New WD Raptor Drive on the horizon...

[MasterBaiter]

I'm gonna probably go over to dual core some time this year (maybe on the new socket change).
Back to on-topic though... I couldn't justify buying these new Raptor drives unless they have improved the arial density on them. The only benefit I can see to these dudes is that they have more capacity so far. I'd be curious to see some benches on them.

 

[Skrying]

Several reasons.

1: Dell doesn't make any.
2: The single core 4000+ was well behind the single core 640 in performance, and extrapolation would say that the 4800+ or Opteron 175 would probably continue that relation with respect to the dual core P4s. 4 execution units do a better job keeping the I/O pipe stuffed full and dealing with the bursty nature of compilation.

I originally pushed to test Opterons, but after testing the 4000+, I'm pretty confident that the Opteron systems would perform worse, and cost more than the equivalent Dell machines with 840/EEs we use for the same level of service we get from Dell.

So your application is different than others? I havent seen a test yet, though I will say upfront I havent done a ton of research, where a dual core Opteron/A64 didnt beat a Intel dual core.

 

[RussSchultz]

So your application is different than others? I havent seen a test yet, though I will say upfront I havent done a ton of research, where a dual core Opteron/A64 didnt beat a Intel dual core.

Have you seen a parallel compilation testbench? I hadn't, so I had to run the tests myself. As mentioned above, on our tools, with our codebase, the 840/EE beat the Athlon 4000+ with 50% shorter compilation times.

Your mileage may vary, but that was the results of our testing.

You could try looking for a linux kernel recompile benchmark results on the net, but I can't seem to find any. Why? I guess there's not enough developers out there wanting to know which machine has the fastest compile times.

 

[BRiT]

Have you seen a parallel compilation testbench?

The first chart here lists compilation testbenches in Linux, dont know much about the Intel platform, but at least there you can see the differences in the AMD realm: http://www.linuxhardware.org/article.pl?sid=05/04/21/1747217&mode=thread

And here is Anandtech's lil blurb on Dual-Core compiling the Linux kernel while doing other tasks. This seems to show the AMd X2 4200+ to be on par with the Intel Pentium D 840: : http://www.anandtech.com/linux/showdoc.aspx?i=2463&p=9

 

[RussSchultz]

I've found that (in our compiler environment), the equivalent of -jn works better when you double (or more) the number of execution units with parallel threads, particularly if the compiler is relatively optimized.

On our old development tools, the fastest athlon system was the best because:
a) The tools didn't offer parallel compilation
b) The compilation was more compute bound than IO bound.

On our new development tools, the fastest intel system was the best because:
a) The tools offered parallel compilation
b) The tools were bursty in nature: IO bound, and then compute bound.
c) The more tasks I could have that didn't suffer from context switching, the compiler could queue up file reads, and put the task to sleep waiting on completion, letting another ready task compile what it had read in.

I think under the windows environment, the key is "tasks that don't suffer from context switching" (which requires flushing the whole cache, paging in completely different sets of stuff, etc). In this case, its the 840/EE with its 4 'processors'. The even though those extra two 'processors' aren't complete processors, they avoid the context switching, which reduces the IO load, which helps alleviate the compilation bottleneck.

 

[Graham]

I run stripe'd 74gb raptors...

At one point I timed the end of post to getting into windows. It was 3 seconds to login, about another 4 seconds to usable desktop (including AV, etc). So yeah, I'm pretty happy with how fast they are.
After using my 7200rpm machine for a while I'm getting seriously annoyed at how much slower everything feels

My thinking was that my pc would only be as good as it's weakest link, so I decided to ballance it and spend approximatly the same amount of money on hdd, gfx, cpu, mb + ram. It seemed to work fairly nicly for a $2200 nz (~1500us) machine. I could have easily spent far more on cpu instead of hdd, for example, but doubt it would have made for a better machine.

 

[ANova]

So your application is different than others? I havent seen a test yet, though I will say upfront I havent done a ton of research, where a dual core Opteron/A64 didnt beat a Intel dual core.

Mind blower isn't it.

/sarcasm

 

[Karma Police]

I've found that (in our compiler environment), the equivalent of -jn works better when you double (or more) the number of execution units with parallel threads, particularly if the compiler is relatively optimized.

On our old development tools, the fastest athlon system was the best because:
a) The tools didn't offer parallel compilation
b) The compilation was more compute bound than IO bound.

On our new development tools, the fastest intel system was the best because:
a) The tools offered parallel compilation
b) The tools were bursty in nature: IO bound, and then compute bound.
c) The more tasks I could have that didn't suffer from context switching, the compiler could queue up file reads, and put the task to sleep waiting on completion, letting another ready task compile what it had read in.

I think under the windows environment, the key is "tasks that don't suffer from context switching" (which requires flushing the whole cache, paging in completely different sets of stuff, etc). In this case, its the 840/EE with its 4 'processors'. The even though those extra two 'processors' aren't complete processors, they avoid the context switching, which reduces the IO load, which helps alleviate the compilation bottleneck.

This is just.........about...........40,000 miles off topic...........roughly. Where's the comments on the new WD raptor? Does everything here have to turn into an Intel/AMD fan*** fight?

 

[RussSchultz]

It stemmed from somebody saying "its a penis waiving purchase, just like the latest processors", where I offered first had experience that such a purchase had its merits in certain situations, depending on your individual needs.

Does everything here have to turn into an Intel/AMD fan*** fight?

*shrug* I'm not a fanboi either way. I use AMD at home because of price/performance, but our testing found the intel platform to the be best suited to our needs at work.

If its any consolation, we use the 74GB raptors in our development systems.

 

[aaronbond]

Didn't read everyone's posts, but the only diff. between "Server" and "Gamer" is

Time Limited Error Recovery (TLER) on the Server Edition

About TLER

http://www.wdc.com/en/library/sata/2579-001098.pdf

 

[MuFu]

What is going on with TCQ these days? Is there still a lack of support in SATA controllers? There was a big hoo-hah about it when the Raptor first came out.

Edit - aha, I missed that the new version will use NCQ (and that means native SATA, w00p).

 

[KimB]

*shrug* I'm not a fanboi either way. I use AMD at home because of price/performance, but our testing found the intel platform to the be best suited to our needs at work.

I'd just like to make a general comment that it really speaks wonders as to how for AMD has come from their rather humble beginnings (in the PC microprocessor industry, anyway) of reverse-engineering Intel designs that you have to spend so much time arguing that you found Intel's processors to be better for a specific task

 

[pcchen]

I'd just like to make a general comment that it really speaks wonders as to how for AMD has come from their rather humble beginnings (in the PC microprocessor industry, anyway) of reverse-engineering Intel designs that you have to spend so much time arguing that you found Intel's processors to be better for a specific task

I think people has been arguing for that since maybe K-6 era.

 

[KimB]

I think people has been arguing for that since maybe K-6 era.

AMD has had strong supporters since then, certainly. But I don't remember ever seeing a discussion where it was 1 vs. all with the one claiming that an Intel processor performed much better, and everybody else having a hard time believing it.

Personally, I think this shows, more than anything, that AMD would gain quite a lot by going with a Hyperthreading-like design in some benchmarks.

 

[Karma Police]

AMD has had strong supporters since then, certainly. But I don't remember ever seeing a discussion where it was 1 vs. all with the one claiming that an Intel processor performed much better, and everybody else having a hard time believing it.

Personally, I think this shows, more than anything, that AMD would gain quite a lot by going with a Hyperthreading-like design in some benchmarks.

Hmmm. Now that I think of it, this Intel/AMD discussion really clears up the whole "New WD Raptor on the horizon" debate!

 

[N00b]

There is a test on Storagereview

Breathlessly waited for by enthusiasts around the world, the Raptor WD1500 improves upon its predecessor by margins of up to 21% when it comes to our single-user productivity and gaming tests.

This, of course, may simply complicate the power user's dilemma. Given $300-$350 in cash, should an enthusiast choose the monstrous 500 GB Hitachi Deskstar 7K500 (a great performer in its own right) or sacrifice more than 2/3rds of the capacity to garner the swiftest speeds ever delivered by a mechanical storage device?

 

[Sxotty]

ARG! Why would they make that stupid case with the plexiglas for $50 extra. I hope they do not take the lack of sales to mean that PC enthusiasts do not like the drives... Anyway I think I may be able to wrangle one of these in a couple of months for myself