Samsung 990 PRO 2TB: Performance Degradation After 7B2QJXD7 Firmware Upgrade - Capacity-Specific Bug?

[SFdrifter]

PLEASE NOTE THAT THIS FIRST POST WAS MOSTLY INCORRECT -- I MISINTERPRETED SLC CACHE EXHAUSTION FOR POOR PERFORMANCE; I DO NOT BELIEVE THERE IS A FIRMWARE BUG INVOLVED

I'm reporting an apparently capacity-specific firmware bug affecting Samsung 990 PRO 2TB drives after updating to firmware 7B2QJXD7. My 4TB 990 PRO with identical firmware performs normally.

---

**AFFECTED DRIVE:**
- Model: Samsung SSD 990 PRO with Heatsink 2TB (used as system drive with 5 partitions)
- Serial: S7DRNJ0XC08579H
- Firmware: 7B2QJXD7 (upgraded from 4B2QJXD7)
- System: Arch Linux, AMD Ryzen 9 9900X

**COMPARISON DRIVE (NORMAL):**
- Model: Samsung SSD 990 PRO 4TB (used as data drive with 1 partition)
- Serial: S7DSNJ0X912378T
- Firmware: 7B2QJXD7 (upgraded simultaneously with the 2TB SSD)

---

**PERFORMANCE DEGRADATION (FIO BENCHMARKS):**

**2TB Drive - BEFORE firmware upgrade, using 4B2QJXD7:**
✓ Sequential Write: 6,650 MB/s
✓ Random Write: 4,436 MB/s (1,135K IOPS)
✓ Sequential Read: 6,650 MB/s
✓ Random Read: 2,970 MB/s (760K IOPS)

**2TB Drive - AFTER firmware upgrade, using 7B2QJXD7:**
xx  Sequential Write: 1,478-3,691 MB/s (45-78% LOSS)
xx  Random Write: 1,375-1,460 MB/s (68-70% LOSS) ← CRITICAL
✓  Sequential Read: 6,752-7,007 MB/s (normal)
xx  Random Read: 2,369-2,582 MB/s (13-20% loss)

**4TB Drive - AFTER firmware 7B2QJXD7:**
✓ Sequential Write: 6,577 MB/s (NORMAL)
✓ Random Write: 4,435 MB/s (NORMAL - matches 2TB baseline!)
✓ Sequential Read: 6,687 MB/s (NORMAL)
✓ Random Read: 4,594 MB/s (EXCELLENT - better than 2TB baseline!)

---

**TESTING METHODOLOGY:**

To isolate the issue, I performed extensive testing:

1. Tested across multiple partitions (root, home, data)
2. Tested from Live USB (eliminating OS activity)
3. Tested both drives with identical parameters
4. Monitored temperatures during tests

**KEY FINDINGS:**
- Performance degradation occurs on ALL partitions of 2TB drive
- Live USB testing (no OS) still shows degraded performance
- 2TB drive runs 26°F hotter than 4TB (113°F vs 87°F)
- Higher temp suggests excessive write amplification?
- 4TB drive with SAME firmware performs perfectly
- Both drives updated simultaneously
- Both have similar write endurance (~22-23TB)

---

**TEMPERATURE COMPARISON:**
- 2TB (Problem): 113°F / 45°C (elevated)
- 4TB (Normal): 87°F / 31°C (normal)

---

**POSSIBLE IMPLICATIONS**

Is this a **capacity-specific firmware regression** affecting 2TB 990 PRO models?  Does the 7B2QJXD7 firmware have different code paths for different capacities, and does the 2TB implementation have a critical bug in random I/O handling?

The elevated temperature on the 2TB drive suggests the firmware is causing excessive internal operations (write amplification, garbage collection, etc.).

---

Dear Samsung -- 

1. Can you please confirm this capacity-specific bug?
2. Can you investigate why 2TB apparently differs from 4TB implementation?
3. Can you provide a firmware fix (8B2QJXD7) or rollback capability?
4. Can you provide guidance for affected users?

**Has anyone else experienced this after updating 2TB 990 PRO to 7B2QJXD7?**

I have detailed benchmark logs, SMART data, and test results available as requested.

---

**System Details:**
- OS: Arch Linux (kernel 6.x)
- Motherboard: AsRock x870e Taichi
- Testing Tool: fio 3.39
- Test Parameters: libaio, direct I/O, various block sizes

[Memory Product Expert -1]

Hello useroIUpCVJ7VK,

Sorry, maybe I misunderstood, Did your computer manufacturer say that you needed to reinstall windows to turn off VMD?

I personally have a Asus laptop that i turned off VMD mode, my laptop just rebooted a few times and once I was back at my desktop, VMD was off and Samsung magician was working fine.

I also have a desktop with MSI motherboard, VMD was off by default, but I can turn it on without having to reinstall windows.

[useroIUpCVJ7VK]

That was Samsung's advice. Because it's HP that's not an option I have to reinstall Windows to turn it off, or I can do the manual firmware updates. So no reason for me to stick with a company that could fix an issue with an update. Instead, make the customer do it, and gaslight them, and tell them VMD has never worked with Magician. If I were building a gaming machine this wouldn't be an issue. But at least I have learned a lesson. Always go with your gut.

[Memory Product Expert -1]

Hello useroIUpCVJ7VK

Understood, if this is not a Samsung laptop, I think Samsung support would only be able to give general advice, 

But, I would still check with HP to confirm, you'd be surprised but a lot of computer manufactures have easy ways to turn off VMD without reinstalling Windows.

Having to reinstall Windows due to a BIOS setting being turned off was needed back in the older days of windows, but if your using Windows 10 or  11, there are easy ways around this. 

But if HP is not providing any information for this, then I agree that it wouldn't be worth the effort in most cases, especially if your SSD is not having any issues, then doing all of this wouldn't even be necessary 

[useroIUpCVJ7VK]

It won't recognize the drive with it turned off. I mean I've had zero issues with this drive. I normally don't do updates until I see what problems they've caused, When I bought the 990 pro I knew about the degradation problem they were denying and then finally admitted it, and put out a firmware update to fix it. I installed the drive clean install of Windows 11, I went straight to Samsung installed the Magician software it said there was a firmware update and it's worked ever since. I have received updates through there until now. Like I said they could fix they're choosing not to fix it.
If I have to manually do updates from them why should I even keep their software installed?

[Memory Product Expert -1]

Hello useroIUpCVJ7VK,

Well, I would ask the system manufacturer why do your SSDs not show up when VMD mode is off,

Regardless of the issues, since its your computer, you should be able to set the computer to any mode and that should not cause your SSD to stop being detected.
Both of my computers let me use the SSD with VMD off or on.

You should be able to just turn off VMD, let windows do it restart and readjustments, then go to Magician and do the update as needed.

[niegowic]

I got the same problem 😞 After upgrade to latest firmware it sucks.. it was working much better before upgrade of firmware

[niegowic]

Before upgrading IOPS dropped by half... 

[Memory Product Expert -1]

Hello niegowic,

Are you able to post screenshots of your performance test results and your system specs?

If you are having an issue, make sure to post all important information showing the problem and why you think this issue is caused by the firmware. Also, make sure you are using the product using the recommended settings and Operating system. 

Otherwise, no one will be able to help or investigate the issue...

[Memory Product Expert -1]

Here are my performance results as of this morning, for reference 

 

 

[SFdrifter]

Ok...I may have created a tempest in a teapot, but I think I have a better sense of what’s happening. Whether it’s by design or a bug, I’m not sure. Let me focus on the random write issue.

First, apologies to Memory Product Expert -1 -- after reviewing all my benchmark data since April, I don’t think what I’m seeing is tied to the new firmware – I’ve been seeing poor random write performance all along. It just coincidentally happened that my benchmark using the 4B2QJXD7 firmware immediately before updating to the 7B2QJXD7 firmware had exceptionally good results for the 2TB drive, and my initial post was prompted by the difference in performance that I saw.

Second, the parameters I’ve been using in fio testing have a large part to play. I’ve previously tested random write performance using this command (call it the “sustained random write” test):

fio --name=write_iops_random --directory=”$TEST_DIR” \

--numjobs=2 --size=10G --time_based --runtime=60s --ramp_time=2s \

--ioengine=libaio --direct=1 --verify=0 --bs=4K --iodepth=32 \

--rw=randwrite –group_reporting=1

For the 990 Pro 4TB drive, average random write speeds for both firmware versions ranged from ~1,100k to ~1,300k IOPS. For the 990 Pro 2TB drive, average random write speeds ranged from ~350k to ~600k IOPS (with the one outlier of ~1,392k IOPS). These results were consistent across both firmware revs (notwithstanding the outlier).

I noticed, however, that, for the 2TB results, the avg IOPS might be significantly below spec, but fio also (but not always) reported max values of ~1,380k IOPS, indicating that the SSD hardware is capable of fast random writes approaching the spec value.

When I modified the above fio command by changing the runtime to 5s and ramp_time to 1s (call it the “burst random write” test), the results changed markedly:

For the 990 Pro 4TB drive, the avg random write speed was 1,391k IOPS; for the 990 Pro 2TB drive, the avg random write speed was 1,393k IOPS. Increasing the runtime to 10s (and ramp_time back to 2s), and the 2TB avg random write speed again dropped significantly below spec to 599k IOPS.

So, here’s what I infer from these results:

** there’s nothing wrong with the 2TB drive hardware.

** the 2TB drive’s SLC cache is being exhausted sooner than that of the 4TB drive – either it’s physically smaller or the drive’s ~50-60% utilization has reduced its capacity. Or the 2TB firmware’s write management is not the same as the 4TB’s: background operations start during sustained random write, garbage collection begins, and write amplification increases, all leading to lower average random write speeds.

** cache state affects the fio testing, causing the range in avg random write speeds that I’ve seen in the 2TB drive

** the 4TB drive might be performing better during the sustained random write test because it has more NAND chips and better heat distribution, because it has more parallel NAND channels, and/or because its firmware handles write management better.

Or something else entirely.

So, what’s the bottom line?

**I don’t think the 7B2QJXD7 firmware caused any degradation in performance that wasn’t already present in earlier firmware;

**burst random write performance of the 990 Pro 2TB drive over 5 seconds roughly meets the spec’ed value (or at least isn’t egregiously lower);

**sustained random write performance of the 990 Pro 2TB drive over 10 seconds or more drops to ~25% of spec, but the worst value I saw (375k IOPS) is still significantly better than the SATA 870 EVO drives I tested or than a hard drive.

**the difference I reported earlier in the Power On Hours from “smartctl -x” for the two drives is still a mystery – if it’s correct, then what’s going on?

What do you think?