On my C2D E8400 @ stock 3GHz, Alex Kan's AK version 8.0 SSE4.1 optimized SETI client crunches WUs in 0:50-1:00, much faster than stock client (avg 2:20). CPU temperature is hotter, ~132-138F on AKv8 up from ~122F on stock client. Using Boinc Manager v5.10.45.

---

Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2015 x64

Client error got worst after AK V8 and still havnt figured out the cause.

I ran memtest and no errors found. I lowered down the memory to 910 from 1190 but still errors occur. One thing to note is that Windows XP64 shows my cpu speed as 4104 but it should be 3.6 as that what i set on the bios.

after lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=830968726
http://setiathome.berkeley.edu/result.php?resultid=830296880

before lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=829419130
http://setiathome.berkeley.edu/result.php?resultid=828901082

Anyone has answers on this?

---

Client error got worst after AK V8 and still havnt figured out the cause.

I ran memtest and no errors found. I lowered down the memory to 910 from 1190 but still errors occur. One thing to note is that Windows XP64 shows my cpu speed as 4104 but it should be 3.6 as that what i set on the bios.

after lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=830968726
http://setiathome.berkeley.edu/result.php?resultid=830296880

before lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=829419130
http://setiathome.berkeley.edu/result.php?resultid=828901082

Anyone has answers on this?

Made links active for you...

Put [url] before and [/url] after each link.

---

Clk2HlpSetiCty:::PayIt4ward

Client error got worst after AK V8 and still havnt figured out the cause.

I ran memtest and no errors found. I lowered down the memory to 910 from 1190 but still errors occur. One thing to note is that Windows XP64 shows my cpu speed as 4104 but it should be 3.6 as that what i set on the bios.

after lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=830968726
http://setiathome.berkeley.edu/result.php?resultid=830296880

before lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=829419130
http://setiathome.berkeley.edu/result.php?resultid=828901082

Anyone has answers on this?

I had a few of similar errors reported. You should back the cpu speed down to 3.0 and see if it works ok there. Many of us found that after switching to the AK V8 code that the cpu works harder and gets hotter than before. Due to this many of us had to back down the OC to get stable again.

Client error got worst after AK V8 and still havnt figured out the cause.

I ran memtest and no errors found. I lowered down the memory to 910 from 1190 but still errors occur. One thing to note is that Windows XP64 shows my cpu speed as 4104 but it should be 3.6 as that what i set on the bios.

after lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=830968726
http://setiathome.berkeley.edu/result.php?resultid=830296880

before lowering the mem speed:
http://setiathome.berkeley.edu/result.php?resultid=829419130
http://setiathome.berkeley.edu/result.php?resultid=828901082

Anyone has answers on this?

I had a few of similar errors reported. You should back the cpu speed down to 3.0 and see if it works ok there. Many of us found that after switching to the AK V8 code that the cpu works harder and gets hotter than before. Due to this many of us had to back down the OC to get stable again.

Do you guys stress test your system with Prime95? Works for me... Much more stressful on the CPU then SETI...

Do you guys stress test your system with Prime95? Works for me... Much more stressful on the CPU then SETI...

Actually, that depends on the specific system and ap. At least one reliable participant has reported that a recent ap here needed a slightly lower CPU speed than was able to run a multi-hour Prime95 test on one of his hosts.

As an old microprocessor design engineer and person involved in testing, the single most common technical error I see propagated through the overclocking community in this area is:

The heat is everything premise. Folks think that the most speed challenging test is the one that heats the CPU the most, and also think that the only thing that makes one ap more speed challenging than another is heat. This is false. No test or application exercises all paths and sequences, and for a specific sample of a part, an application can have a more speed challenging requirement than any particular test or benchmark.

---

Do you guys stress test your system with Prime95? Works for me... Much more stressful on the CPU then SETI...

My thoughts exactly. I won't let a SETI WU near my rig until it is 8 hour+ Prime95 stable.

Also X-Files 27 said "Client error got worst ..." which suggests there were regular Client errors before loading the new App. That, in itself, is an indication that the host was not healthy.

F.

---

Do you guys stress test your system with Prime95? Works for me... Much more stressful on the CPU then SETI...

Actually, that depends on the specific system and ap. At least one reliable participant has reported that a recent ap here needed a slightly lower CPU speed than was able to run a multi-hour Prime95 test on one of his hosts.

As an old microprocessor design engineer and person involved in testing, the single most common technical error I see propagated through the overclocking community in this area is:

The heat is everything premise. Folks think that the most speed challenging test is the one that heats the CPU the most, and also think that the only thing that makes one ap more speed challenging than another is heat. This is false. No test or application exercises all paths and sequences, and for a specific sample of a part, an application can have a more speed challenging requirement than any particular test or benchmark.

I can't disagree with that but Prime95 stable is the best indicator we have. My (Q9450) rig will run 8+ hours of Prime95 (either short FFTs or Blend) at 3.6GHz. Then it turns up its toes after about 3 hours of SETI crunching. So I've wound it back down to 3.5GHz - it has done 60+ hours straight at that speed before and I have no indication that it is likely to fail (though I am a little concerned that IE seems to lose contact with my router now and again!)

F.

---

I'll see if lowering down the CPU OC will help mitigate the problem.

BTW, did anyone find if SSE4.1 has better results compared to SSSE3? So far in my case SSSE3 is faster by 1000secs.

---

I'll see if lowering down the CPU OC will help mitigate the problem.

BTW, did anyone find if SSE4.1 has better results compared to SSSE3? So far in my case SSSE3 is faster by 1000secs.

Not too surprising. Some sse4.1 machines will prefer the ssse3x version because it has an architectural feature designed to relieve some of the pressure on FSB/Cache/RAM. That could account for about 5%-> 10% improvement using ssse3x , on very pressured machines such as those with very High OC Quad Core & slower Bus. In testing that improvement works particularly well on the 65nM Core2's that have small cache and low FSB (667MHz or 800MHz) and can show an even larger difference (over a build made without the x-tweak).

On non-pressured SSE4.1 machines, 4 different builds ssse3, ssse3x, sse4.1 & sse4.1x showed no significant difference, but a slight preference (<0.5%) for sse4.1(non-x) in Mid ARs. SSSE3 only machines all preferred the 'x' variant, so the two fastest apps ssse3x & sse4.1 were the ones chosen for release.

ssse3 (non-x) : Deleted (always slower)
ssse3x : promoted to dual purpose role
sse4.1: non-pressured 45nM
sse4.1x: Redundant, identical performance to ssse3x

making the release builds SSE3, SSSE3x and SSE4.1, also helped reduce confusion about which app to chose, which was an important consideration for release, as well as reducing the development workload by cutting the number of apps from 10 down to 6 (3 apps in each of 32 & 64-bit).

Naturally when refinements are made down the track, the choice of best builds may change.

Having said all that, 1000 seconds difference in a WU really sounds like a different AR. There are some points across ARs where there is a sudden change in the amount of work required, perhaps the WUs in question lie on opposite sides of one of these boundaries.

Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

No major improvement on my AMD but the #'s are FLYING by on my Dual-Core!

Nice work guys!

---

I'll see if lowering down the CPU OC will help mitigate the problem.

BTW, did anyone find if SSE4.1 has better results compared to SSSE3? So far in my case SSSE3 is faster by 1000secs.

Here is how SSSE3x vs SSE4.1 performs on my o/c'd Q9450:


Direct Link

Note that there are only 3 results for SSE4.1 at VLAR so little confidence in the comparison in that AR band.

Linking with Jason's comments, I am running 2G of PC6400 RAM in dual-channel at 1:1 so RAM is running at 876MHz. I guess that might qualify my rig as high o/c low FSB; it is 45nm and 12M L2 cache.

Conclusion is that if you get a SSE4.1 capable processor, don't assume that the SSE4.1 App will be the best; it is worth trying the SSSE3x.

F.

---

No errors so far after lowering the CPU OC to 3.5 and up the RAM OC to 1190 with a little increase of juice (voltage).

I'll switch my 45nm to SSSE3.x now.

---

I have an E6600 and a Q6600, running on the same model motherboard (Gigabyte 965P-DS3) at the same settings save for voltage (3.006 GHz, about 1.37V requested for the Q, about 1.4 for the E) on WinXP SP2. Both are running 25% SETI since the AK V8 Windows port release, with the current Einstein Windows power user ap.

Across the two machines I've had at least three spontaneous reboots since the V8 port release. The E6600 had not done that in many months, and the Q6600 was doing it about once a month.

Possibly I am another instance of someone whose hosts won't quite keep up with the ported AK ap at formerly satisfactory settings.

By the way, both systems were already running a voltage higher than that required for hours of Prime95 success, as the Einstein ap I was using required a bit more. However its failure symptom was occasional failed Work Units--not spontaneous reboots.

I'm not interested in going much higher in voltage, so if I get another reboot soon I'll drop the clock rate a little.

---

No errors so far after lowering the CPU OC to 3.5 and up the RAM OC to 1190 with a little increase of juice (voltage).

I'll switch my 45nm to SSSE3.x now.

That's cool........or should I say hot? Sounds like you prefere your OC right at the edge just like Mark.

No errors so far after lowering the CPU OC to 3.5 and up the RAM OC to 1190 with a little increase of juice (voltage).

I'll switch my 45nm to SSSE3.x now.

That's cool........or should I say hot? Sounds like you prefere your OC right at the edge just like Mark.

LOL......the kitties squeeze them little electrons 'till they squeal......

And I have found that the SSSE3x does better on the Frozen Penny than the SSE4.1....I have more cpu clock than RAM bandwith on tap, and the SSSE3x seems to work better with all 4 cores loaded with Seti.......

Just lowered my fsb on the Penny from 458 to 456 x 10......trying to get it a bit more stable....

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

My temps have gone up from 57 Celsius to 60 Celsius with the new app

---

My temps have gone up from 57 Celsius to 60 Celsius with the new app

> hmmm - mine dropped by 6-7 ;)

---

BOINC Wiki . . .

Science Status Page . . .

My temps haven't changed.

NO FAIR, I feel left out.

BOOHOO!

---

Pleased to report that despite ambient temperatures of over 30C today where the computers live that none of the quads suffered errors or crashes despite the higher core temperatures.

---