64 Bit - What should we expect

Message boards : Number crunching : 64 Bit - What should we expect
Message board moderation

To post messages, you must log in.

AuthorMessage
Youngblood
Avatar

Send message
Joined: 3 Apr 99
Posts: 5
Credit: 59,811
RAC: 0
United Kingdom
Message 76481 - Posted: 4 Feb 2005, 18:10:58 UTC

What should we expect Seti@Home wise when (if) 64 Bit becomes fully supported and a 64 Bit client is available?

Answers on a postcard

PS any reply's please rate my new avatar.
"Death Warmed Up" - Just one of the interesting dishes in my new cookbook.
ID: 76481 · Report as offensive
Profile mikey
Volunteer tester
Avatar

Send message
Joined: 17 Dec 99
Posts: 4215
Credit: 3,474,603
RAC: 0
United States
Message 76509 - Posted: 4 Feb 2005, 20:03:36 UTC - in response to Message 76481.  
Last modified: 4 Feb 2005, 20:06:28 UTC

> What should we expect Seti@Home wise when (if) 64 Bit becomes fully supported
> and a 64 Bit client is available?
>
> Answers on a postcard
>
> PS any reply's please rate my new avatar.
>
Right now the AMD 64 chips are optimised for the way how Seti is crunched. The part of the cpu that is 64 bit is not the FPU, which is the part that Seti uses.
Do a search for John McLeod VII, he answered this awhile back.
Avatar looks nice.
ID: 76509 · Report as offensive
Youngblood
Avatar

Send message
Joined: 3 Apr 99
Posts: 5
Credit: 59,811
RAC: 0
United Kingdom
Message 76584 - Posted: 4 Feb 2005, 22:25:22 UTC

John's response is below, thanks for letting me know about this Mikey.

"S@H is mostly floating point which uses the FPU. It is the IPU that is 64 bit, the FPU has not changed much. Eventually, I expect 64 bit versions, but since they will not buy that much, they are not a priority."

At least it means that peoples seti farms won't suffer from a boost in technology, but in what way are the 64 chips optimised for crunching?




"Death Warmed Up" - Just one of the interesting dishes in my new cookbook.
ID: 76584 · Report as offensive
Profile Paul D. Buck
Volunteer tester

Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 76603 - Posted: 4 Feb 2005, 23:22:28 UTC - in response to Message 76584.  

> At least it means that peoples seti farms won't suffer from a boost in
> technology, but in what way are the 64 chips optimised for crunching?

If you talk about a chip like the G5 in the Apple Macintosh, you can do standard double precision math with one fetch as that is the size of the stored number for double precision floating point numbers. The G5 also has two FPUs (not sure of all the versions this is true for), also there are SIMD instructions that could be used to improve throughput.

SETI@Home, because it does not need high accuracy only uses single precision numbers (to the best of my knowledge).

But, *MY* experience is that the optimized versions of the SETI@Home Science Application are not significantly faster than the reqular flavor. As I am going to be lowering my contribution to SETI@Home as the other projects mature and start being "full-time" I have not spent my time looking at them ...

Ok, the "meat"; a 64-bit CPU's biggest advantage is the large virtual memory space it brings to the table. This is of most use when you are talking databases.

If you are interested, go on the Apple site, goto the store, click on the G5 silver box and drill down on the "Why buy" and they have a bunch of nice discussions on what the G5 64-bit chip brings to the table. What is true for it is also true for the AMD 64-bit chips.

The only other reason that some of the 64-bit chips seem to be faster than earlier generations is that the chip designers learn by doing. In computer design, as in most complicated engineering, you have to build it, fly it, figure out what it only soars with the turkeys, and then try, try, try again.

AMD has been learning things, if you compare them and Intel you can see differences in the trade-offs the teams select.
ID: 76603 · Report as offensive
Youngblood
Avatar

Send message
Joined: 3 Apr 99
Posts: 5
Credit: 59,811
RAC: 0
United Kingdom
Message 76626 - Posted: 5 Feb 2005, 0:26:29 UTC

Paul, thanks for the post and the G5 suggestion. I've had a look through and now it seems clear that 64 bit is more about size (like you said huge amounts of virtual memory, and apparently upto 4 terabytes of ram are possible in the future) than a hefty processing speed increase.

Thanks
"Death Warmed Up" - Just one of the interesting dishes in my new cookbook.
ID: 76626 · Report as offensive
Profile Paul D. Buck
Volunteer tester

Send message
Joined: 19 Jul 00
Posts: 3898
Credit: 1,158,042
RAC: 0
United States
Message 76782 - Posted: 5 Feb 2005, 10:59:02 UTC - in response to Message 76626.  

Youngblood,

> Paul, thanks for the post and the G5 suggestion. I've had a look through and
> now it seems clear that 64 bit is more about size (like you said huge amounts
> of virtual memory, and apparently upto 4 terabytes of ram are possible in the
> future) than a hefty processing speed increase.

Well, for a database this will translate into a speed increase. Because the addresses used are "native" there is no need to resort to "tricks" to be able to directly address that memory space.

Keep in mind that the address space is, for a long time to come, mostly of interest in relationship to virtual memory, which is stored on disk with a significantly smaller amount of physical memory.

> Thanks

You are very welcome.
ID: 76782 · Report as offensive

Message boards : Number crunching : 64 Bit - What should we expect


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.