Spinrite 6.1 Benchmark vs Readspeed vs TD Tune

[Bplayer]

It was possibly time a tuneup of the SSD on my laptop, so I fired up the Windows To Go with VM. Checked the performance with RS and HD Tune Free.

RS: 193.2 1388.4 302.8 299.0 341.5

The benchmarks suggested an L3 on the full disk might be beneficial. SR worked fine but it took a bit longer than the initial estimate suggested. There were some interesting results. The huge spikes in the TD Tune graph pre SR were likely due to unused sectors. The L3 activity would have marked them as active as shown in the relatively flat graph below.

Two measurements were taken for RS and SR Benchmark after the L3 run. The results were quite different.

A. Results using the drive configured using an IDE controller:

RS: 203.4 321.9 317.0 323.7 320.9
SR: 239.178 322.109 326.079

B. Results using the drive configured using an AHCI controller:

RS: 516.7 1327.1 1346.4 1314.9 1298.9
SR: 188.438 261.793 263.023

Analysis:
1. SR & RS using the IDE controller give relatively similar results, but slower results over the first section.
2. RS is significantly faster using AHCI (assume this is due the the different built-in drivers in these two apps)
3. SR is slower using the AHCI controller than the IDE controller (also due to built-in drivers)

Conclusions:
1. TD Tune seems to be a reasonable tool to determine if/when/where an L3 should be performed.
2. Readspeed using an AHCI controller provides unrealistic results
3. The relatively slow results, by both SR and RS, over the first part of the drive is inconsistent with HD Tune. L3 should have improved this area. Not sure why this does not show in the results with RS and SR, but does with TD Tune..

 

[peterblaise]

Free HDD Scan seems more accurate and consistent,
though it takes longer than free HD Tune.

This HDD Scan 4.1 Read Graph 32768
( matches SpinRite transfer block size ) took 2:23:13 h:m:s
SanDisk SSD PLUS 2TB SATA3 in SATA2

The peak number is 270 KB/sec, SATA2.

Note the number of 'slow' sectors - 119,233 measured as
taking more than 150 ms to access.

HDDScan Drive Test Report
Model: SanDisk SSD PLUS 2000GB
Firmware: UP4504RL
Serial: 24294H800063
LBA: 3907029168

Report By: HDDSCAN v4.1
Report Date: 2025-03-28 06-06-31 AM

- - - - -

HD Tune 2.55 Fast took only about half a minute:

HD Tune 2.55 Benchmark Fast 8MB
SanDisk SSD PLUS 2000GB

Transfer Rate Minimum : 5.3 MB/sec
Transfer Rate Maximum : 260.3 MB/sec [ SATA2 ]
Transfer Rate Average : 208.3 MB/sec
Access Time : 0.4 ms
Burst Rate : 34.1 MB/sec
CPU Usage : -1.0%

- - - - -

Transfer Rate Minimum : 5.3 MB/sec is enough to make
me go back to HDDs.

I don't care if SpinRite Level 3 might bring an SSD back to
some level of even 1/2 peak performance for a little while.

I know it will drop back down to 5 MB/sec again.

I think SSDs are a way for chip makers to make sales.

But SSDs are not a way for customers to get good value.

- - - - -

HD Tune Accurate took a little longer, about 7 minutes:

HD Tune 2.55 Benchmark Accurate 8MB
SanDisk SSD PLUS 2000GB

Transfer Rate Minimum : 79.5 MB/sec
Transfer Rate Maximum : 262.3 MB/sec [ SATA2 ]
Transfer Rate Average : 235.9 MB/sec
Access Time : 0.4 ms
Burst Rate : 35.2 MB/sec
CPU Usage : -1.0%

- - - - -

See also free ( old ) HD Tach:

- - - - -

Summary and suggestion: Make SpinRite and ValiDrive
GRAPH modules. Both programs collect copious
information that would inform an industry-unique
multidimensional spectral matrix presentation, would
communicate so much in one glance, and would be a
stellar-standout in the world of data storage information
measurement.

Some sample multi-dimensional graphs scrounged from
the web:

I'd love to see a long-term performance representation
of my data storage media.

 

[Bplayer]

My objective is to determine the general condition of the drive which will indicate if full or partial running of SR would be beneficial. From my small sample of SSD's it seems that the first 25-40% of the drives are slower area. So it is practical to stop SR level 3 at 25% then run the benchmark again. Then restart SR if the drive as necessary. HD Tune running for 30 seconds provides enough information for this exercise.

The longer running benchmarks do not provide any additional information for this purpose.

 

[peterblaise]

@Bplayer "... My objective is to ​

determine the general condition of the drive which will ​

indicate if full or partial running of SR would be ​

beneficial. From my small sample of SSD's it seems that ​

the first 25-40% of the drives are slower area. So it is ​

practical to stop SR level 3 at 25% then run the ​

benchmark again. Then restart SR if the drive as ​

necessary. HD Tune running for 30 seconds provides ​

enough information for this exercise. The longer ​

running benchmarks do not provide any additional ​

information for this purpose ..."​

Note the many 'slow sectors' well beyond the front 25%
area in all the graphs I shared.

- - - - -

Free HD Tune 2.55 has not been updated since February
12, 2008, and is by implication inappropriate for SSDs
when compared to fee ( paid ) HD Tune Pro 6.10 version's
claims, read the details under "What's New" at
https://www.hdtune.com/ today, which has been updated
periodically in the 17 years since 2008 with these notable
inclusions:

added support for NVMe drives​

added support for JMicron NVME to USB bridge​

improved support for OCZ and Samsung SSDs​

added support for Windows 10​

Added support for more SSD drives​

Improved support for +4TB disks​

Improved support for +2TB disks​

Improved support for SSD​

added support for more SSDs​

I have noticed a significant difference on SSDs between
the results of the free HD Tune 2.55 and the results from
a demo of the current fee ( paid ) HD Tune Pro 6.10, just
as I have noticed a significant difference between
subsequent runs of free HD Tune 2.55, each run
producing a significantly different graph, especially on
the active system boot drive. I get more similar graphs
when booting from Windows PE and then testing the
system drive when it is not the active boot drive.

- - - - -

Free HDD Scan 4.1.02.9 presumably last updated on
2019-08-31 is available at https://hddscan.com/ and has
a list of supported SSDs there. I find it consistent on any
drive regardless.

- - - - -

I find SSDs inconsistent between models and between
tests, where different transfer block sizes seem to produce
significantly different results, as if I am searching for a
match or break of internal cache, page, and whatever
controller sizes a drive may be using.

I've not found any software yet that iteratively tests and
reports on what may be the internal construct of an SSD's
cache / controller / page / whatever scheme.

I think SSDs depend on Windows cache to provide what
appears to the end user as smooth and fast performance
. . . until it doesn't, then we scurry around looking for ways
to return our computers to peak performance once more.

SSDs by any measure do not perform as promised
regardless of attempted 'maintenance' rewriting. Even
cloning out and SE secure erase does not present a drive
ready to meet promised performance specifications at
every or any sector that is in any way equivalent to the
promise and fulfillment of HDD performance promises
and delivery . . . except SSD peak data transfer rate at
least once someplace across the drive.

I find that HDDs generally have a range from a maximum
of peak performance down to one-half performance
across a smooth-ish arc curve from the front to the end.

SSDs appear to have a peak occasionally anywhere at any
time, but most predominantly present one-half
performance anywhere at any time, and one-tenth or less
performance wherever whenever <-- THIS is the challenge
that SpinRite LEVEL 3 attempts to address, and the results
are terrific . . . for a moment, then poor performance
inexorably seems to return.

- - - - -

All that being said, if we all started sharing Fast 8MB
Benchmark graphs of our SSDs from even free HD Tune
2.55, then we'd have a body of comparison from which to
make more-informed insights.

HD Tune 2.55 Benchmark Fast 8MB:
JAJS600M2TB

Transfer Rate Minimum : 394.2 MB/sec
Transfer Rate Maximum : 517.3 MB/sec
Transfer Rate Average : 492.1 MB/sec
Access Time : 0.2 ms
Burst Rate : 66.2 MB/sec
CPU Usage : -1.0%

I don't believe it, the drive goes way slow and overheats
when it gets 'stuck' on slow sectors.

I'll try HDD Scan 4 - the drive is the boot drive, so it's 'live':

OK, one 'hole' full of, slow sectors.

I'll run HD Tune 2.55 Accurate 8MB Benchmark:

HD Tune 2,55 Benchmark Accurate 8MB
JAJS600M2TB

Transfer Rate Minimum : 165.7 MB/sec
Transfer Rate Maximum : 511.8 MB/sec
Transfer Rate Average : 356.1 MB/sec
Access Time : 0.2 ms
Burst Rate : 72.9 MB/sec
CPU Usage : -1.0%

Yeowwza!

That 'feels' more like a representation of what it is to sit at
this computer and -w-a-i-t- for it to become responsive to
any program activity.

- - - - -

I think graphing software needs to show the environment
information along with the graph:

Date Time Temperature Range​

Drive details​

Active Boot or non-system non-boot drive​

Attachment scheme - USB, SCIS, BIOS, IDE, ATA, AHCI,​

SAS, NVMe ...​

Drive age and power on time​

Drive's self-reported lifetime reads and writes​

Graphing software name and version and any settings​

Not so we can see and 'fix' slow sectors.

But so we can see and share how bad - or good - our
SSDs really are, and refine our observational schemes to
better, more accurately represent the user experience.

- - - - -

Note, SpinRite 6.1 wise, we can calm DynaStat way down if
it's spending a lot of 5-minute sessions NOT recovering
any additional data from slow sectors:

SPINRITE NORAMTEST LEVEL 3 DYNASTAT 1​

. . . and turn DynaStat off completely if the data is of no
consequence, if we're preparing a drive for reuse:

SPINRITE NORAMTEST LEVEL 5 DYNASTAT 0​

 

[DanR]

it seems that the first 25-40% of the drives are slower area. So it is practical to stop SR level 3 at 25% then run the benchmark again. Then restart SR if the drive as necessary. HD Tune running for 30 seconds provides enough information for this exercise.

Yes. The front part of the SSD drive is where the OS and apps are located. Thus, this area is subjected to frequent reads during normal use of the drive. This subjects the front area of a SSD drive to read fatigue, where the NAND charge decreases with repeated reading of the NAND, making these areas harder and slower to read. A refresh, i.e. SR level 3, would restore the NAND charge and thus the read performance.

I would expect SpinRite 6.1 to provide more accurate benchmarking than ReadSpeed v1. RS v1 was released in Dec 2020. Then, SR 6.1 development, starting where RS v1 ended, began with much effort going into improving not only the drive detection and enumeration code, but also the native AHCI and ATA/IDE driver code. Many code improvements were achieved during this initial SR 6.1 development activity for SR 6.1's benefit.

Normally the [noticeable] difference between RS and SR 6.1 should not be all that much. In your OP however, there are some RS results that don't make a lot of sense.
Question: Are these drives connected to a SATA controller, or to an M2 controller? If the latter, and if they are "seen" as SATA,, that might explain the RS numbers.

 

[peterblaise]

@DanR "... I would expect SpinRite 6.1 to provide more
accurate benchmarking than ReadSpeed v1. RS v1 was
released in Dec 2020. Then, SR 6.1 development,
starting where RS v1 ended, began with much effort
going into improving not only the drive detection and
enumeration code, but also the native AHCI and
ATA/IDE driver code ..."​

We use our drives under Windows or other sophisticated
multi-thread multi-process operating systems, not
single-thread DOS, where ReadSpeed and SpinRite do
their work, so I never expected ReadSpeed and SpinRite
benchmarks to be anything other than hardware
throughput via ReadSpeed's and SpinRite's drivers,
optimized for function and speed but only in comparison
to BIOS function and speed, not penultimate function
and speed, just whatever ReadSpeed and SpinRite
needed, nothing more.

- - - - -

@DanR "... Are these drives connected to a SATA ​

controller, or to an M2 controller? ..."​

I though they were NVMe, hence the peak data transfer
rate of 1960 MB/sec.

@Bplayer saw 1960 KB/sec as "... huge spikes ..." rather
than everything else as huge dips. Free HD Tune 2.55
showing 250 MB/sec to 300 MB/sec could be HD Tune's
mismatch to NVMe rather than the drives being M2 SATA.
We cannot compare to new OOBE graphs - new out of
the box experience. I always run a full suite of references
on new drives so I have something to compare to later.
We do not know what the drives are capable of as
installed in in the system.

- - - - -

The "NVMe MTFDKBA512QFM-1B" refers to a Micron
2400 512GB NVMe PCIe Gen4 M.2 SSD, a fast and
high-capacity solid-state drive. [1, 2, 3]​

• Micron 2400: This indicates the specific model of the
SSD from Micron Technology. [1, 3]​

• 512GB: The storage capacity of the drive, meaning it
can store up to 512 gigabytes of data. [1, 3]​

• NVMe: Stands for Non-Volatile Memory Express, a
protocol that enables faster data transfer speeds
compared to traditional SATA SSDs. [1, 3, 4]​

• PCIe Gen4: Refers to the Peripheral Component
Interconnect Express (PCIe) interface used for
communication, with "Gen4" indicating the fourth
generation of the PCIe standard, which offers faster
data transfer speeds. [1, 3, 5]​

• M.2: The form factor of the SSD, a small, compact
connector type that allows for high-performance
storage. [1, 3, 6]​

• MTFDKBA512QFM-1B: This is the specific part
number or model identifier for this particular SSD.
[1, 3]​

​

[1] https://www.micron.com/products/sto...t-catalog/part-detail/mtfdkba512qfm-1bd15abyy
[2] https://www.cdw.com/product/micron-2400-ssd-512-gb-pcie-4.0-nvme/7845894
[3] https://www.ebay.com/itm/325929345384
[4] https://www.liquidweb.com/blog/nvme-vs-ssd/
[5] https://www.corsair.com/us/en/explo...rence-between-m2-pcie-sata-and-nvme-for-ssds/
[6] https://www.ibm.com/think/topics/nvme-vs-m2​

- - - - -

At this point, we'd need to see their behavior after SE
secure erase in order to have something 'pristine' to
compare to.

 

[Bplayer]

We clearly have different goals in this process. I only want an indication if there are slow sections that would benefit from running SR. This benchmark would be run every 6-12 months. I do not need to know if my SSD is the fastest on the market, how it compares across my devices, or how it performs compared to the previous generation. That is a different goal that I have no interest in.

Maybe the benchmark in SR is sufficient for my needs, but I found the 15 year old Free HD Tune seemed to provide just a bit more information. Maybe it is accurate or maybe it is providing misleading information, but it seems, in combination with SR benchmark, to provide a reasonable high level view.

My expectation is that the SSD's in my desktop and laptops will be in service for about 7 years. If less then they are replaced, or a bonus if more. Just running SR level 3 on the entire drive every 6 months will be unlikely to put a dent in that lifespan. If the SSD's provide 5 good years of performance with SR vs 7 years of mediocre response without SR, then I will take the 5 years. So for maximum effectiveness it comes down to determining if a tuneup is run every 6 or 12 months, or just a tuneup of the first 30% of the drive at those intervals.

 

[DanR]

So for maximum effectiveness it comes down to determining if a tuneup is run every 6 or 12 months, or just a tuneup of the first 30% of the drive at those intervals.

Agreed. Per my previous post this thread I would do the first 30% or so as benchmarking indicates.

For additional perspective, you might take a look at this: https://forums.grc.com/threads/results-samsung-840-series-250-gb.923/#post-14759

 

[peterblaise]

https://forums.grc.com/threads/results-samsung-840-series-250-gb.923/#post-14759 seems SATA2,
right?

See also https://forums.grc.com/threads/readspeed-results-for-3-year-old-samsung-840-evo-ssd.393/
which seems SATA3.

They both show poor performance well below peak
manufacturer's promise, from only 5% to 25% of peak
data transfer rate - ouch.

On the one hand, it's good that there may be ways to
return performance . . . for a while.

On the other hand . . .

 

[DanR]

https://forums.grc.com/threads/results-samsung-840-series-250-gb.923/#post-14759 seems SATA2, right?

Right. SATA III drive on a SATA II controller. Scroll up in the thread. The first post explains it.

 

[peterblaise]

And it's not just the USB2-equivalent data transfer rate
speed and slower that our SSDs collapse on down to, or
USB1-rates, or worse.

For me, it's the stuck hesitation, where I watch the
temperature go up as the SSD is doing something internal,
I dunno, maybe working real hard on 'fixing' and
reallocating whatever it is having trouble reading, and all I
can do is wait for it to finish, I suppose, and I see the
temperature to go down again, and then I can use my
computer again.

If anyone has experience of any brand name and model
SSD that never looses performance, please do share.

- - - - -

HDD Hard Disk Drives, specifically CMR Conventional
Magnetic Recording HDDs, are more of a miracle than I
thought.

And even 'Hammer' HAMR Heat-Assisted Magnetic
Recording Seagate and WD 18TB+ seem surprisingly
responsive, consistent, and reliable.

Even my SMR Shingled Magnetic Recording drives have
been 100% reliable and consistent - just comparatively
too lethargic for boot OS operating system compared to
CMR or HAMR drives.

- - - - -

This topic is refreshing SSDs.

I have 100% success rate in doing that by many means,
including SpinRite 6.1 rewriting levels 3, 4, and 5..

100% of those SSDs revert to the 25% on down to 5%
and slower data transfer rate performance, some
immediately, but all eventually.