Team Group GE Pro 2 TB

[Mr. Perfect]

And games still don't use the extra bandwidth. I'm a little disappointed DirectStorage isn't getting used yet to really make these faster drives worth wile for gaming.

 

[W1zzard]

DirectStorage isn't getting used yet

Ratchet and Clank does use DirectStorage, but as you can see, no significant difference in load times

I wont only a comparison between Phsion E26 and the Inno with the same heat sink, that would be great.

I will add that, no worries, just waiting for the heatsink

 

[InVasMani]

It'll be interesting to see how DirectStorage is with a few newer generations of GPU's that handle the decompression more quickly.

 

[chrcoluk]

Anyone who needs high performance with fsync should seek an enterprise SSD with actual hardware-based PLP, such as the Kingston DC2000B, announced just yesterday at TPU. Those have their PCBs filled with capacitors, hence fewer NAND chips and lower memory capacity. Other SSDs have software-emulated PLP, it probably works, but fsync IOPS is its very weak point.

Also, database engines can recover data from if the power was cut, and writing interrupted, at any given moment. Only uncommitted transactions are lost, as they should be. (I know that about Oracle but other DBs keep transaction log files with the same purpose.) That's not even the greatest danger that fsync guards against. The crucial part here is that data has to be written to permanent storage in the same order the server sent it for writing. If a certain sector is written but one that arrived earlier is stuck in some cache and lost, the mess is complete. I think the same is true of journaling filesystems such as NTFS. This is what fsync guards against.

And of course, power loss is just unacceptable in any application that juggles valuable data, financial or not.

I get your point, but I definitely like that this test was added, and even across the consumer drives tested we see quite a big variance of fsync performance.

 

[ARF]

And games still don't use the extra bandwidth. I'm a little disappointed DirectStorage isn't getting used yet to really make these faster drives worth wile for gaming.

Probably the best is to still stay with the old mechanical discs HDD... I wonder what the loading times are with those...

 

[Hagal77]

I will add that, no worries, just waiting for the heatsink

Best man, big THX!

 

[Wirko]

I get your point, but I definitely like that this test was added, and even across the consumer drives tested we see quite a big variance of fsync performance.

It's always good when another data point is added. But it should be thoroughly explained that fsync isn't for everyone, by far. The numbers make some SSDs (Samsung, Lexar) look bad when they really aren't, at least not because of fsync performance. At the other end there are the DRAM-less WDs, suspiciously too good in that regard. (And then they can't handle ZFS.) I'm sure there will be discussions here like "what ssd should i buy?" -> "avoid the 990 pro, it's so poor at fsync".

It'll be interesting to see how DirectStorage is with a few newer generations of GPU's that handle the decompression more quickly.

Hm, I think this will just (further) resolve the bottleneck that really isn't. The bottleneck is probably mostly because of random read IOPS, which can only be improved if the software uses queueing more. But DirectStorage API also has some provisions for queueing. It can't improve anything by itself but maybe it makes developers' work easier.

 

[W1zzard]

But it should be thoroughly explained that fsync isn't for everyone, by far

"This provides important integrity guarantees that are used by databases like MySQL, SQL Server, high-availability filesystems and the etcd service that's the backbone for all Kubernetes clusters."

that's not sufficient?

 

[Assimilator]

Good god that is absolute rubbish. Innogrit and Team Group should be ashamed of releasing a product in such an obviously broken state.

 

[W1zzard]

Thermal testing has been added: https://www.techpowerup.com/review/team-group-ge-pro-2-tb/8.html

 

[Wirko]

Thermal testing has been added: https://www.techpowerup.com/review/team-group-ge-pro-2-tb/8.html

The GPU fans probably don't create any significant airflow around the SSD heatsink, right? But if they do, it could be benefical (and maybe even possible) to associate the fan control with the SSD temperature. It's a very big heatsink, inefficent in still air but even slight air movement might prevent thermal throttling in a setup like yours.

 

[edlee]

thanks for testing with heatsink, all i can saw wow, even with a large heatsink gen5 ssd's still need a fan if you are hammering at them to prevent thermal throttling.