At this stage it looks as though initial releases will likely be (6 of):

32 bit Windows:
sse3 (AMD & Intel that support sse3)
ssse3x
sse4.1
64 bit Windows:
sse3 (AMD & Intel that support sse3)
ssse3x
sse4.1

Sorry, no graphics (Yet), and no less than SSE3 (Yet). We'll be exploring those items, along with Linux and potentially other ports, in the not too distant future. We are however hoping to squeeze a little more performance out of these builds after release, as Alex is likely off working on version 9....

Jason

That's nice.. I hope my old AMD K7 friend will have soon his MMX version..
Or is MMX to old to support?

---

Oh and for those Who'd like to keep track of what's going on, "V" is now "Who?" again and He's added another project(Primegrid).

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

Final results from the test-pilot-in-chief prior to the BIG DAY.

Shame he can't post for himself... (grumble grumble)

And that code is equally suited to AMD processors? That is, no more of the "Naughty Intel" 'dirty' trick?

Great work!

Happy crunchin',
Martin

It is unknown whether the SSSE3 or SSE4.1 builds would run on some future AMD CPU with those capabilities, but I doubt it. Don't take tests of those builds on Intel systems as directly indicative of performance of the SSE3 builds which do run on AMD systems. The SSE3 builds provide a very considerable speed increase on both Athlon 64 X2 and Phenom, though.

                                                               Joe

Got my new processor installed, now I just need a new worker bee.

---

And herewith the official results of QSilver's test of Alex's Core2 vs Xeon flavours of optimisation on the Mac:

Direct Link

At first sight, it seems obvious that the Xeon flavour has an advantage (between 0.5% and 5%); and that may well be true but the advantage may be less than first sight suggests. The second chart offers further information:

Direct Link

Again, the columns represent the number of WU's in each AR band which is read from the left hand scale. The "Error Bars" represent the range of CPU times within the AR band (read from the right hand scale) with the Mean value indicated. The blue columns are for the Core2 optimisation, the yellow for the Xeon.

In all cases where there are data for both flavours within an AR band, the spread of CPU times for the Xeon flavour is smaller than for the Core2 (perhaps a reflection of better cache management?). And in all cases where there are more than 5 results within an AR band, the whole spread of the CPU times for the Xeon flavour is between the min and max values for the Core2.

So, for example, it is clear from the position of the Mean value for the Xeon flavour in the 0.225 - 0.4 AR band that most of the results were close to the lower boundary. However, without extending the spread it is quite conceivable that future results could be nearer the upper boundary and pull the Mean value closer to the Core2 Mean value.

Judgement call? Yes, the Xeon flavour is more efficient than the Core2 flavour - but, in this case, not by very much.

F.

{snipped the charts from above quote}

Overall, I agree with Fred's assessment. At best, it's 5% or so improvement where most WUs are generated (Angle Ranges between 0.36 and 0.45) with a smaller standard deviation between WUs. It's possible that there will be no improvement depending on the mix of WUs.

If it works, I've attached below a couple of tables showing my MacBookPro (T8300 at 2.4GHz) running the Alex Kan Core2 and Xeon workers. My split of the ARs is different than Fred's.



I'm going to collect data for another week or so to see whether these results hold up. I'll probably just leave the Xeon app running after that because, hey, a few percentage points is worth something. And, it'll have to tide me over until v9!

QS

---

At this stage it looks as though initial releases will likely be (6 of):

32 bit Windows:
sse3 (AMD & Intel that support sse3)
ssse3x
sse4.1
64 bit Windows:
sse3 (AMD & Intel that support sse3)
ssse3x
sse4.1

Sorry, no graphics (Yet), and no less than SSE3 (Yet). We'll be exploring those items, along with Linux and potentially other ports, in the not too distant future. We are however hoping to squeeze a little more performance out of these builds after release, as Alex is likely off working on version 9....

Jason

That's nice.. I hope my old AMD K7 friend will have soon his MMX version..
Or is MMX to old to support?

I use S@H V5.15 Rev. 2.4V SSE3 opt. app for my QX6700, because it's faster as SSSE3.
I see this mostly at the ~ 64 Credit WUs..
The others are nearly the same..


How it will be with the new opt. app with Alexs code?
The SSSE3(x) would be faster than SSE3?

---

I use S@H V5.15 Rev. 2.4V SSE3 opt. app for my QX6700, because it's faster as SSSE3.
I see this mostly at the ~ 64 Credit WUs..
The others are nearly the same..


How it will be with the new opt. app with Alexs code?
The SSSE3(x) would be faster than SSE3?

This is how the pre-release versions of SSE3 and SSSE3x performed:


Direct Link

There are few readings at VHAR (only 9 / 10) so the apparent reversal of effectiveness should be viewed with caution.

F.

---

You can download it now. It's out.

http://lunatics.kwsn.net/discussion-forum/releases-all-released-versions-are-announced-here.msg7491.html;topicseen#msg7491

---

You can download it now. It's out.

http://lunatics.kwsn.net/discussion-forum/releases-all-released-versions-are-announced-here.msg7491.html;topicseen#msg7491

Got it up and Running ...THANKS GUYS!!!!!!!

---

Official Abuser of Boinc Buttons...
And no good credit hound!

You can download it now. It's out.

Yay. Now to try and remotely install it....

---

The result of the first WU has been impressive, but was started with the version of Crunch3. When complete a whole with the V8, I think it will be amazing ...

Best regards and congratulations to all the people involved in this challenge!!!

---

Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)

Up and running with AK_v8_win_x64_SSE3.exe on my AMD Opteron 280's... So far, everything looks very, very good. Thanks to all the coders and testers for the hard work!!!

---

. . . Kindest Regards & Thanks to All for their Contributions to the Project

. . . Windows Port of Alex Kan's v8.0 - Science App Release

---

BOINC Wiki . . .

Science Status Page . . .

OPS, sorry for my wingman, downloaded wrong AK_app... Lost around 20 units! Great work! Thanks to AK and whole group of porters. I'm waiting for Cruch3rs port for FreeBSD ;)

---

i'm watching temps as was warned.
actually on my quads...seems temps went down.
haven't backed off overclocks either.
i'll keep track of them for sure

---

i'm watching temps as was warned.
actually on my quads...seems temps went down.
haven't backed off overclocks either.
i'll keep track of them for sure

Hey Phud...2 more upgrades and I will catch back up. I have been watching temps and they are staying where they were.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

Holy freek'n cow!!! My P4 Prescott 670 is Love'n this thing, I just installed it & WOW!!! Thank YOU!
You guys should get an AWARD!

Up and running with AK_v8_win_x32_SSE3.exe. I'm seeing completion times that are half of what I was getting w/Cruch3rs port(which in itself was a VAST improvement of the stock app)!

What's will happen when it hits the 13feb08ac's that I just got?
OK, back to lurking...

rdr

---

I even noticed a side effect here, When I restarted Boinc after installing the AK Seti app, Now when I surf the web My web browser doesn't wait for some other process like It did under 2.4V. :D

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

Holy freek'n cow!!! My P4 Prescott 670 is Love'n this thing, I just installed it & WOW!!! Thank YOU!
You guys should get an AWARD!

Up and running with AK_v8_win_x32_SSE3.exe. I'm seeing completion times that are half of what I was getting w/Cruch3rs port(which in itself was a VAST improvement of the stock app)!

What's will happen when it hits the 13feb08ac's that I just got?
OK, back to lurking...

rdr

I suspect It'll crunch them.

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

And some results.Running Win32XP SP2,CPU8400@4050, 2G RAM. http://setiathome.berkeley.edu/result.php?resultid=829936918 R-2.4V|xP|FFT:IPP_SSE3|Ben-Joe - CPU time 2841.109375
http://setiathome.berkeley.edu/result.php?resultid=830032515
SSE4.1 Win32 Build 41 - CPU time 1406.406
GREAT Improvement.Thanks to all of you, about great work.Happy crunchin.

---