So what were the results of the

So what were the results of the upgrade?

In order to validate the work I did, I upgraded one TS-470 box with the Intel 3770T CPU and 16 gigs of RAM and had my second TS-470 box factory original with no upgrades.  I then connected both boxes to an 8 port GigE switch and connected a HP Z600 workstation with 192gigs of Ram, Kingston SSD Drive and dual processors for a total of 12 cores, 24 hyperthreaded.  I then copied 20 gigs of data from the z600 (running Windows 7 ultimate) to the NAS’s over their Sama server.  My goals wasn’t to find the fastest client copy possible, my goal was to compare from a standard windows computer how fast could the NAS’s accept the copy processes.  The 20 gigs was comprised of 3-4 gig DVD ISO’s and 4 gigs of small data files. 

To see how different conditions affected the copy, I copied the data from the Workstation to each NAS.  On the NAS’s I had a 5400 rpm Western Digital Green drive (really quiet) with a non-encrypted volume and an encrypted volume.  I also had an SSD with an encrypted volume.  Finally, I did the copy from the workstation to all of those volumes and then I copied the data back to the workstation on the target drive which was an SSD (non-encrypted).  Here were my results:

20 gigs of data

volume

time

cpu spike

1

write

470-Stock-hd

4:05

37.5

2

write

470-Stock-hd-enc

5:33

90

3

write

470-Stock-ssd-enc

3:26

100

4

read

470-Stock-hd

3:49

20

5

read

470-Stock-hd-enc

4:49

77

6

read

470-Stock-ssd-enc

3:17

100

7

write

470-Ultimate-hd

4:09

23

8

write

470-Ultimate-hd-enc

5:54

25

9

write

470-Ultimate-ssd-enc

3:15

27

10

read

470-Ultimate-hd

3:58

5

11

read

470-Ultimate-hd-enc

4:49

5

12

read

470-Ultimate-ssd-enc

3:31

5

 

 

 

 

 

At first glance, it would seem that the upgrade actually hurt my performance.  If you look at line 1 compared with line 7 I lost 4 seconds of performance.  Again line 2 vs line 8 I lost 30 seconds of performance.  However the trends stops when we get to line 3 vs line 9 where I gained 11 seconds in performance during the write.  However lines 4-6 vs 10-12 I lost some performance again.  However, if you look at the CPU usage during all of these tests the story is different.  For the stock TS-470 when I was writing data I was pegging out the CPU and when I was reading data I again was pegging out the CPU whenever I was reading or writing to the encrypted volumes and especially to the SSD encrypted volume.  However the worst I was doing with the Ultimate model was using ¼ of the cpu and during the read traffic it was really dramatic on the Ultimate, my traffic barely registered with the cpu.

However this is only part of the story.  The TS-470 has only two cores.  Therefore one of the cores was pegged reading or writing encryption to an SSD that could take it as fast as the CPU could pump it out and during the write tests both CPU’s were being engaged.  The TS-470 Ultimate has 4 cores that are hyperthreaded to  8 because it uses hyperthreading.  The cpu performance you see on the chart is only one cpu.  There other seven were sitting there almost  idle.   My goal of not taxing the cpu was achieved by installing the chip with the AES extension.   This can all be illustrated by my cpu performance monitoring during the test.  Below is the actual cpu performance of my test table above.  I have marked each section with the number that corresponds to the chart above.  On my ultimate model you will see two quick spikes which I believe was the OS doing something else however the remainder of the test ran constant.  The first chart is the stock QNAP TS-470:

Stock QNAP TS-470

QNAP TS-470 Performance chart

As you can see, when encrypting and decrypting data, the CPU was chewed up on the stock QNAP TS-470.  Moreover, using SSD drives which can read and write faster meant the CPU was extremely busy to the point of being pegged to keep the SSD’s busy.

Here is the cpu performance on my QNAP TS-470 Ultimate.  Again I put red numbers to correspond to the table above:

 QNAP TS-470 Ultimate

QNAP TS-470 Ultimate cpu Performance

 

As you can see above, only 1 out of 8 cores were busy on the QNAP TS-470 Ultimate and it wasn’t that busy.  Since I don’t have a TS-470 Pro I can’t compare but I suspect it would land somewhere in the middle between the pegged cpu on a stock TS-470 and a cpu that only climbs to about 25% on the Ultimate.

One test I did not run was an SSD without encrypted volumes.  Honestly, I forgot to put that in the mix and I have already upgraded both boxes so I can’t go back now.  Since I run encrypted volumes my question was answered.

Conclusion:

 The upgrade achieved what I wanted.  A cpu that isn’t pegged and uses less power than the TS-470 pro.  However, this may not be the best configuration for you.  As you can see the stock TS-470 beat out my ultimate in every test except for the SSD with encryption, not by a huge margin so the trade-off might be worth it.  If you’re not using SSD’s you may have better performance with 7600 rpm or 10K rpm drives and a TS-470 Pro, if speed of transfers (without encryption) or even with encryption is your only criteria.  7600 rpm and defiantly 10k rpm drivers are much noisier than a 5400 rpm drive with acoustic management turned on and the SSD’s run silent.

 Some other people have said they chose the Ultimate upgrade because of transcoding on the fly.  That may be another reason to choose the Ultimate upgrade or at least choose the Pro model.

Finally, the Intel 3770T defaults to 2.5ghz with burst up to 3.7ghz.  I haven’t investigated to see if the TS-470 is pushing the cpu beyond the default speed.  If there is a turbo boost mode that hasn’t been unlocked yet, it may well outperform the TS-470 in every test as well as the TS-40 Pro.  I hope to figure that out.

          GO TO: You can read about my goals here.

               GO TO:    Why did I upgrade?

          GOTO:       You can see the testing I did before I did the upgrade here.

          GOTO:      You can see how I upgraded the TS-470 from the basic model to what some have dubbed the Ultimate model, however I would argue more like the “best for security and performance“ model.

             Return to Main Page