SAS SSD power consumption identifier

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Whaaat

Active Member
Jan 31, 2020
393
220
43
Hi everyone! I struggled to find an answer here on forums, but didn't succeed, so decided to post it here.
Accidentally I bought one Toshiba PX04S off ebay for a very good price in a brand new condition. Then I bumped into a comprehensive review of this drive. What immediately caught my attention is that the drive was tested in two performance modes (9W and 12W) which differ by 30% in sequential write speeds and IOPS. In the faster mode the drive also consumes about 1W more power confirmed by measurements. Similar to CPU overclocking, right? Ok, I don't care if a single drive will consume 7W or 8W during rare moments of burst write in home environment, but +30% to performance looks very desirable. So the question is: how do I change the performance mode? Original article doesn't help at all:
We take this opportunity to thank Toshiba for permitting Myce to test the drive in both power modes. The way in which the power mode is configured is only disclosed to customers once they have entered into a non-disclosure agreement.
Then I started googling and found an article, in which Toshiba engineers confirm that by default they are dumping 30% of write performance to limit the drive power consumption. Ok, it may be relevant for multi-drive backplanes in corporate environment, but, what the hell, there is not a single user manual document from Toshiba which unveils how to switch this mode. You can notice that almost every fast SAS SSD produced after 2014 has several power modes according to the datasheets, ranging from 6W to 14W usually, which is user selectable. From Hitachi and Micron user manuals I found that there is mode page that is responsible for power consumption id, and every manufacturer has it's own default value and set of values to choose from for different power modes.

Well, I'm also using one Sandisk Lightning Ascend Gen II, which according to it's datasheet doesn't have different power consumption levels, only one 'typical 5.9W'. But Sandisk scli utility shows us how deep actually the rabbit hole goes:
Code:
C:\Program Files\SanDisk\scli\bin64>scli.exe pwrconsumptionid disk3 -l

disk3 : Power Consumption Values
=================================
Identifier  Power Consumption Value
----------  -----------------------
   0          9 Watts
   1          10 Watts
   2          11 Watts
   3          12 Watts
   4          13 Watts
   5          14 Watts
   6          15 Watts
   7          16 Watts
   8          17 Watts
   9          18 Watts
   10         19 Watts
   11         20 Watts
   12         21 Watts
   13         22 Watts
   14         23 Watts
   15         24 Watts
   16         25 Watts

Command Executed Successfully.

C:\Program Files\SanDisk\scli\bin64>scli.exe pwrconsumptionid disk3 -g

disk3 : Power Consumption Identifier
=====================================
Page Control Type  Power Identifier
-----------------  ----------------
Current            3
Changeable         255
Saved              3
Default            3

Command Executed Successfully.
So we have a lot of values to choose from, with the default id=3 which is 12W. One can change it with sdparm or scli. Let's see what we have for PX04S:
Code:
C:\Program Files\SanDisk\scli\bin64>scli.exe pwrconsumptionid disk2 -l

disk2 : Power Consumption Values
=================================
Identifier  Power Consumption Value
----------  -----------------------
   0          9 Watts
   1          12 Watts

Command Executed Successfully.
Bingo, scli shows that indeed we have only two modes and which id corresponds to the higher consumption. Alter it with sdparm and that's it:
Code:
    Power consumption [ps] mode page:
  ACT_LEV       0  [cha: n, def:  0, sav:  0]  Active level
  PC_ID         1  [cha: y, def:  0, sav:  1]  Power consumption identifier
 

Whaaat

Active Member
Jan 31, 2020
393
220
43
Looks like the power mode on intel ssds impacts the performance in sequential workloads (still better than hdds :D)
yep, the most strange is that in the most power limited case intel writes faster in 4k random (200 kiops ≈ 800 MBps) than in 64k sequential -> 600 MBps only, - that's odd.