Slow intel 3500?

[Noetic]

I recently purchased an 800GB intel 3500 with low hours off of ebay. CrystalDiskMark was topping out at 410MB/s on reads and 460MB/s on writes. As a control a Samsung 840pro ran at 550MB/s and 510MB/s respectively on the same circuit.

Perhaps it was a problem with this particular drive, so I RMAed it to intel and they sent me a brand new unit. Unfortunately, yielding the same results. I test the drive on two other machines, all with different flavors of Win7 64.

ATTO did display a read peak at 500MB/s, but HDtest concurred with CDM.

Has anyone else had similar experiences with this model?

-David

 

[Kaos_Drem]

You might want to try some other benchmarks as well, I've got a bit of experience with both the 730 series and the S3500 that it was based on and CDM and Atto are not always what I would trust implicitly for results. Try AS SSD, Anvil Storage Utilities, and maybe HDtune. I'm not however going to dispute that the numbers you have are low, but the 840pro and S3500 are targeted at entirely different markets.

 

[-Dragon-]

The point of these drives isn't throughput it's latency. They aren't better in a consumer system with low QD they're for servers with high QD.

 

[Noetic]

I do understand the targeted market for the s3500, however the benchmarks intel claims puts its throughput at 500/450 MB/s. It is 20% off on its reads. I was wondering what might be the cause of it. All the intel tuning apps have been run on it as well as the firmware upgrade. HDTune results were no better, BTW.

Truth to tell, I can live with it. For the 350 I paid, having the reliability of a new intel 800GB drive is a good deal.

 

[doz]

Are you running on the Intel SATA port w/ Intel RST driver installed?

 

[Noetic]

Good question. It was first install in a Dell inspiron 17r special edition. I assumed that it was an Intel SATA port. As for the Intel RST drivers, I assumed that the Intel tool kit would have install them. With the other systems, they might have been a non-Intel SATA ports. I will check.

 

[-Dragon-]

Did you secure erase it? Shouldn't really affect reads but who knows. You mentioned low power on hours but didn't mention total TB written, would be hard to get near the warrantied endurance of 460TB, but $350 for a >$800 drive that's less than a year old (and supposedly barely used) is awfully cheap for there to be nothing wrong.

 

[Noetic]

When purchased, it was powered on less than 70 times (using intel tools), i do not recall the amount written, though it was low. Intel gave me a brand new one as an RMA. Moreover, this is why I am curious why it reads 100MB/s less than spec, and if there is anything I can do to remedy it.

as for the price, it was an ebay auction of a low mileage demo. The same fellow sold the superior s3700 800GB for $530.

 

[doz]

Have you tried the drive on your desktop to check speed? I dont ever trust crippled laptop motherboards.

 

[Noetic]

Have you tried the drive on your desktop to check speed? I dont ever trust crippled laptop motherboards.

It took the place of a Samsung 840 Pro 256GB, which yielded the typical top notch response (550/510 MB/s respectively). So far, this hold true with now four different machines. Since this is a replacement for a drive with the same numbers, I was wondering if this is typical of these drives.

 

[darkunitzero]

as someone else mentioned here. the 3500 was meant for dc usage. have you tried testing with something like iometer and increasing the qd to 32+?

 

[Noetic]

I understand that the drive was designed for servers, still why would it not display Intel specification under lighter load. It might be interesting to see which test they use to make their determination.

 

[drescherjm]

I would benchmark it with a queue depth of > 32 perhaps 64 or 128.

 

[darkunitzero]

I understand that the drive was designed for servers, still why would it not display Intel specification under lighter load. It might be interesting to see which test they use to make their determination.

Maybe you didn't know what to look for at the time you purchased the SSD?

http://www.intel.com/content/dam/ww.../product-specifications/ssd-dc-s3500-spec.pdf

Take a look at table 4.

Performance measured using Iometer* with 128 KB (131,072 bytes) of transfer size with Queue Depth 32.

Sequential Read (SATA 6Gb/s) - 500
Sequential Write (SATA 6Gb/s) - 450

Edit: If you plan on running the test with IOmeter, remember to secure erase the drive, calculate the test duration time, then run the test until you hit steady state.

 

[Noetic]

Maybe you didn't know what to look for at the time you purchased the SSD?

http://www.intel.com/content/dam/ww.../product-specifications/ssd-dc-s3500-spec.pdf

Take a look at table 4.

Performance measured using Iometer* with 128 KB (131,072 bytes) of transfer size with Queue Depth 32.

Sequential Read (SATA 6Gb/s) - 500
Sequential Write (SATA 6Gb/s) - 450

Edit: If you plan on running the test with IOmeter, remember to secure erase the drive, calculate the test duration time, then run the test until you hit steady state.

Thank you for the clarification. It is clear that the numbers come from a different modality, I confess that I am not as well versed as I should be with server throughput. I have numerous SSDs, and thought little of the difference, jumping on what appeared to be a good deal.

What I do find striking, is that the Intel personnel did not have the answer when I posed it to them. In fact, the issue was escalated to a higher level and then they decided to give me a brand new drive.

Well, I do appreciate the education, thank you.

 

[darkunitzero]

Np, we learn something new every day.

As for the Intel personal, if you were just talking to support, I highly doubt that they would know use case scenarios for the SSD.