Samsung 9100 Pro 2 TB

[csendesmark]

Some of us in the workstation market view no heatsink as a Pro too. I've bought heatsink drives that were (on sale for cheaper than the alternative ) only to then have to shuck them to install in PCIe add-in cards.

Well, this product has "PRO" in it's name, but for actual workstation stuff I maybe you are better off with u.2 drives, which I never used - so can't give real advices.
And sorry for your trouble with the heatsink versions with the AIC adaptor, that must been really bad feeling.

PCI-E 3.0 SSD's are gold for the PC gaming. Still PCI-E 5.0 SSD's are useless and expensive...

This SSD is not aimed for "the games",
For gaming you need capacity and low latency over anything, @Ruru is right, speeds over 3GB/s (PCIe gen3's) are plenty.
Think about RDR2 or the upcoming GTA6

 

[Chrispy_]

For gaming you need capacity and low latency over anything, @Ruru is right, speeds over 3GB/s (PCIe gen3's) are plenty.
Think about RDR2 or the upcoming GTA6

I think most games (at least games that don't use DirectStorage) care more about CPU performance than SSD.

My 9900X loads into game levels slightly faster than my 5800X3D, despite the 9900X having a PCIe 3.0 x2 SSD for the game library and the 5800X3D using a vastly superior WD SN850X in the primary (CPU-connected) Gen4 slot. My guess is that the decompression of compressed assets is CPU-limited and the 9900X is unsurprisingly faster at that, whilst still not decompressing fast enough to outpace the 1800MB/s of the Gen3 drive in an slot with only two PCIe lanes.

 

[_roman_]

Positive: the drive have dedicated DRAM (It'S very sad that this common feature is now a positive point. In my point of view this is what i expect and every drive should have)

Negative: Samsung, Lots of Firmware issues with Samsung, Low quality in the past of Samsung SSDs, outdated windows software to update the drive, pcie 5.0 hoax sticker with barely any performance for pcie 5.0., high price, same low performance as any other drive recently

that drive would be acceptable when labeled and sold as Gen 4. AFAIK gen 4 caps around 7500 MB/s or MiB/S (units does not matter in this case anyway as its close to gen4 )

 

[Chrispy_]

Positive: the drive have dedicated DRAM (It'S very sad that this common feature is now a positive point. In my point of view this is what i expect and every drive should have)

Negative: Samsung, Lots of Firmware issues with Samsung, Low quality in the past of Samsung SSDs, outdated windows software to update the drive, pcie 5.0 hoax sticker with barely any performance for pcie 5.0., high price, same low performance as any other drive recently

that drive would be acceptable when labeled and sold as Gen 4. AFAIK gen 4 caps around 7500 MB/s or MiB/S (units does not matter in this case anyway as its close to gen4 )

That sequential write speed test is single-threaded. You won't see Gen5 speeds on any drive without multi-threaded requests

 

[rusTORK]

Hm... is there any "Full Performance" mode for this SSD in Samsung software? Like previous 990\980 Pro?

 

[PixelTech]

Is 2280 standard too constraining of space for performance people are looking for? Are the components crammed on the PCB?
I've noticed that most mobos M.2_1 slot is between PCIe_1 (GPU) and the CPU. Putting a M.2 drive between 2 hot things. Would it be better if we put the drive on the backside of the mobo? Some mobos do it already, but not for the main slot that has the full bandwidth.
The gaming industry has had time to adopt DirectStorage or equivalent technologies but still hasn't. Do gaming only enthusiasts even buy at these prices?

 

[W1zzard]

Crucial T705

T705 is Phison E26, which is represented by Corsair MP700 Pro in my tests

Micron 4600

Sorry but no, review of another SSD with that controller coming soon

 

[Chrispy_]

Well, this product has "PRO" in it's name, but for actual workstation stuff I maybe you are better off with u.2 drives

U2 is just a connector. It's still NVMe protocol over PCIe lanes.

Realistically, if a product has OPAL hardware encryption, DRAM and uses at least TLC NAND to prevent garbage-tier performance when the cache is exhausted then it's "pro" enough for most enterprise workstations, IME.

Few "pro" users actually work on local storage anyway, company policy will almost always have them working off a network share.

 

[W1zzard]

Oh, I know that. Just trying to understand. I thought queues would share the bandwidth, I didn't realize one queue cannot saturate the bus.

Maybe to help understand, in this context "Queue Depth" means "requests executing in parallel", not "requests waiting for access to a single-execution queue which executes one request at a time"

 

[10tothemin9volts]

1.56W consumption in idle is one of the worst values (and too much for me personally) and 0.2W in idle L1 ASPM isn't that good either, when other drives have 0.05W. In a mobile device this may make a difference. Is this due to SAMSUNG's process node, PCIe 5.0 or what could be the reasons?

 

[Balzoro]

Immature?

Samsung SSD controllers have always been far behind other popular vendors in decompression benchmarks, going back at least a decade. If you want reason and meaning, go and interview some Samsung engineers. Feel free to speculate, but it won't be anything more than speculation which is equally unscientific and presumably unacceptable to you?

I'm just saying what I see across 15+ years of benchmarks from here and places like Anandtech, TechReport, Tom's - and if you don't like casual language then perhaps you ought to stay away from internet forums.

I hope I won’t get banned for this - but there’s just always a person that is so condescending that he/she doesn’t even understand what they’re even arguing about and just want to be right. That is literally (in the most literal sense) you, right now.

In the decompression charts, the 990 pro is significantly faster than the 9100 pro. Same as with most other drives, Samsung or not Samsung. It does not correlate with any of the other benchmark results, and also does not make sense on paper. Again - even compared to Samsung own drives. Even the outdated 860 passed it comfortably.

So instead of just saying ‘Samsung always sucks’ which is immature and inflammatory, it would be useful to understand WHY that result happened. You do understand that for the sake of the argument, this could have been an error, or a hardware issue with that specific SSD, right?

You can tell that someone is a hater by the fact that they talk negatively first about a brand instead of exploring the more possible options that caused that result anomaly

 

[W1zzard]

In the decompression charts, the 990 pro is significantly faster than the 9100 pro. Same as with most other drives, Samsung or not Samsung. It does not correlate with any of the other benchmark results, and also does not make sense on paper. AGAIN - even compared to Samsung own drives.

No improvement, not even with the drive just half full.

https://imgur.com/LJgRUjv

Note how Active Time spikes. This looks like a controller/firmware/design issue, not a testing problem

 

[csendesmark]

U2 is just a connector. It's still NVMe protocol over PCIe lanes.

Realistically, if a product has OPAL hardware encryption, DRAM and uses at least TLC NAND to prevent garbage-tier performance when the cache is exhausted then it's "pro" enough for most enterprise workstations, IME.

Few "pro" users actually work on local storage anyway, company policy will almost always have them working off a network share.

That is true, u.2 is just a connector,
but maybe it is not that clear cut that the m.2 is (generally speaking) for "home users" like us
The u.2 is for enterprise/server environments, where often they not even giving TBW but DWPD or Drive Writes Per Day.
Of course if you are a power user - you might have u.2 in your PC or just have a really kick ass m.2 drive with "extras", the overwhelming majority is just average.
So Samsung's "PRO" tag on this drive is more or less - just good old marketing.
And don't get me wrong, I have 970 PRO 980 PRO and 990 PRO - hopefully 9100 PRO in the near future

No improvement, not even with the drive just half full.

https://imgur.com/LJgRUjv

Note how Active Time spikes. This looks like a controller/firmware/design issue, not a testing problem

Well, it will not the first time for them...
Once I had to RMA my 980 PRO It failed because subpar firmware.

 

[evernessince]

T705 is Phison E26, which is represented by Corsair MP700 Pro in my tests

Both the T700 and T705 are Phison PS5026-E26 drives with a significant performance gap between the two.

By extension, I don't believe the MP700 pro is a good stand in for other drives with the same controller. Heck the comparison should at the very least be against the Corsair MP700 Pro se, which is again an updated version with better performance.

The performance difference of the T705 vs T700 and MP700 Pro vs MP700 Pro Se is greater than the lead the Samsung 9100 Pro has in these charts. 2ns shaved off latency (which is significant when latency is 28ns) and an additional 2 GB/s sequential isn't nothing, especially when you are talking about which drive is the fastest.

 

[Balzoro]

No improvement, not even with the drive just half full.

https://imgur.com/LJgRUjv

Note how Active Time spikes. This looks like a controller/firmware/design issue, not a testing problem

Thanks for checking again. I’m completely stumped on this one given all of the other benchmarks provided. Seems like such a weird anomaly.

I assume it’s coming either from the firmware or thermals (although the latter won’t explain the rest of the benchmarks as it cruised by).

 

[W1zzard]

Corsair MP700 Pro se, which is again an updated version with better performance.

Corsair MP700 Pro SE 4 TB Review

The Corsair MP700 Pro SE solid-state-drive boasts impressive speeds of up to 14 GB/s, thanks to the lightning-fast PCI-Express 5.0 interface. In our review we're testing the 4 TB version, which offers tons of space and incredible performance that's the best of all the drives we tested so far.

www.techpowerup.com

or thermals

it's not the thermals, all drives are active-fan cooled during the performance tests to eliminate that variable

 

[TomWeng]

Looking forward to testing the ZHITAI TiPro9000

 

[Balzoro]

Just went and checked a few other tech blogs - no one seems to be doing a quick and easy zip decompression test. Everyone is mostly obsessed with synthetic benchmarks or gaming (..although that’s also the reason I prefer TPU over others, as the benchmark suits here are more comprehensive!).

I’ll continue keeping an eye out, but if anyone finds a review that shows decompression please let me know/DM. I know it won’t be comparable to the results here, but I’d like to see their 9100 results compared to their own previous bench results with other drives and see if the 9100 is still at the bottom when it comes to decompression or the situation is completely different

 

[low]

Hm... is there any "Full Performance" mode for this SSD in Samsung software? Like previous 990\980 Pro?

 

[QuietBob]

I think most games (at least games that don't use DirectStorage) care more about CPU performance than SSD.

True, although those that use DS are still limited by CPU/RAM performance. The default implementation -- which is what we've seen in games so far -- relies on the CPU for asset decompression.

I believe there have been only two or three titles that actually used the GPU for this purpose, but I can't recall their names.

 

[Guwapo77]

Will you be reviewing the Crucial T705? I bought that one two days prior the "surprise" launch of the Samsung 9100.

 

[Makaveli]

drive is kinda meh.

Now that Sandisk owns WD I really hope we still see something based on this and they don't just cancel it.

Western Digital Previews M.2 2280 PCIe 5.0 x4 NVMe Client SSDs: 15GBps at Under 7 Watts

www.anandtech.com

 

[AsRock]

Price is a killer, glad there is a option for heatsink or not as i like to get mine without as it allows me to better the temps.

 

[chrcoluk]

Looks like there is no improvement to NAND quality vs the Samsung gen 4 pro drives as SLC exhausted write speeds are inferior to both the 980 pro and 990 pro. It somehow manages to get an overall lower speed despite the advantages of gen 5 to help pSLC portion for the write test. However they bumped the size of the pSLC meaning its now less likely to see the lower performance. As someone else pointed out its barely above gen 4 cap for writes, so most of the pSLC improvement is probably controller at expense of the extra heat and power draw.

I think gen 4 drives are still king.

 

[QUANTUMPHYSICS]

$300? I'm getting a 4TB Nvme