Your PC is a few years old. It stutters during Steam updates, occasionally throws a full freeze that forces a hard reboot, and task manager shows 100% disk usage during what should be routine background activity. Your instinct is to start pricing out a new build.
Hold off for a moment.
There’s a specific failure pattern that shows up on older systems - particularly those with QLC NVMe drives running near full capacity - that looks exactly like the symptoms above. It’s not your CPU aging out. It’s not your RAM failing. It’s your storage drive throttling under load because of how QLC NAND behaves when the drive is nearly full.
Replacing the SSD costs a fraction of a full rebuild. And in a lot of cases, it fixes the problem completely.
This guide explains the pattern, how to tell if it applies to your system, and what to do about it - including a real-world example and practical SSD recommendations if you decide a replacement is the right call.
What This Failure Pattern Looks Like
The symptoms tend to cluster together:
- 100% disk usage in Task Manager during Windows updates, Steam game installations, or antivirus scans - even when the drive should have plenty of headroom to handle it
- System freezes or full lock-ups that last 10-60 seconds before the system recovers, or that require a hard reboot
- Sluggish app launches and slow file copies that seem disproportionate to the task
- Crashes during write-heavy operations - game installs, large file transfers, OS background indexing
- Drive performance that degrades over time, even though the hardware is “working”
The freezes feel random because they don’t happen every time - they happen when the drive is being pushed hard and the conditions that cause QLC throttling are met. That intermittency makes it easy to blame drivers, Windows updates, or general “old PC wear.”
Why QLC + High Fill Level Causes This
Most affordable NVMe drives sold in the 2019-2022 era - including the widely used Intel 660p - use QLC (Quad-Level Cell) NAND. Each cell stores four bits of data, which lets manufacturers pack a lot of storage density into a single chip at low cost. The trade-off is endurance and sustained write performance.
To paper over the inherent write speed limitations of QLC NAND, drive manufacturers implement an SLC write cache: a portion of the drive that behaves like faster SLC (Single-Level Cell) NAND to absorb incoming writes quickly. Under normal conditions, this works fine - the cache handles the burst of writes, and data gets migrated to the main QLC NAND in the background when the drive is less busy.
The problem is what happens when the drive is nearly full.
When available free space drops significantly - particularly below roughly 15-20% on most consumer drives - the SLC cache either shrinks dramatically or stops functioning altogether. There’s simply not enough free NAND to maintain the cache. With no SLC buffer, writes land directly on the QLC cells, which are meaningfully slower. On the Intel 660p, sustained sequential write speeds can fall from 1,800 MB/s (cached) to under 100 MB/s once the cache is exhausted - a collapse of more than 18x.
Now layer on a Steam update that’s writing several gigabytes of game data, or Windows doing background index work, and the drive simply can’t keep up. The OS stalls waiting on writes. Task Manager shows 100% disk utilization. The system freezes.
This isn’t a defect. It’s the drive behaving as designed under conditions it wasn’t optimized for. The likely cause of these crashes is QLC cache saturation combined with high fill levels - but it’s worth noting it isn’t a guaranteed cause of every freeze. RAM instability, thermal throttling, and bad drivers can produce similar symptoms. The distinction matters because misdiagnosing a RAM or thermal problem won’t be fixed by a new drive.
The Intel 660p Specifically
The Intel 660p was one of the most popular value NVMe picks of its generation. It offered competitive sequential speeds on paper, reasonable pricing, and the Intel brand name. Millions of these shipped in prebuilt systems and as aftermarket upgrades.
The 1TB model has a rated TBW (Terabytes Written) of around 200 TB - which is modest by today’s standards and reflects the QLC endurance trade-offs. In our CrystalDiskInfo readout, the drive was around 40 TB host writes (how much has been written so far), which is different from the full rated TBW limit. More practically, the 660p’s SLC cache behavior is aggressive: it starts narrowing the cache noticeably once the drive passes the 60-70% full mark, and at 85-90%+ utilization, sustained write performance falls off sharply.
For systems where the 660p is the only drive and it’s serving as both OS boot volume and game library storage, the math gets tight fast. A 1TB drive with Windows, apps, and a few modern games can easily sit at 700-850 GB used. At that fill level, you’re routinely operating in the zone where the cache can’t reliably protect performance.
My Case: When “Good Enough” Stopped Being Good Enough
I had been running an Intel 660p 1TB as the sole drive on an older system for a few years. Performance had always been acceptable - until it wasn’t.
The crashes started showing up during Steam updates. A game update would kick off, disk utilization would spike to 100%, and the system would either freeze for a minute while the drive caught up or lock up entirely and require a hard reboot. At first it seemed like a Steam issue. Then it looked like a Windows problem. Drivers were updated. Nothing helped.
The drive was sitting at around 90% capacity.
A friend described an almost identical pattern: recurring freezes under high-capacity usage, always during write-heavy operations, always on a system with a well-used NVMe drive near full.
The fix was straightforward. I installed a WD Black 4TB SSD and used it to offload the Steam library and bulk data storage. The original 660p stayed as the OS boot drive - but now it had meaningful free space. The crashes stopped.
The lesson isn’t necessarily “throw out the 660p.” The lesson is: a QLC boot drive running near capacity is a liability. Moving the write-heavy workload off the primary drive - or replacing the drive entirely if it’s aged - solves the problem.
Other Things Worth Ruling Out First
Before committing to a drive replacement, it’s worth doing a quick sanity check on other common culprits that produce similar symptoms:
- RAM stability: If you’re running XMP/DOCP memory profiles and haven’t stress-tested stability, run Memtest86 or Windows Memory Diagnostic. Unstable XMP can produce intermittent freezes indistinguishable from storage issues.
- Thermals: Check CPU and GPU temperatures under load using HWiNFO64 or similar. Thermal throttling can cause frame stutter and slowdowns, though it typically doesn’t produce 100% disk usage symptoms.
- Drivers: Outdated storage controller drivers or a recent problematic Windows Update can cause disk-related issues. Check Device Manager for any flagged devices.
- Drive health: Use CrystalDiskInfo to check the SMART data on your existing drive. Look for reallocated sectors, high error counts, or a remaining life/endurance indicator below 50%.
If those checks come back clean and the symptoms align with the QLC/high-fill pattern - particularly if your drive is over 80% full - the drive is the most likely explanation.
Action Plan: Fix It Step by Step
Step 1: Check Drive Fill Level and Health
Open CrystalDiskInfo (free download) and check:
- Current fill level (or check in Windows Explorer - right-click the drive, Properties)
- SMART health status - look for “Good” and note the remaining life percentage if shown
- Any flagged attributes
There are a few CrystalDiskInfo theme editions available. We used the MSI Mei Mihoshi Edition because (A) we like MSI, and (B) we’re weebs. Same core diagnostics, just better vibes while troubleshooting.
If your drive is above 80% full and QLC-based (660p, 670p, SN550, or similar), you’re in the risk zone regardless of health indicators.
Step 2: Move Your Game Library Off the OS Drive
This is the lowest-cost first step. Steam, Epic, and most game launchers support multiple library locations.
In Steam: Settings → Storage → Add Drive, then move games to a secondary drive or a new storage drive you’ve added.
If you don’t have a secondary drive, a QLC drive is actually fine for this use case - game storage is mostly sequential reads, which QLC handles well. It’s write-heavy workloads on a nearly-full boot drive that create the problem. A budget QLC 2TB drive for pure game storage is a reasonable and inexpensive option.
Step 3: Keep the OS Drive Well Under Capacity
The comfort target for QLC boot drives is below 50% used. That keeps the SLC cache healthy and prevents the write performance collapse. Avoid sustained operation above 85-90% on any QLC drive that’s doing real write work.
For reference: Windows 11 + standard apps + a few small programs typically occupies 60-80 GB on a modern installation. A 1TB boot drive with only OS and apps leaves plenty of margin. The mistake is letting game libraries and bulk downloads accumulate on the boot drive over time.
Step 4: Replace the Drive If Symptoms Persist or Endurance Is Low
If your drive health is degraded, your TBW endurance is running out, or simply moving data off doesn’t fully resolve the crashes, replacement is the right call. The good news is that modern TLC NVMe drives are significantly faster, more durable, and not expensive.
A new drive + clean Windows install is a much lower-cost path than a full platform upgrade. If the crashes are storage-driven, a new CPU won’t fix them.
Replacement SSD Picks
These are quality NVMe options across price ranges. All use TLC NAND, which handles sustained writes meaningfully better than QLC and doesn’t carry the same fill-level sensitivity.
High-End
WD Black SN850X The current benchmark for consumer NVMe performance. Gen4 PCIe, consistently strong sequential and random read/write speeds, and a proven track record. If you have a Gen4-capable platform (Ryzen 5000/7000 or Intel 12th gen+), this is a top-tier pick with no meaningful trade-offs.
Buy WD Black SN850X on Amazon →
Samsung 990 Pro Samsung’s flagship consumer NVMe. Competes directly with the SN850X, with slightly different performance profiles depending on the workload. Either pick is excellent; it often comes down to current pricing. The 990 Pro is often the better price-per-GB option at 2TB and 4TB.
Buy Samsung 990 Pro on Amazon →
Value TLC Pick
WD Black SN770 The strongest value TLC option at its price point. Gen4, reliable performance across real-world workloads, and well below the cost of the high-end picks. If you don’t need top-tier sequential speeds, the SN770 is a smart choice for a replacement boot drive or secondary storage. No DRAM cache, but real-world performance is strong for the price.
Buy WD Black SN770 on Amazon →
Secondary/Game Storage Note
For a purely read-dominant game storage drive, QLC is acceptable. If you’re adding a bulk storage drive specifically to offload your Steam library, a QLC drive at a good price-per-GB is a reasonable budget option - just don’t use it as your OS boot volume under heavy load.
How This Fits Into a Broader Upgrade Decision
If you’re on an older Intel platform (8th-10th gen) and asking whether to replace the storage or the whole system, the storage-first path has a clear logic: it addresses the most likely cause of your symptoms at the lowest possible cost. If a new drive fixes the crashes, you’ve bought yourself another year or two on the existing platform while you save for a proper rebuild.
A new CPU and motherboard won’t make your QLC boot drive perform better. And if the crashes are storage-driven, the platform upgrade won’t fix them.
That said, if you’re already planning a platform refresh for other reasons - gaming performance, RAM capacity, software compatibility - a storage upgrade at the same time makes sense. The drives recommended above will carry over to any new build.
For related upgrades that complement a refreshed desk setup, see our guides on the best wireless mice for work and gaming and the best monitor arms for dual-monitor setups. If you’re considering building out a home server to offload bulk storage entirely, our home server build guide covers that path in detail.
FAQ
Do I need a new CPU first?
Not if the symptoms are storage-driven. High disk utilization, write-triggered freezes, and slow installs are not CPU symptoms. A new processor on the same system with the same full QLC drive will produce the same crashes. Confirm the storage is the issue first before committing to a platform upgrade.
How full is too full for an SSD?
For QLC drives under sustained write load: above 80% is genuinely risky, and 85-90%+ is where the performance cliff appears. The ideal operating range for a QLC boot drive is below 50% used. For TLC drives, the threshold is less severe, but the same principle applies - leaving 15-20% free is good practice.
Can I keep the old SSD as a backup or secondary drive?
Yes, with caveats. If the drive health is good (CrystalDiskInfo showing “Good” with no flagged attributes and reasonable endurance remaining), it’s fine as a secondary read-dominant drive. Avoid using it for write-heavy tasks. Don’t rely on it as your sole backup medium - back up to a separate drive or cloud storage. If SMART shows degradation, retire it from active use.
Will reinstalling Windows fix the problem?
Potentially, if the drive fill level was the main issue and a fresh install gets you well below the performance cliff. A clean install typically uses 30-50 GB. If the drive is 1TB and you start fresh, you may recover normal performance without buying anything. That said, if the drive health is declining or you’re going to reinstall the same game library anyway, a new drive combined with a clean install is the cleaner long-term fix.