Reviewed today is the MK6022GAX, the highest performance notebook drive in Toshiba's lineup: A 5400 RPM, 2.5" drive with 16 mb of cache. It surely performs, but does it plumb the sub-20 dBA depths or remain in more audible >20 dBA territory? Here's our acoustic report on the Toshiba MK6022GAX.

http://www.silentpcreview.com/article157-page1.html

Interesting stuff, Mike. I wish I could come over with my MHT2080AT and see how it compares to the rest of the drives you've done so far, but we're about as far away from each other as it gets without leaving the North American continent.

I think the importance of cache size needs to be addressed; different operations rely different amounts on cache. It's also important to note that the way of looking at differing cache sizes is critical as well. Unlike in some cases where it would require continuous doubling of memory size to attain double the performance gains, cache works differently. For example, 8MB of cache does not necessarily command a double 2X gain of performance gain from having cache memory as that of 2MB; no, it's 6MB more cache to work with in general. Not only does it allow four times the amount of same-size data samples to be cached, but it enables caching of much larger data samples at the same time, so increasing the cache from 2MB to 8 MB is more effective than people would be led to believe.

Going from 8MB to 16MB is an 8MB gain, which may look like a doublling of cache capacity, and perhaps, to some people, maybe an effective doubling of cache-performance-increase effect over 8Mb, but that, again, is not true. Not only can the drive now cache twice as many samples of the same size as 8MB cache, but now it can cache data samples ranging above 8MB up to 16MB in size, which opens up incredible gains in performance.

What I'm saying is, let's say you're caching 1MB samples, for example. The 2MB cache only can cache 2 samples. The 8MB cache can cache 8 and the 16MB cache can cache 16 samples. That would indicate a 4X increase in cache hits for 8MB over 2MB, and 8X increase in cache hits fot 16MB over 2MB, but that's barely skimming the surface. Now look at 3MB data samples. The 2MB cache cannot even cache a 3MB sample; therefor, for a 3MB data sample, there isn't any benefit from caching at all. Meanwhile, the 8MB cached drive can cache two whole 3MB data samples, and the 16MB cache can handle a full 5 samples at 3MB! You're looking at "infinite"% gain over 2MB cache for 3MB samples and 150% gain over 8MB cache for the 16MB cache.

I'm not saying that I believe your results when the benches indicated the 16MB cached 60GB Toshiba to outperform a 180GXP, since that's highly unlikely, but that this drive can crush anything else in the category with 8MB of cache, especially 2MB of cache? Oh heck yes!

Now, the main question is: how much does your software benefit from caching? Again, if you're a user that's constantly hitting the same data over and over, the cache will help you much more than someone who works with brand new or extremely large datasamples with each drive hit. In other words, for reading operations, browsing the web will accelerate incredibly with larger cache, but a Photoshop artist working with 200Mb, 16-bits/channel CMYK high-resolution images probably will not benefit nearly as much. In those cases, it's all about raw transfer rate from the physical storage medium; nothing helps that out better than RAID. On writes it's a little different, as the cache memory can act as buffer memory.

-Ed

Bravo to the reviewer for realizing that Benchmarks != Real Life. 16 MB is pretty uncommon so I'm guessing that played with the benchmarks somewhat.

Is it really "the highest performance notebook drive in Toshiba's lineup"? Even more than the 7200rpm drive (also 16mb)? Interesting...

I was not aware that Toshiba made 7200rpm 2.5" drives (Hitachi does, in their 7K60 and E7K60 series drives); perhaps Mike wasn't aware, either.

hyperslug, could you please link us? I can't seem to find it at Toshiba's site.

-Ed

How odd. It isn't on Toshiba's site, is it? Anyway, model # is MK5024GAY. There is some info in this well hidden pdf on page 2 which indicates the rotational speed.

Also interesting is this link, mwave, that says discontinued.

I just picked up this drive from newegg, unfortunetly I haven't had time to use it that much yet as I have to re-install my OS. But from my DOS useage it seems pretty fast and is acceptably quiet. I can hardly hear it when the case is closed up, and that only if I really try hard. I believe if I suspend it I won't hear it at all, but we shall see.

I would be wary of making any simplistic assumptions on what improvements a doubling of cache size from 8 MB to 16 would give. Need to know a lot of parameters about the drive caching system _before_ even considering usage patterns (home, workstation, server). Time for the reviewers to try and scour storagereview.com for possible info.

Hello:

Does this drive (or any of this type) come with the IDE adaptor (that I *think* I'm seeing in the articles photos)? In other words could it be used easily in a desktop system plugged into a desktop motherboard?

TIA

NeilBlanchard --

AFAIK, none of the notebook drives come w/an adapter, but these are really cheap and seem readily available. FrontierPC, who lent me the drives, sells 2.5">3.5" adapters for less than CA$12.

If you pay $10US or more for one (before shipping), either it's made of gold or platinum, or you're getting ripped off.

-Ed

MikeC-

You didn't mention anything about AAM status of the drive. Does this drive support AAM? If so, is there a substantial noise difference between performance and seek modes, and what is the default setting?

Edward-

While while you're quite right that additional hard drive cache does not increase performance in a directly proportional fashion, I think the examples that you've given don't accurately represent the way these caches work. For example, you write that there's no benefit from caching if there's only 2MB of cache and a 3MB sample size. Either the point is trivial (no one would design a cache that way), or you're confusing the size of data chunks in the cache (which are about the size of file system clusters, about 32kB) with the size of the files being accessed. Point being, if a 3MB file is being cached in 2MB worth of cache, than 2/3 of the data from that file will be in the cache, not none of it. Even in the example you gave of working with a 200MB photoshop file, a read-ahead cache can be of benefit if your system is bottlenecked somewhere else (say at the CPU, because you're applying a filter).

Did not check on AAM, the seek noise is low.

In that case you are talking about the cache memory acting as a buffer, and as I mentioned, when writing, the cache memory acts as a buffer, so my comments do not apply, since I made no explicit implications about buffer function, just cache function.

I did not mention cache size specifically relating to file size; I said sample size. As for Photoshop, 95% of the time I apply filters, they're on the entire image, so the sample size is almost the entire file, anyway.

If you're doing something like professional sound editing, where you work more often with just a small portion of the file, then yes, the sample size is going to be much smaller than the actual file size and yes, the cache memory will help more.

Cluster/chunk size is not the same as cacheable data sample size is not the same as file size. The file size is just that, the entire file. The cluster/chunk size is related to the way the file is divided up on your hard drive (depends on NTFS vs. FAT, depends on single drive vs. RAID, depends on settings in RAID, and myriad items). The sample size is the portion of a file that is accessed more often than the rest; if your filters are on the whole image all the time, the sample is almost the entire file, if you're doing sound editing and you're focusing on a small time portion, then that small portion is your data sample. For example if I'm performing match color adjustments on an image, it's for the entire image, but if I'm using fader/synth effects on a 15-second portion of a 4-minute song, it's a much smaller portion of the file, and probably, as a data sample, fits in cache.

I honestly do not know just how important a read buffer is; as far as I'm concerned a bottleneck at the CPU, i.e. the drive is waiting on the CPU, isn't nearly as bad as a bottleneck at the drive, i.e. the rest of the system is waiting on the hard drive, so write buffering is far more important than read buffering, in which case we're talking about new data, not in a cache situation where it's the sample data sample getting repeated hits.

Again I'm not making any direct comments on buffering as I do not know what sort of impact that has on different applications, I'm only commenting on caching. I.E. repeat hits to the same data sample, not accelerating movement of new data samples in or out of the drive.

-Ed

Edward-

Thanks for the clarification, and reminding me that you were just talking about cache functions as opposed to buffer functions. For performance considerations, though, one might as well consider them together since they both depend on the amount of on-board memory on the drive, and since the goal of a smart read buffer is basically to increase cache hits. Consequently, I tend to think of them together, especially in the context of 2MB vs. 8MB vs. 16MB.

But even just considering cache functions, unless a drive uses a really dumb least recently used (LRU) cache algorithm, some of the sample will wind up in the cache. Perhaps a trival amount (2MB of a 200MB file), but sometimes enough to matter (2MB of a 3MB sample). The amount of benefit in such situations depends on the caching algorithm, but is not unequivocally "no benefit", and having the extra cache to fit in that last 1MB is not an "infinite"% gain, even metaphorically.

Anyone have a 16mb cache 2.5" drive? Does it run hot?