Sign in to follow this  
KingGremlin

15k.5 numbers posted

Recommended Posts

The performance database has 15k.5 numbers posted now. Despite incredible low level performance, workstation performance is terrible. Eugene sure wasn't lying there. Is this the second sample, Eugene? Multi-user performance is pretty good, but certainly nothing exceptional when considering how tanked the single user performance is. Hopefully this is just a firmware issue that will be fixed before the drive reaches market.

Share this post


Link to post
Share on other sites
Guest Eugene
The performance database has 15k.5 numbers posted now. Despite incredible low level performance, workstation performance is terrible. Eugene sure wasn't lying there. Is this the second sample, Eugene? Multi-user performance is pretty good, but certainly nothing exceptional when considering how tanked the single user performance is. Hopefully this is just a firmware issue that will be fixed before the drive reaches market.

Yes, this is the second (68-pin) sample... the first (80-pin) unit apparently hit some pretty bad hiccups with the 80-68 pin converter we were using...

Honestly, this shouldn't be discordant to many. STR long ago wrote its way of the equation. Yes, 135 MB/sec is imrpessive, but if STR accounted for, say, 5% of ovcerall performance, what difference would it make if it was 135 MB/sec or 999 TB/sec?

Share this post


Link to post
Share on other sites

Yes, lowlevel performance has been downplayed for a while now, but there should still be a limit to how little performance a drive manufacturer can pull out of the raw numbers a drive can produce. You can't judge the performance of a car simply by looking at HP numbers, but if a car has 400HP and records a 0-60 time of 11 seconds, people are going to be left scratching their head thinking what on earth did the engineers do to mess up the potential of that car so much?

but if STR accounted for, say, 5% of ovcerall performance, what difference would it make if it was 135 MB/sec or 999 TB/sec?

Absolute difference? Not much. However, if it's 45% better than the previous leader, with the 3rd fastest access time measurement to date, I certainly expect relative performance that is better than 40-60% slower than the previous generation competition in workstation performance which supposedly stresses localized access patterns which should benefit from the extreme STR numbers. This drive lost 4 out of 5 single user marks to the 250GB Samsung Spinpoint 120 (14.3ms, 71.3MB/s-41.5MB/s). I don't care how much you want to minimize the impact of lowlevel performance, that's embarrassing. No wonder Seagate bought Maxtor. Give this drive to Maxtor or Fujitsu engineers I bet they would have extracted significantly better performance out of it.

Is this drive going to have a desktop and server mode like the 15k.4?

Share this post


Link to post
Share on other sites

I don't believe STR accounts for only 5 percent of the overall performance of a drive, I think it has much greater influence. In the end however, it's the whole package - STR, access times, command queuing optimizations, caching strategies, even interface transfer rate - which leads to a certain performance level. If there is unbalanced performance in on these low level characteristics, a drive will not reach its expected performance level.

In this case, the problem is probably with the caching strategies. The 15K.4 and 15K.5 seem to be totally de-optimized for workstation usage. If the 15K.5 has to perform a lot more seeks than other drives with proper caching strategies, it will still be behind in performance, despite its stunning sequential transfer rates. That doesn't mean higher STR won't help. Compare the performance of the 15K.5 to the 15K.4 and you will see the 15K.5 performs substantially better in four out of five of SR's RankDisk benchmarks. Since both drives are poor in caching and access times are about equal, the difference is probably caused by the higher STR of the 15K.5.

Share this post


Link to post
Share on other sites
Guest 888

I'll speculate this drive would be the best in very high-load multiple video-streaming server... May-be...

Share this post


Link to post
Share on other sites
The performance database has 15k.5 numbers posted now. Despite incredible low level performance, workstation performance is terrible. Eugene sure wasn't lying there. Is this the second sample, Eugene? Multi-user performance is pretty good, but certainly nothing exceptional when considering how tanked the single user performance is. Hopefully this is just a firmware issue that will be fixed before the drive reaches market.

Do you really think this drive was ever intended to be a workstation drive?

Not a chance.

Brendan

Share this post


Link to post
Share on other sites

Well, even the server performance is pretty average...

Hopefully Seagate will retain Maxtor's Atlas 15K and Atlas 10K firmwire writers to fix this sort of decidedly unremarkable performance...

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this