SSE2 vs SSE3 uptake on X86

MP
MP
Joined: 9 Feb 05
Posts: 29
Credit: 761523
RAC: 0
Topic 195673

Almost universally, most PCs these days have MMX and SSE by default. To not have an SSE version of some kind for a distributed computing application is rare.

There are quite a few mobile PC chips out there that used to have almost no support for SSE3, but this is fading. I think I own one, but don't have Vista running on it so it is not so easy to test for.

SSE means :
http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions

I suspect that if E@H did a user base differential :

SSE2 users
---------- = x
SSE3 users

that it would be less than 2.

For SSE/SSE3 it would be greater than 3, I assume. SSE2 and SS3 instructions (as compliers implement them) do offer some fantastic speed advantages. Yet the uptake for these instructions is still slow outside multimedia users.

Myself -- I have never liked the X86 instructions, memory nor register design. Yes, the CPUs work -- but one is only a few instructions away from disaster. The X86 CPUs -- you can only love if you like your finite state machines cobbled together.

The 680x0 family was always better, but a mere stopgap until better ideas could come along like

[url]http://en.wikipedia.org/wiki/Cell_(microprocessor)[/url]

mikey
mikey
Joined: 22 Jan 05
Posts: 12666
Credit: 1839065474
RAC: 4204

SSE2 vs SSE3 uptake on X86

Quote:


Almost universally, most PCs these days have MMX and SSE by default. To not have an SSE version of some kind for a distributed computing application is rare.

There are quite a few mobile PC chips out there that used to have almost no support for SSE3, but this is fading. I think I own one, but don't have Vista running on it so it is not so easy to test for.

SSE means :
http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions

I suspect that if E@H did a user base differential :

SSE2 users
---------- = x
SSE3 users

that it would be less than 2.

For SSE/SSE3 it would be greater than 3, I assume.

SSE2 and SS3 instructions as compliers implement them do offer some fantastic speed advantages, but the uptake for these instructions is still slow.

I just checked my units and saw this
"Global Correlations S5 HF search #1 v3.06 (S5GCESSE2)". You can see that it is using SSE2 for crunching, your pc's are hidden though so I can't check your units. All of my pc's are elsewhere for this month as my Team is crunching in remberence of a deceased members favorite project so I don't have alot of units.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250251847
RAC: 35705

What I get from our GW

What I get from our GW Application (S5GC1):

- SSE3 doesn't help this application much, so there are no such application versions built and I can't say anything about the distribution

- we get more results from the Mac OX PPC AltiVec app than from all non-SSE2 x86 Apps (SSE and "standard" FPU, Windows and Linux) together. In fact I'm wondering whether it does make sense to continue supporting non-SSE2 CPUs at all.

BM

PS: I agree that of all CPU architectures available at any given time since the days of the 8080, the architecture of the Intel CPUs had always been among the worst. Every other CPU design that I can remember (Motorola 6809 & 68k, SPARC, MIPS, Alpha, PowerPC) was better and did overcome a couple of limitations that x86 CPUs still suffer from. But it's not primarily hardware properties that sells computers, it's software (availability & cost).

BM

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0

RE: What I get from our GW

Quote:

What I get from our GW Application (S5GC1):

- SSE3 doesn't help this application much, so there are no such application versions built and I can't say anything about the distribution

- we get more results from the Mac OX PPC AltiVec app than from all non-SSE2 x86 Apps (SSE and "standard" FPU, Windows and Linux) together. In fact I'm wondering whether it does make sense to continue supporting non-SSE2 CPUs at all.

Most of those non-SSE2 x86s probably belong to me. ;)

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250251847
RAC: 35705

RE: Most of those non-SSE2

Quote:
Most of those non-SSE2 x86s probably belong to me. ;)

Ok, then it does make sense ;-)

BM

Edit: seriously I'll probably drop the SSE-only App for the next run. There will be a SSE2 App and a "compatibility" App.

BM

M. Schmitt
M. Schmitt
Joined: 27 Jun 05
Posts: 478
Credit: 15872262
RAC: 0

No, no, no. :) I still run an

No, no, no. :)
I still run an old Athlon XP 3000 24/7 since many years on the first motherboard without any errors! I doubt that todays technology(mainboards, disks) will last that long. But anyway I would use it for WCG, if non SSE CPU's will not be supported here any more.

Happy WE

ExtraTerrestrial Apes
ExtraTerrestria...
Joined: 10 Nov 04
Posts: 770
Credit: 576676892
RAC: 195014

RE: if non SSE CPU's will

Quote:
if non SSE CPU's will not be supported here any more.

If I understood Bernd correctly they'd run the plain x86 codepath on the traditional FPU instead of the SSE1 unit.

MrS

Scanning for our furry friends since Jan 2002

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.