Upload & Sell: On
thanks for posting the benchmarks. I think I know why some of your numbers are off.
I don't know what software this is, but if you are using 2MB to 10MB read/write file sizes, all you are doing in hitting the 64MB cache on the drives. For sustained read/writes you need large file sizes so the cache does not influence the results, and you actually test the disk speed.
That's not quite right. Caches don't work like that AFAIK. For file writes, maybe. For reads, doubtful. And I had the caches turned off in every test. Although I have to admit it's a little confusing what caches are being turned off and which are even able to be turned off.
In any case I completely and totally do not care HOW I got those speeds. All I care is that I get speeds like that with my 15 to 20MB images files. And I do most of the time - with these drive - while I didn't with any other previous drives. So I'm just happy with the results and don't care how it all works. It's interesting to discuss but it's not meaningful IMO. With 2 drives in RAID0 that's 128MB of combined cache and for 4-drive RAID0 it's 256MB. That's about 10 and 20 images respectively. It's also all of the icons from several folders of 2,000 images each - and so on. So please, yes, let us count the cache!
Bench 2, AJA:
With a 16GB file size you are getting 377 MB/s read, which is right in line with what other people are reporting. Also, the the Seagate 3TB drive is not that great at I/O workloads as the tests show, and are topped by the WD Black and VR drives.
OK, but then why all the hits in the 420+ MB/s range all across the graph? The almost 400MB/s writes and the almost 380MB/s reads in the text above are the averages from quite a wide swath of I/O rates. Not only that but one never gets a linear result from multiple drives in RAID0. At least I've never seen it. There's an implied overhead of between 10 and 20MB/s per drive within the stripe. So I was also considering that when talking about these units operated singly. That may have been a mistake in this case. But if so it's the first time I've seen such a linear increase in performance. Another good thing about these drives then... Hmmm, undecided...
Bench 3, Disk Speed Test:
Again, sustained reads on the order of 360 MB/s, which is consistent with what others are reporting.
Doesn't BM do an encode and decode pass on the data-streams? I read that it did. So those speeds are the result of trying to achieve a less synthetic result.
Unknown benchmark. Again, with a 20MB file size all you are doing is testing the cache, and not the disk.
I understand your enthusiam, because I feel that way about technology too. The thing is that some people make buying decisions based on what they read at forums like FM, and you want to check your facts before you post so they are not misled.
If you read my first post in this thread, you will find that I agreed with your recommendation for the new Seagate 3TB 3 platter drives, and that I encouraged the OP to look at using an...Show more →
1) cool then they would be as stoked as I am.
2) Good on ya bro! We can be Seagate brothers. Now I'm waiting for the SSHD drives to hit the 3TB and 4TB range. 8 or 16GB of intelligent NAND cache... Yeeeahhh-BUDDY!
3) Yeah, sorry to disappoint you but that's actually true. Benchmarks are pretty much meaningless in this regard. It's user experience and how much time one spends doing what needs to be done. I can take a MacPro1,1 from 2006, add 64GB RAM of 800MHz RAM to it, overclock the 8 core 3.0GHz processors to around 3.5 to 3.8 GHz, place the system on an SSD, add in the fastest newest Quadro K5000 which makes a lot more difference in many apps than CPU speeds - for example: http://helpx.adobe.com/en/photoshop/kb/photoshop-cs6-gpu-faq.html#mercury Force the machine to load 64bit kernel for continued updates, and so on. A RAM Drive in such a system gets 3.5 gigaBYTES per second.
And while a newer system whether your hack-box or the 2012 12-core MacPro, might even double some of those figures the user won't feel it. 3.5 GB/s feels the same to a photographer as 7 GB/s. The most he'll likely ever do while editing is 1 secon'd's worth of I/O per image. And so it is with many other tasks too. At the end of an 8 or 10 hour day he will not have saved 10min at the edit bay between a tricked out 2006 MacPro and the newest over-clocked system you can realistically imagine! I know we both love high-spec speedy systems but this is something most of us guys completely over-look. Remember The MacPro I'm talking about besides the zinggy gfx card has a potential of 8x3GHz (24GHz) computing power natively. A well tuned 8-year-old workstation class system with a cherry picked application selection will kick ass on (or hold its own against) a modern over-clocked desktop grade system anyone here is likely to build, and at a significantly reduced price-point too - that's just a fact. I can't help that one. Some of us may not WANT to believe it because new shinny things are usually more fun to play with than old things but that's just our own personal problem.