        2714832896 |       662801 |      21.57 |    5523.00 |    4.818 |     2.974



total:
  %-ile |  Read (ms) | Write (ms) | Total (ms)
----------------------------------------------
    min |      0.045 |      0.920 |      0.045
   25th |      0.115 |      3.136 |      0.119
   50th |      0.135 |      4.495 |      0.151
   75th |      0.183 |      6.030 |      0.652
   90th |      0.481 |      7.436 |      4.502
   95th |      0.712 |      8.377 |      6.034
   99th |      2.193 |     10.829 |      8.383
3-nines |      3.269 |     15.200 |     11.804
4-nines |      5.600 |    169.471 |     19.060
5-nines |     21.486 |    249.792 |    203.439
6-nines |    203.152 |    250.639 |    250.286
7-nines |    249.540 |    250.639 |    250.639
8-nines |    249.540 |    250.639 |    250.639
9-nines |    249.540 |    250.639 |    250.639
    max |    249.540 |    250.639 |    250.639
PS D:\>
 

oddball

Active Member
May 18, 2018
166
53
28
39
And an HP NVMe drive (no RAID)

Code:
Input parameters:

        timespan:   1
        -------------
        duration: 120s
        warm up time: 5s
        cool down time: 0s
        measuring latency
        random seed: 0
        path: 'testfile.dat'
                think time: 0ms
                burst size: 0
                software cache disabled
                hardware write cache disabled, writethrough on
                performing mix test (read/write ratio: 80/20)
                block size: 4096
                using random I/O (alignment: 4096)
                number of outstanding I/O operations: 8
                thread stride size: 0
                threads per file: 4
                using I/O Completion Ports
                IO priority: normal



Results for timespan 1:
*******************************************************************************

actual test time:       120.00s
thread count:           4
proc count:             32

CPU |  Usage |  User  |  Kernel |  Idle
-------------------------------------------
   0|  23.39%|   2.25%|   21.13%|  76.61%
   1|  26.52%|   2.21%|   24.31%|  73.48%
   2|  25.33%|   2.60%|   22.72%|  74.67%
   3|  24.96%|   2.23%|   22.73%|  75.04%
   4|   0.70%|   0.52%|    0.18%|  99.30%
   5|   0.21%|   0.14%|    0.07%|  99.79%
   6|   0.70%|   0.56%|    0.14%|  99.30%
   7|   0.18%|   0.09%|    0.09%|  99.82%
   8|   1.29%|   0.26%|    1.03%|  98.71%
   9|   0.85%|   0.09%|    0.76%|  99.15%
  10|   3.19%|   1.77%|    1.42%|  96.81%
  11|   0.01%|   0.01%|    0.00%|  99.99%
  12|   0.29%|   0.23%|    0.05%|  99.71%
  13|   0.13%|   0.09%|    0.04%|  99.87%
  14|   0.22%|   0.17%|    0.05%|  99.78%
  15|   0.21%|   0.08%|    0.13%|  99.79%
  16|   0.43%|   0.21%|    0.22%|  99.57%
  17|   0.14%|   0.10%|    0.04%|  99.86%
  18|   1.12%|   1.05%|    0.07%|  98.88%
  19|   0.40%|   0.30%|    0.10%|  99.60%
  20|   0.18%|   0.16%|    0.03%|  99.82%
  21|   0.27%|   0.13%|    0.14%|  99.73%
  22|   0.10%|   0.05%|    0.05%|  99.90%
  23|   0.74%|   0.42%|    0.33%|  99.26%
  24|   0.90%|   0.65%|    0.25%|  99.10%
  25|   0.03%|   0.01%|    0.01%|  99.97%
  26|   0.31%|   0.23%|    0.08%|  99.69%
  27|   0.09%|   0.08%|    0.01%|  99.91%
  28|   0.04%|   0.03%|    0.01%|  99.96%
  29|   0.33%|   0.27%|    0.05%|  99.67%
  30|   6.51%|   2.17%|    4.34%|  93.49%
  31|   0.18%|   0.09%|    0.09%|  99.82%
-------------------------------------------
avg.|   3.75%|   0.60%|    3.15%|  96.25%

Total IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |      1947635712 |       475497 |      15.48 |    3962.46 |    2.017 |     1.400 | testfile.dat (20GiB)
     1 |      1946431488 |       475203 |      15.47 |    3960.01 |    2.018 |     1.465 | testfile.dat (20GiB)
     2 |      1956315136 |       477616 |      15.55 |    3980.12 |    2.008 |     1.487 | testfile.dat (20GiB)
     3 |      1963872256 |       479461 |      15.61 |    3995.49 |    2.000 |     1.432 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:        7814254592 |      1907777 |      62.10 |   15898.08 |    2.011 |     1.446

Read IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |      1558355968 |       380458 |      12.38 |    3170.47 |    1.963 |     1.450 | testfile.dat (20GiB)
     1 |      1557176320 |       380170 |      12.38 |    3168.07 |    1.965 |     1.455 | testfile.dat (20GiB)
     2 |      1565786112 |       382272 |      12.44 |    3185.59 |    1.953 |     1.547 | testfile.dat (20GiB)
     3 |      1571045376 |       383556 |      12.49 |    3196.29 |    1.947 |     1.393 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:        6252363776 |      1526456 |      49.69 |   12720.41 |    1.957 |     1.462

Write IO
thread |       bytes     |     I/Os     |    MiB/s   |  I/O per s |  AvgLat  | LatStdDev |  file
-----------------------------------------------------------------------------------------------------
     0 |       389279744 |        95039 |       3.09 |     791.99 |    2.234 |     1.151 | testfile.dat (20GiB)
     1 |       389255168 |        95033 |       3.09 |     791.94 |    2.230 |     1.487 | testfile.dat (20GiB)
     2 |       390529024 |        95344 |       3.10 |     794.53 |    2.227 |     1.192 | testfile.dat (20GiB)
     3 |       392826880 |        95905 |       3.12 |     799.21 |    2.213 |     1.559 | testfile.dat (20GiB)
-----------------------------------------------------------------------------------------------------
total:        1561890816 |       381321 |      12.41 |    3177.66 |    2.226 |     1.359



total:
  %-ile |  Read (ms) | Write (ms) | Total (ms)
----------------------------------------------
    min |      0.131 |      0.335 |      0.131
   25th |      1.384 |      1.622 |      1.418
   50th |      1.793 |      2.034 |      1.852
   75th |      2.273 |      2.523 |      2.339
   90th |      2.881 |      3.203 |      2.956
   95th |      3.705 |      4.272 |      3.811
   99th |      5.642 |      6.293 |      5.744
3-nines |      7.585 |      8.272 |      7.739
4-nines |     29.182 |     21.595 |     24.122
5-nines |    222.177 |    113.180 |    221.932
6-nines |    223.051 |    222.843 |    223.051
7-nines |    223.068 |    222.843 |    223.068
8-nines |    223.068 |    222.843 |    223.068
9-nines |    223.068 |    222.843 |    223.068
    max |    223.068 |    222.843 |    223.068
 
  • Like
Reactions: i386

oddball

Active Member
May 18, 2018
166
53
28
39
I'd never run this before, but I'm convinced....

Going to probably dump our 960 Pro's and buy these for SQL Server log drives. A 5x gain!! The write performance is what I care about. We write a LOT per day and that's what's slow. We are going from memory to SQL Server, data drives are SSD, and logs NVMe, but apparently not quick enough.
 

i386

Well-Known Member
Mar 18, 2016
1,936
505
113
31
Germany
I'd never run this before, but I'm convinced....

Going to probably dump our 960 Pro's and buy these for SQL Server log drives. A 5x gain!! The write performance is what I care about. We write a LOT per day and that's what's slow. We are going from memory to SQL Server, data drives are SSD, and logs NVMe, but apparently not quick enough.
Be sure to read the benchmark (representing a workload that I have @work) parameters: it's not using all threads/cores available (workload is running a software that is licensed per cores, we use 4 core high frequency cpus), it disables cache (hardware and software!) and uses 4k blocks (dbs use 8k)
 

EricE

New Member
May 11, 2017
12
1
3
54
Optane makes awesome dedicated L2 cache with PrimoCache: PrimoCache Overview

For $120 there isn't a better (or at least cheaper) way to accelerate your Windows server, especially if you can't afford to go all SSD.

Heck I got the $20 for my workstation and use the 64GB M2 Optane as L2 cache. Even with a Samsung EVO M2 SSD I saw a noticeable difference in boot and performance. I'm a huge fan of PrimoCache.
 

whitey

Moderator
Jun 30, 2014
2,770
866
113
38
p4800x for this, WOW! Yeah even for $100-200 more I'd take the ent class p4800x series over the P905. They are both badass optane devices, I own a 900P...crazy good, but that true 'Intel validated/signed off' extra goodies no matter what you think B.S. or not makes me feel all warm n' fuzzy.

(bits shamelessly 'borrowed' from Patricks review)

As an enterprise part, the Intel® Optane™ SSD DC P4800X offers multiple data protection features that the Intel® Optane™ SSD 900P does not, including DIF data integrity checking, circuit checks on the power loss system and ECRC. The DC P4800X also offers a higher MTBF/AFR rating.
 
Last edited:

Patrick

Administrator
Staff member
Dec 21, 2010
11,908
4,871
113
@oddball sync writes and low QD is where optane shines. It also does not need to do expensive erases like NAND so performance over time is very consistent.

While you can look at max read/write and think the 960 pro is better, Optane is in another performance league for servers.
 

oddball

Active Member
May 18, 2018
166
53
28
39
@oddball sync writes and low QD is where optane shines. It also does not need to do expensive erases like NAND so performance over time is very consistent.

While you can look at max read/write and think the 960 pro is better, Optane is in another performance league for servers.
Yes, seems like it.

For random reads our performance is very fast.

Where we have pain is sustained writes. We write 10-20GB a night and disk is our bottleneck. Right now the setup is RAID 10 with SSD and transaction logs on NVMe. But as I've seen in this thread we could potentially bump things up a few notches by going Optane.