Spod, on 03 March 2010 - 01:50 PM, said:
Hmm, 1 page review... I like it, clicking through lots of pages (and waiting for them to load) is a pain compared to just scrolling down.
You will only see one page reviews here. If there's a reason to load a bunch of benchmarks or something on a second page, we'll do that if needed. I hate the 10 page reviews some sites do.
Quote
So, this is an official SR review, not just a "reader review", right?
Reasonable spread of benchmarks, though real world desktop usage was very notable for its absence - I think that is a critical necessity if you want to maintain the reputation of SR for being better than everyone else's... erm... storage reviews (sorry).
Give me the numbers! The graphs and comparative comments are all well and good, but for comparison with other similar designs, the actual numbers need to be available. Label it on the graph, or chart it, even just stick it in the Performance Database (call it Testbed 5).
Yes this is an official StorageReview review. This is the layout going forward, just incomplete in its current form. As you can see comparisons are weak with our first “official” internal drive review but that will change quickly as more reviews are published. It is a chicken and egg problem right now with only one published review to pull numbers from. Since this is also a new test bed for StorageReview we couldn’t pull in data from past reviews either. Trust me when I say it was pretty awkward and embarrassing to have only a single drive plotted on those charts. Give it a month and we will have plenty of datasets to include in future reviews.
The hardest thing so far has just been figuring out what tests we will run and retesting drives over and over again to make sure we have consistent results. The benchmarks shown in this review represent less than half of our final goal. Real-world benchmarks will be expanded in upcoming reviews. Right now it is limited to a simple folder transfer and the rest are synthetic benchmarks. We have finally worked out a method for capturing low-level disk I/O which was the major stumbling block up until now. With that we can finally add in real-life scenarios which are discussed in our new test bed thread. Expect to see those soon. Once we prove internally that our method gets repeatable benchmarks across multiple drives we will make those public and add them to reviews. We don’t want to jump the gun and have to correct it later on if we find an error in our method. Once things are set in stone if we still have older drives in the office, those will be back-tested and the review will be updated with that information.
Long-term plans are for the performance database to be updated with tests on our new Dell XPS 9000. This will give you the chance to compare multiple drives against each other that we might not mention in any given review. We will generally keep the sample size down to
4-5 drives to not clutter standard reviews.
Quote
Near the end... "but it is probably one of the most extensively tested for compatibility and reliability." Can you back this statement up? How much testing does Intel do on its drives, for example?
This is a WD claim, the hour count, and they indicated it's probably the most extensive. We can't know for sure though as OCZ, Intel and others haven't published this data. I will say though, that they ran HDD routines which are certainly more intensive that most, if not all SSD guys run.
Quote
Finally, that trim performance bug thing - it looks like a big deal to me. Why buy this drive with that issue when I can get another SSD without the issue for a similar price? How often did it happen? How bad was it when it happened? Was it just noticeably slower, or was it as slow as a mechanical HDD? Or worse? I'd definitely keep the updates flowing on that issue - by picking this up and following it through to the bitter end, you can help WD make a better drive, make life better for SSD consumers everywhere, and maybe do something to establish SR's reputation as a review site that finds and cares about the little niggles as well.
Now about the Western Digital SSD itself… The TRIM problem we ran into was triggered primarily with synthetic benchmarks that dump random data onto the drive. In the real world this doesn’t generally happen. Western Digital was able to replicate the problem and should be able to find a way to lessen the problem with a new firmware release. When the problem was active (as you can see in the HDTune Write image) write performance was terrible. Slower than notebook drive slow… like 2-3MB/s sequential write speeds bad.
This was a huge concern in our labs and prompted us to rerun tests yesterday after a long conference call with their product designers.
We localized the problem with their help and were able to work around the problem by limiting the scope of the IOMeter test to a smaller portion of the drive. If random write activity is spread out over the entire surface the drive had a mental breakdown of sorts.
Quote
Still, good first effort! Keep it up!
We will. And we greatly appreciate your input. I'm sure you guys will give us a little time to work out the bugs
We just didn't want to sit on this and other drives, without posting a review, even if it's not the final review. Anything we publish now, will be updated with additional tests as we develop them.
Sign In »
Register Now!
Help
MultiQuote
