[SOLVED] How to Benchmark Firecuda 520?

[warc1]

I have a new PCIe 4.0 PC build using a ROG x570-e motherboard with a Ryzen 9 3950X CPU and a PCIe 4.0 1TB Firecuda 520 system drive. I've benchmarked the system drive using

1. CrystalDiskMark,
2. AS SSD Benchmark and
3. Samsung Magician

I get the following sequential read/write results (MB/s):

1. 5002/1019
2. 4072/3077
3. 5000/1016

1. and 3. are consistent and shows read speeds at spec and write speeds at only 1/5 of spec. 2. shows significantly higher write speeds but both read and write performance is well off specs. Are any of these to be believed? If so, do I have a problem with my drive given the low write speeds? Is there a definitive benchmark tool I should be using instead? I can't find one from Seagate.

 

[Maxxify]

Update the firmware if possible, I've heard that helps with some E16 drives. Be aware that the listed write speed of the drive is the "up to" SLC cache speed at high queue depth. While that will/should show in benchmarks, the real low queue depth (transfer) speed of the base TLC is actually around 1000 MB/s or so. The E16 drives tend to have full-drive SLC caching (which means the cache is 1/3 the size of unused capacity, roughly).

 

[NightHawkRMX]

When Windows are installed on a drive its performance will be reduced in tests as Windows is eating up some of the performance of the drive.

In most cases, a Firecuda is very little faster than a barracuda. The only difference is a cache but this doesn't always show in benchmarks.

Also, marketing for hard drives/SSD is very misleading/optimistic in general

 

[Gam3r01]

When Windows are installed on a drive its performance will be reduced in tests as Windows is eating up some of the performance of the drive.

In most cases, a Firecuda is very little faster than a barracuda. The only difference is a cache but this doesn't always show in benchmarks.

Also, marketing for hard drives/SSD is very misleading/optimistic in general

Agreed. You will never see rated/benchmark speeds in daily use, its just not possible given the type of information you would be handling, other drives in the system, etc.

 

[warc1]

Thanks all for the input. I tried checking for new firmware but Seagate Support reports there is no firmware for the 520 even though Seatools reports a firmware version for my installed drive.

Seatools is very frustrating. There is a firmware update function that is apparently not supported on the 520 because it is impossible to select that drive. Same goes for the diagnostic function in Seatools for which the 520 cannot be selected but my two Samsung EVO 860s can. There's also a tuning function for which I can select the 520, but if I try to run it, crashes.

I do understand how specs can be unrelated to real world performance. However, the numerous reviews I've found on the 520 use the same AS and CrystalDiskMark benchmark tools I've used and they report as much as 4 times the sustained write speeds and as much as 30 times the 4k random write IOPS. Any ideas on why the discrepancies?

 

[popatim]

Remove the drive and install windows to a sata ssd.. Shut down and reinstall the firecuda, Boot from the sata SSD and now run your benchmark.

You absolutely cannot accurately benchmark your boot drive when Windows and other apps are using the drive at the same time.


As for your low writes issue, make sure your NVME drive is in the M2 slot that is connected to the CPU and not the one connected to the chipset.

 

[warc1]

As for your low writes issue, make sure your NVME drive is in the M2 slot that is connected to the CPU and not the one connected to the chipset.

I don't know which is which. It's currently in the slot closest to the CPU on the Asus X570-E motherboard.

 

[Maxxify]

Also make sure you have write caching (in Windows) enabled for the drive, plus that the drive is 4K aligned (AS SSD will tell you this).

 

[warc1]

Remove the drive and install windows to a sata ssd.. Shut down and reinstall the firecuda, Boot from the sata SSD and now run your benchmark.

Did that with the following before and after results for sequential read/write (MB/s):

CrystalDiskMark Before: 5002/1019 After: 4991/1278
AS SSD Benchmark Before: 4072/3077 After: 4226/3816
Samsung Magician Before: 5000/1016 After: 4992/4291

CrystalDiskMark shows only very minor improvement with writes around 25% of specs. AS SSD shows more significant write improvement but reads and writes are about 80% of specs. Samsung Magician reports performance close to specs. I don't know what to make of this wide disparity.

@Maxxify - I did confirm that write caching is enabled on the drive and it is 4k aligned.

 

[Maxxify]

AS SSD is low queue depth (QD1) so it will be different than programs that use higher QD like CDM. Those numbers look good outside of CDM's write result, although it's possible it's just hitting the TLC rather than SLC.