Significant write speed drop on HDD Seagate EXOS X16
Open source disk encryption with strong security for the Paranoid
Brought to you by:
idrassi
Menu
βΎ
β΄
Hi everybody. My system is 12 core Intel i7-12700K, 32GB DDR5 6000MT/s, CL32, MB with Z790 + Windows 10. Whole thread is about copying big files over 1GB.
I'm using VC since beginning and before, I used TC as well. In years, I encrypted tens of HDDs in capacities under or equal 4TB. Every HDD I encrypted had no performance issues. But now I got several Seagate EXOS X16 HDD with 16TB of space. All these disks suffer for significant performance issues while encrypted.
No encryption read/write both close to 300MB/s.
Encrypted, read about the same ~300MB/s, but write dropped to about 50MB/s.
When I starts to copy many 1GB files on freshly created and formatted encrypted disk, I immediately hear disk heads to constantly seeking like crazy. On fresh disk without any files, heads shouldn't be heard. When those files is read back, speed is 300MB/s without head noise. So files can't be fragmented. Then why heads is moving like crazy when writing? According to Seagate, this disk does not have SMR technology that can explain this behavior.
What I tried:
I encrypted entire disk with/without RAW partition
I encrypted entire disk using stored encrypted file-container on unencrypted disk
Different encryption algorithm, or hash. I tried also TrueCrypt. Same thing.
Tried different, older PC (CPU Intel i7 4790K, 32GB DDR3 on MB with Z77) + Windows 7, 10
This is my first bigger jump from smaller disks to a big ones with helium inside. I do not know if this issue is related to my HDD model, or not. Can you share if anyone has same issue or not? If you have bigger disk about 14-20TB, do you experience this issue? If you do, or don't, tell your system specs, used OS and model of used HDD. I'm desperate. Thank you.
Last edit: John Samsiton 2024-04-27
Another, but 8TB Seagate EXOS 7E8 behave the same way. Also I tried to encrypt those disks with another open source solution DiskCryptor and it works flawlessly. Same speeds if I'm writing to it as well as reading. I tried also Bitlocker. Good performance as well. So It looks like Seagate has done something in HDDs firmware against VeraCrypt because older and smaller disks works great and also other encrypting SW works great on modern disks :( Can anyone verify it if you have same issue? Please...
Last edit: John Samsiton 2024-05-15
I don't have that big of a hard drive and would suggest using a file-based volume (*.hc) to see if performance drops.
Hi. Thank you for response. In my first post I wrote "I encrypted entire disk using stored encrypted file-container on unencrypted disk". I was creating big file. Almost that big like available free space to use as much size as possible. Did you meant something else? I also tested smaller one. Like 200GB. Same thing. Writing still very slow. Every time write speed is very good (above 200MB/s), but after few seconds it drops and stays there. Time to time speed jumps above 200MB/s, but for few seconds. I did not mentioned that earlier because it's short and average is still about 50MB/s.
When creating an encrypted volume, try unchecking the Dynamic
That is due to all drives have internal memory cache in a FIFO (first-in-first-out) process writing or keeping Most Recently Used files for faster access. Once the cache is full, you are limited by the physical drive's read/write to the HDD.
I know this does not explain why DiskCryptor and BitLocker have better results. It may due to VeraCrypt's single-threaded IRP completion design.
https://sourceforge.net/p/veracrypt/discussion/general/thread/f6e7f623d0/?page=20#8362
Another possibility probably just applies if you didn't purchase these drives brand new, since I believe they are shipped with 512-byte sector emulation (512e) which is normally quite compatible with VeraCrypt. These drives have a physical sector size of 4096 bytes, referred to as 4Kn. However, they are switchable in firmware between appearing to have either 512 or 4096 byte sector sizes. I remember reading of VeraCrypt write speed issues with hard drives using their native 4Kn sectors (cached performance would seem normal). So if you think this could apply to your drives, open an administrative command prompt and type this command...
fsutil fsinfo ntfsinfo d:(or whatever drive letter).The 10th line of output for this drive will always read "Bytes Per Physical Sector : 4096". However, if set for 512e operation, the 9th line should read "Bytes Per Sector : 512"
Switching between modes of operation takes only a couple of minutes using Seagate SeaChest, but it is totally data destructive.
Thank you for an idea. I tried to change sector size. I was getting "Bytes Per Physical Sector : 4096" and "Bytes Per Sector : 512". After sector size change to 4096 via SeaChestLite, I see now "Bytes Per Sector : 4096". Sadly same issue. Slow write, and fast reading that back (100GB). Is it possible to VC is not fully compatible to this kind of Advanced format disks?
Last edit: John Samsiton 2024-05-16
Sorry nothing seems to work for you, but as I said in my earlier post, I would have expected the 512e firmware setting (default shipping state) to be the most compatible with the VeraCrypt driver. I'm familiar with these Seagate X16 drives, but I purchased 2 self-encrypting drives from this model line and use a separate locking/unlocking utility, so I don't have experience using VeraCrypt on them. But I do have an older Toshiba HDD with 4Kn physical sectors using 512e firmware, and that drive performs perfectly with VeraCrypt with no apparent speed issues. I was hoping that 512e sectors might work for you on the Seagate as well.
Edit: I just remembered that my Seagate EXOS X16 drives (purchased about a year apart) both shipped with SS02 firmware, which I immediately updated to SS04 using SeaChest. Having the latest firmware may improve your results if you don't already have it.
Last edit: Gary Marks 2024-05-16
I have a similar issue. I have a Seagate 12TB ST12000VN0008. Windows 11. AMD Ryzen 5 3600.
I use 500GB file containers, writes are fast at first and drop to ~50MB/s for large transfers. The drive is in 512e mode and was brand new when I bought it. My previous 6TB Seagate did not have the issue.
I also have the partition encrypted with bitlocker and there are no speed issues with that. The drive also has the latest firmware. It's not a huge issue for me, but I thought I'd add to the post.
Thank you Alex R, for sharing your experience. I wanted to know if it's in model, or line of manufacturing process that have this issue.
Thank you ι£δΉζζ³ for hint. I never saw the option "Dynamic" to be active to control. That option is greyed out and unticked. I always creating primary partitions.
Gary Marks, I upgraded firmware to latest SN04 (I had SN03 on all 5 disks I mentioned). The write speed actually went up to average 80MB/s. For curiosity I wrote a script to save write speed every second into graph where I copying big files into blank disk encrypted by VC. It's weird that after a long time, when all caches are surely filled, write speed went up to full speed for 40 seconds. This happened twice in almost 2 hours of copying. I noticed in that 40s window, disk was quiet as it supposed to be. At lower speed, disk was loud by heads moving.
Enigma2Illusion do you think this IRP limitation can cause different behavior between 4TB, and 8TB disk? Because all other in system is the same.
In attachment on Y axis is speed and on X is time in seconds. I attached one more graph for comparison. Same HW, same system, same copied data from same source, different encryption tool - DiskCryptor
Last edit: John Samsiton 2024-05-22
It is probably a contributing factor. DiskCryptor does not have the ability for file containers nor hidden volumes. The technical difference is discussed in the thread below.
https://sourceforge.net/p/veracrypt/discussion/general/thread/9cb021ad/?limit=25#df5a/a805
Another possible contributing to the issue of the IRP:
https://sourceforge.net/p/veracrypt/discussion/general/thread/f6e7f623d0/?page=20#8362
Be aware of the most recent post by the developer explaining his personal life and the huge limitations that will no doubt impact the future development & support of Veracrypt.
https://sourceforge.net/p/veracrypt/discussion/general/thread/6d0c0dfdc8/#ab89
If you cannot accept these limitations, then the time has come for you to find a different encryption software that will meet your expectations. :-)
Last edit: Enigma2Illusion 2024-05-25
Thought I would add more info to this myself as this is a issue to keep coming up. Running x4 6tb exos drives in a raid 5 setup on a supermicro board that uses intel vroc, and write to speed is less then 1mb to the point that Windows shows 0mb on the gui graph. This is mounted file of 5tb in aes. Have tried everything. Threre is defiantly something with these segate exos drives firmware, note show firmware tn04
Adding to this thread since I'm not sure where else to put it, and this one deals with Seagate EXOS disks.
I've recently purchased 8Tb Seagate 7E10 model st8000nm017b. Its' SMART attributes say it's been unused before me, and it has latest SB10 firmware (at least SG program doesn't find anything fresher for this disk). Testing it with Victoria program didn't show any glaring issues except write speeds were lower then expected - 70 to 50Mb/sec. But it wasn't a big enough issue for the shop to accept the return of the disc.
I've created a partition of maximum possible size on this disk in Windows 10 22H2, which gave me 2 partitions on it, one 16Mbs and one 7.3Tbs. Then I encrypted the bigger partition with VC 1.26.24 (latest at the time of writing). During encryption, VC showed write speeds around 70mb/sec and concluded the process in expected 28 hours. Encrypted partition then mounted normally.
However, when copying files from other disks to it I encountered a problem with write speeds. It would initially go up to 250mb/sec or even more (explainable by disk buffer), then settle to expected 70-50Mb/sec, but after some time it'd go down to 5-3Mb/sec and stay there until the end of copy procedure. Should I pause the process, I could see in Resource Manager that disk would keep writing for a minute or so with speed 5Mb/sec (probably emptying the buffer) and eventually stop. Resuming the operation would have the speed restored to 70-50 for some time, then slowdown to 5-3 would occur again. The duration of normal speed doesn't seem to be consistent, neither in time passed, not in Gbs written (in some cases it'd go for 200Gbs before slowdown). There might be a correlation in number of files written, but I can't be sure.
Read speed is entirely unaffected, staying around 250Mb/sec in sequential reads, regardless of how long the read procedure is going on.
Logical sector size is 512, Extended disk control codes are enabled (VC is installed, not portable). Disk Active Time hovers at around 100%, average response time around 500Ms during whole write procedure, both during normal and reduced speed.
Just in case, I tried installing latest VC beta 1.26.27 since its' release mentioned I\O request handling. but the behaviour doesn't change.
Other hardware includes Gigabyte B560 HD3, Intel Core i5-11400 with 16Gb RAM.
I do not know what to try and where to dig to get to the source of this behaviour, maybe someone here can help?
EDIT: Another data point: on a whim I decided to try and disable windows write caching for the disk and the behaviour changed. Now I first get initial burst of write speed courtesy of internal disk cache, then the speed goes down to around 20 Mb/sec jumping as low as 8 and as high as 45 depending on what's being copied right now. Bigger files seem to reach higher top speeds. It stays consistently around 20 though and doesn't drop to 5 except momentarily.
This goes contrary to my expectations - top speeds being lower was expected, but not bottom speeds being higher.
Last edit: Alexander Narovlansky 2025-10-16
Not only EXOS are affected. Different brands, different results. But all have common behavior. If bigger, it's slower. Veracrypt bad write speeds and DiskCryptor writes as unencrypted. Test was CrystalDiskMark v8.0.4 x64 and 8GiB, SEQ1M-Q8T1. I tested 7 random disks I have with different capacities and brands and it looks like VC issue starting at 4TB disks.
!SEAGATE EXOS X16, 16TB speed DC 225MB/s, VC 50MB/s (80 with newest FW)
!SEAGATE EXOS 7E8, 8TB speed DC 200MB/s, VC 50MB/s
!WD model WD40EFRX, 4TB speed DC 155MB/s, VC 97MB/s
!SEAGATE ST4000VM000, 4TB speed DC 185MB/s, VC 116MB/s
WDC WD2002FAEX-007BA0, 2TB speed DC 144MB/s, VC 139MB/s
TOSHIBA DT01ACA100, 1TB speed DC 204MB/s, VC 200MB/s
SEAGATE ST3500413AS, 500GB speed DC 130MB/s, VC 130MB/s
I've never heard of this program (DiskCriptor). I just re-encrypted 2 disks with veracrypt...after many problems (sigh).
Do you think it's better in terms of performance? And in terms of privacy, is it secure ?
Unfortunately, performance with Veracrypt drops dramatically.
This is the performance of my main NVMe drive before and after encryption with VeraCrypt 1.26.27:
before:
https://imgur.com/a/l0EFlIV
after:
https://imgur.com/a/y2j2qAo
Last edit: standard user 2026-03-20
Performance drop for any SSD and HDDs should not at all because:
SSDs is way faster than HDDs in sequential speeds as well as response times. So, in HDDs, the bottleneck is disk performance all the time. But SSDs sequential speeds is too fast for CPU to handle, so CPU is bottleneck here. CPU is not usually involved in performance tests, but when disk is encrypted by software, SPU is always handling it, so you are limited by CPU speed. AES in my case in VeraCrypt's benchmark is above 20GB/s (I have fast CPU 12900k). And this speed only matter in sequential speeds. Not the random. Instruction for CPU taking also some small time and when you testing speed from random places from disk, the CPU can be fast, but if you have tons of small delays for read different places from disk, this delays adds up to significant drop of random r/w speed. And without SW encryption, the delays from CPU does not exist, so there is the drop. But HDDs has painfully slow response time compared to SSDs, so the added CPU delay is insignificant and you shouldn't notice. That's why on HDDs, on SW encryption, there shouldn't be difference. But it is on VeraCrypt only. That's why I started this thread.
About DiskCryptor. Yes, it should be secure. It's also open source as VeraCrypt, but DC have tiny community, if any = support is way worse, but its performance is as without encryption, fantastic to HDDs. Beware that also DC will have significant performance drop in random disk operations on SSDs due to CPU bottleneck.
Sure, thanks for the heads up. I already knew that with modern SSD/NVMe drives, the CPU makes the difference. In fact, i didn't understand why i was having such a dramatic drop in 4K reads, considering i'm using AES. I have a Ryzen 9700X (over 21Gb/s in decryption performance): https://sourceforge.net/p/veracrypt/discussion/general/thread/843f2e41db/?limit=25#f259 .
So i think the integrated Veracrypt benchmark should be offered with 4K performance. Is there a 4k comparison with Diskcriptor ?
Last edit: standard user 2026-03-30