Don’t buy a 1TB PCIe 5.0 SSD

The first PCIe 5.0 SSDs for consumers came out early this year, capping out at 10,000MB/s, but now we’re already getting several PCIe 5.0 drives capable of 12,000MB/s. These second-generation PCIe 5.0 SSDs are further pushing what’s possible with the best PCIe 5.0 SSDs, and Gigabyte is entering the fray with its Aorus Gen5 12000, named after its claimed speed of 12,000MB/s.

While I have no doubt in my mind that the Gen5 12000 2TB is likely tied for first with other second-generation PCIe 5.0 SSDs, the same can’t be said for the 1TB model, which is what I got for this review. The lower capacity model of the Gen5 12000 is fast, but not as fast as its 2TB counterpart, and it’s even slower than first-generation PCIe 5.0 SSDs in many areas. Due to much worse bang for buck in performance and capacity, it’s hard to recommend any 1TB PCIe 5.0, including the Gen5 12000 1TB.

About this review: Gigabyte sent me the Aorus Gen5 12000 1TB for the purposes of this review. Gigabyte did not see the contents of this review before publishing.

Gigabyte Aorus Gen5 12000: Pricing and availability

The Gen5 12000 has been available since September, and Gigabyte offers a 1TB model (the one I’m reviewing) and a 2TB model, which cost $160 and $270 at the time of writing. That’s on par with Crucial’s T700 and Teamgroup’s Z540, which are also second-generation PCIe 5.0 SSDs rated for 12,000MB/s. 10,000MB/s SSDs are also priced almost the same, but of course they offer much less performance, which is a point in the Gen5 12000’s favor.

While many PCIe 5.0 SSDs either offer a model with a heatsink or one without, or come with a heatsink preinstalled by default, Gigabyte has merely included a fairly high-end heatsink that you can install yourself if you want. This heatsink is actually made of copper heat pipes and metal fins, unlike the heatsink you get on drives like MSI’s Spatium M570 or the heatsinks that come with motherboards.

How the Gigabyte Aorus Gen5 12000 was tested

To test the Gen5 12000, I put it in my Intel test bench, which has the Core i9-14900K, ASRock’s Z790 Taichi Lite, and 32GB of DDR5 RAM running at 5,600MHz and CL40 timings. Testing on an Intel CPU helps hit the maximum possible performance on storage devices, as AMD hardware right now struggles a little with these faster PCIe 5.0 SSDs, though hopefully this changes in the future. As for software, everything was updated as of Oct. 17 and all tests were performed on the most up-to-date software at that time.

My benchmarks of choice for testing the Gen5 12000 (plus three of its rivals) are CrystalDiskMark, 3DMark’s storage test, and IOMeter, which each respectively show theoretical maximum performance, performance in games, and sustained writing speeds. In my opinion, these are the three main things you’re likely to care about in an SSD, and these benchmarks should give you a good idea of where each SSD stands.

The Gen5 12000 comes with a custom heatsink, but I didn’t use this for testing, instead relying on the heatsink that comes with the Taichi Lite. I did however test this heatsink separately and will share the data later on.

Performance

CrystalDiskMark comes with six tests with preset parameters for the type of workloads (sequential or random), block size, queue depth, and thread count. In the real world, these values depend on the software, so testing such a wide variety of tests should give us a decent idea of what the Gen5 12000 and other drives can do.

Scores are organized by read/write and are measured in MB/s.

Although the Gen5 12000 1TB is faster than the FireCuda 540 2TB and Spatium M570 2TB in sequential reads, it actually loses to those two first-generation drives as well as the Z540 2TB in sequential writes, and not by an insignificant margin. It also racks up a loss in the high thread count random test, which is a bit more niche admittedly.

These results are a big reason why it’s hard to recommend the 1TB variant of the Gen5 12000. In many cases, an SSD’s maximum performance is limited to higher capacity models, and this is definitely true for the Gen5 12000 and perhaps 12,000MB/s rated drives in general. The Z540 1TB is another SSD with similarly low performance compared to its larger 2TB counterpart.

The storage test in 3DMark is very much unlike CrystalDiskMark as it tests real games and how they load and save games, including Overwatch and Battlefield V. SSDs score more points by transferring data faster and at lower latency.

Despite being disadvantaged in CrystalDiskMark, the Gen5 12000 1TB pulls out a win here in 3DMark’s storage benchmark. While being 1TB clearly has a negative effect in CrystalDiskMark, here it’s not so much of a problem, though I do suspect that 3DMark’s storage test is a bit out of date given DirectStorage (which isn’t tested here at all) is on the horizon.

Lastly I have IOMeter, which is an old benchmarking program that is useful for measuring the performance of SSDs in long writing workloads. These kinds of workloads can wear down SSDs and cause them to slow down even after just a few minutes. One reason for this is that SSD storage is composed of two parts: a small but fast cache and the rest of the SSD, which is fairly slow. Even a short amount of writes can deplete this cache and lower performance. Plus, filling up an SSD reduces its writing performance in general, which just compounds the whole cache issue.

This first benchmark in IOMeter tests the four SSDs when they’re 50% full, and it lasts for 15 minutes.

Scores are measured in MB/s.

Unfortunately for the Gen5 12000 1TB, its performance here is by far the worst. For about a minute it was able to sustain its roughly 9,500MB/s write speed until it plummeted to less than 2,000MB/s, ending with an average of 2,082MB/s over the course of the 15-minute test. It’s nearly half the speed of the FireCuda 540 on average, a first-generation PCIe 5.0 SSD. I would assume the 2TB model would however be at least as fast as the Z540 if not on par with the FireCuda 540.

This next benchmark shows performance at 10%, 50%, and 90% full, with the graph focusing on just the Gen5 12000 1TB and the table showing every other SSD.

At 10% full, the Gen5 12000 1TB looks more acceptable, though you can see around the 5-minute mark the Gen5 12000 started to perform strangely. Still, it ended up averaging 8,262MB/s, which is fine enough even if it is only in third place, and a distant third at that. Still, it’s just not good that the Gen5 12000 can lose to a previous generation SSD like the FireCuda 540 in this area, and the M570 even manages to come out ahead of the Gen5 12000 in the 50% and 90% filled tests too.

Thermals are an interesting thing to look at with the Gen5 12000 1TB since it comes with its own heatsink, and I wanted to test it and see if it actually improved anything. PCIe 5.0 SSDs get very hot and at minimum need a passive heatsink, and although the Gigabyte heatsink isn’t actively cooled, it’s clearly much higher-end than the heatsink that comes with the Z790 Taichi Lite or any motherboard for that matter. However, since filling up a drive also has significant performance consequences, thermal throttling will only really matter when a drive is almost empty.

In IOMeter with the drive filled up 10%, the Gen5 12000 peaked at 80 C roughly five minutes in, which was also the point where the drive started to suffer performance loss. With the custom heatsink installed, the Gen5 12000 hit just 69 C and averaged at 9,404MB/s, much higher than when it ran under the Taichi Lite’s heatsink. The results with 10% of the drive filled illustrates pretty clearly that even the 1TB version of the Gen5 12000 will thermal throttle without good cooling, meaning you might not want to use your motherboard’s passive cooling solution if you want the best performance.

However, this heatsink is really tall, and could prove to be cumbersome for PC building. But thankfully, there’s no fan on this cooler; it’s just fins and heatpipes, making it an attractive alternative for those who don’t want to listen to a tiny, loud fan. Still, you can definitely decide to not use the custom heatsink, as it’ll only improve performance when your drive isn’t filled up very much at all, and even with thermal throttling it’s not like the Gen5 12000 is useless.

Should you buy the Gigabyte Aorus Gen5 12000?

You should buy the Gigabyte Aorus Gen5 12000 1TB if:

• You want higher-end PCIe 5.0 SSD performance
• You like using a custom heatsink to decrease temperatures
• You’re OK with keeping this SSD mostly unfilled for maximum performance

You shouldn’t buy the Gigabyte Aorus Gen5 12000 1TB if:

• You want better bang for buck and can afford the 2TB model
• You want even better performance and are willing to wait for 14,000MB/s PCIe 5.0 SSDs
• You need more storage than just 1TB

While I’m sure Gigabyte’s Aorus Gen5 12000 2TB is pretty fast, the simple truth is that the 1TB model is not even close. It only won in a single test, tied some others, and lost in many. The bang for buck of the Gen5 12000 1TB is already worse compared to the 2TB model and other 2TB PCIe 5.0 SSDs in terms of capacity, but it also being worse in performance makes it hard to recommend. However, the Gen5 12000 1TB definitely wins some points with its great custom heatsink and top-end gaming performance.

If you are looking for a 12,000MB/s SSD, the Gen5 12000 2TB is what you’ll want, assuming you can afford it. You can also try waiting a few months for the upcoming 14,000MB/s PCIe 5.0 SSDs, which will definitely be faster than both the 1TB and 2TB models of the Gen5 12000. The 1TB version of the Gen5 12000 has a tough time justifying its existence given the circumstances, though at least I can’t say it’s a bad SSD — it’s just a worse version of a good SSD.

Source: Gigabyte

Gigabyte Aorus Gen5 12000

Fast but not good value

$155 $160 Save $5

Gigabyte’s Aorus Gen5 12000 SSD is rated for reads and writes of up to roughly 12,000MB/s. You can get it in either 1TB or 2TB capacities, and it comes with a passive heatsink.