Western Digital's 4 TB WD4001FAEX Review: Back In Black

[machinegundom]

I still haven't use half of my 1TB, but having this extra space would be nice.

 

[murphyslaw1978]

Guys, can someone help me out with this? For one of the benchmarks - streaming 4KB reads - these drives are pushing well over 100MB/sec. If you run the IOPS conversion (IOPS = data*IOsize), then you have 25,000 IOPS (100,000KB*4KBIO size). So doesn't think mean that for this specific, particular test, the drive is pushing 25,000 IOPS? I ran a similar test using IOmeter on my laptop with a 7200 RPM WD 500GB drive. I can push 20MB/sec with 4KB streaming reads all day long. IOmeter shows around 5100 IOPS. Am I looking at this right?

 

[blazorthon]

[citation][nom]murphyslaw1978[/nom]Guys, can someone help me out with this? For one of the benchmarks - streaming 4KB reads - these drives are pushing well over 100MB/sec. If you run the IOPS conversion (IOPS = data*IOsize), then you have 25,000 IOPS (100,000KB*4KBIO size). So doesn't think mean that for this specific, particular test, the drive is pushing 25,000 IOPS? I ran a similar test using IOmeter on my laptop with a 7200 RPM WD 500GB drive. I can push 20MB/sec with 4KB streaming reads all day long. IOmeter shows around 5100 IOPS. Am I looking at this right?[/citation]

You're comparing sequential throughput to random throughput. They aren't the same. IOPS is generally used to measure random throughput, something that hard drives do far worse than they do sequential throughput.

 

[murphyslaw1978]

It is not out of the ordinary to assume that any brand new SSD drive on the market (or an array of 100 SSDs on an EMC SAN for example) cannot due either random or sequential of 5100 IOPS at 4KB block. So with everything else the same or even, the IOPS would be the same. Can anyone disprove this?

 

[blazorthon]

[citation][nom]murphyslaw1978[/nom]It is not out of the ordinary to assume that any brand new SSD drive on the market (or an array of 100 SSDs on an EMC SAN for example) cannot due either random or sequential of 5100 IOPS at 4KB block. So with everything else the same or even, the IOPS would be the same. Can anyone disprove this?[/citation]

The drive in this article is a hard drive, not an SSD. The same is true about your drive.

IOPS is data transfer rate divided (not multiplied) by transfer block size.

Any new SSD on the market these days pushing a mere 5KIOPS is very slow. Even half-decent models should push several tens of thousands of IOPS. Hard drives, on the other hand, are much slower with random throughput.

If you're comparing your hard drive to the drive in the article and asking why there is a huge discrepancy in performance, then maybe your hard drive is simply a much slower model. It has much lower capacity (and thus lower data density) and might be an otherwise older model with inferior performance. For example, if two hard drives are equal in every way except for capacity, then the higher capacity drive would have approximately greater performance as found by the percentage increase in the square root of each drive's capacity. There can also be differences in other ways that affect the drive's performance such as the cache capacity, cache performance, and much more.

SSDs have far greater performance than hard drives and it would be unreasonable to think that any new SATA SSD on the market can't push a mere 5K IOPS in 4KB block sizes. Even a half-decent model should be able to push several tens of thousands of IOPS at that block size.

If this doesn't answer your question, then I'm not sure of what it is you're asking.

 

[murphyslaw1978]

I agree with everything you said. By your definition of IOPS (data transfer rate divided by transfer block size), then my crappy, slow SATA drive that reads 20,000KB/sec divided by 4KB block size, then the IOPS is 5100. Notice, I said nothing about streaming versus random because that has nothing at all to do with the mathematical formula of IOPS. Trust me, I'm not comparing the slow SATA drive to a modern SSD in terms of random performance. But everyone is saying that SATA drives can do 100-150 IOPS and SSDs can do way more. My point is, with streaming workloads, that's NOT the case.

 

[Touko]

The reviewers wish at the end of the review would have already happened if not for the total regulatory fail allowing a duopoly in the hard drive market.

Now they can just call each other on the phone and milk everybody like cows.

Until by some miracle SSD's break the 4TB barrier and people go to burn down SG and WD headquarters.

 

[mart666]

Yes the seagate are faster, but i saw on NewEgg that Seagate have a lot more DOA drives or failling drivers shorter after buying than Western Digital.

To advoid loosing time with defective drives i ordered two of these drives (WD4001FAEX)