I might have found a bug. That or my dual core X2's are having issues. Can anyone confirm this and maybe help eliminate certain machines (I.E narrow it down a bit).

Anyway, after installing 5.8.4, I noticed the benchmarks (Whetstone) were very LOW on my X2 5200. I ran some tests on prior boinc versions and found them occuring on versions all the way back to 4.05 boinc-gui.exe.

For my tests on my X2 4800 tonite, I restarted the puter, exited out of everything in the systray except "probe II" so I could see temps, and opened taskmanager. I uninstalled boinc, renamed the boinc folder and started by installing 4.05 and working my way up to 4.19. Just installed right over the old and NEVER attached to a project, so I know project applications weren't running. After 4.19 I uninstalled it, and installed 4.20, then continued to just install over the old all the way to 5.8.4. I ran the benchmarks three times each version, and each test immediately followed the earlier one. There were NO breaks, NO other apps started.

Here's my data:


As you can see it happens with basically ALL versions including 5.4.9, 5.7.5 and such.

Now, my question,

Is this just happening to my X2's, is it all X2's, All dual cores, does it heppen with single cores?
I'm not talking about small normal variations, so please don't point out those, I'm looking for the Biggies, like 33-50% smaller Dhrystone benchmarks

thanks

tony

Also, I'm only looking for data from STOCK Boinc clients.

Also, you can probably just test with the version you have. No need to do what I did. It seems to not be version related, atleast not solely.

@Astro, I have had some strange results when Boinc runs CPU Benchmarks. I just ignore them, I know that the ones on my DP500MHz mac are correct so I know when boinc screws them upon the C2D. I have often found that if I ask it to run them they are more accurate than when Boinc decides to run it's own test.

G4 500MHz DP 335 and 741 at the minute
Intel C2D 1.66GHz 1518 and 3315 or there abouts.

Thanks SATAN (never thought I'd say that, LOL)

I'm a Boinc Alpha Tester, If we can narrow down the problem, I'll pass it on to Rom/David and they can attempt a fix. That is if it's a Benchmark problem and not just something about my X2's. The Boinc Alpha mail list now has two seperate reports of this happening, I'm just fishing for more.

tony

Also, you can probably just test with the version you have. No need to do what I did. It seems to not be version related, at least not solely.

I have an Intel P4DEE 3.2ghz. Meaning it is a dual core with HT on each core, Boinc thinks I have 4 cpus. I run version 5.4.2 and when Boinc ran the benchmarks several days ago the numbers were 1416/1724, when I just ran them manually they were 1413/1667.

---

That's a small (what I'd call normal) variation IMO, Now if it changed from 1416/1724 to 1416/900 and back, that's what I'm looking for. If people could stop exit out of all the stuff running in the background, suspend all projects, open taskmanager and monitor cpu usage(to ensure something else isn't getting the cpu), while running the benchmark 10 times in a row and report them, then we'd see.

David Anderson said "the benchmarks are part of boinc.exe,
and they run for 20 seconds (floating-point)
then 15 seconds wait, then 20 seconds (integer)."

So you should see Boinc.exe get 99% of cpu for 20 seconds, then System Idle process at 99% for 15 seconds, then Boinc.exe at 99% for the next 20 seconds, then it should spit out the benchmark numbers.

He also says, "The benchmark code measures CPU time, not elapsed time.
So in principle it shouldn't matter what else is running,
as long as that benchmarks are getting nonzero CPU time." So this monitoring of cpu usage might not be necessary.

thanks again

tony

That's a small (what I'd call normal) variation IMO, Now if it changed from 1416/1724 to 1416/900 and back, that's what I'm looking for. If people could stop exit out of all the stuff running in the background, suspend all projects, open taskmanager and monitor cpu usage(to ensure something else isn't getting the cpu), while running the benchmark 10 times in a row and report them, then we'd see.

David Anderson said "the benchmarks are part of boinc.exe,
and they run for 20 seconds (floating-point)
then 15 seconds wait, then 20 seconds (integer)."

So you should see Boinc.exe get 99% of cpu for 20 seconds, then System Idle process at 99% for 15 seconds, then Boinc.exe at 99% for the next 20 seconds, then it should spit out the benchmark numbers.

He also says, "The benchmark code measures CPU time, not elapsed time.
So in principle it shouldn't matter what else is running,
as long as that benchmarks are getting nonzero CPU time." So this monitoring of cpu usage might not be necessary.
thanks again tony

Does L2 cache size have anything to do with it? After the 1st time the location would be in the cache.

---

I get the same hiccup at times also Tony. I'm using the 5.5.0 Client (can't remember what it's based on) and at times a benchmark will show the "Measured integer speed" up to 5000 million ops/sec lower than normal. This happens on both of my X2 rigs. When i see that it has happened during an automatic benchmark, I re-run it manually and it usually resets back to normal. Sometimes it takes a couple manual benchmarks to get it to normal values. The "Measured floating point speed" rarely varies when this happens.

Weird....

Also, you can probably just test with the version you have. No need to do what I did. It seems to not be version related, at least not solely.

I have an Intel P4DEE 3.2ghz. Meaning it is a dual core with HT on each core, Boinc thinks I have 4 cpus. I run version 5.4.2 and when Boinc ran the benchmarks several days ago the numbers were 1416/1724, when I just ran them manually they were 1413/1667.

I just upgraded to version 5.4.11 and of course Boinc ran the benchmarks:
1416/1277. I did not shut down, nothing else changed.

---

Here's the report testing format which would help:

1) computer description
2) number of cores
3) h/t yes/now
4) boinc version
5) list of benchmarks

For example, I just ran it on my AMD64 X2 5200, I'd report it like:

AMD64 X2 5200
dual core
No h/t
5.8.4
2624/2877
2626/4705
2627/4703
2627/4748
2623/3368
2625/4710
2628/4705
2627/3242
2626/3212
2626/3391

Intel P4 3.2
dual core
with HT
5.4.11
1416/1277
machine name: workstation

---

AMD64 3700+
Single core
no h/t
5.8.4
2396/4339
2389/4341
2389/4356
2391/4352
2387/4357
2394/4354
2398/4340
2394/4353
2396/4355
2389/4359

NO significant variation on my single core AMD

AND

Mobile AMD64 3700
Single core
no h/t
5.8.4
2243/4071
2254/4074
2243/4066
2248/4089
2248/4082
2248/4096
2250/4073
2250/4072
2240/4089
2252/4077

No significant variation on this AMD single core

AMD XP 1833mhz
single core
no h/t
5.4.9
1720/2878
machine name: Seti 2

AMD 1250mhz
single core
no h/t
5.4.9
1168/1961
machine name: Boinc 1

Intel Celeron 1400mhz
single core
no h/t
5.4.9
1262/2269
machine name: Boinc 2

AMD 64 2200mhz
single core
no h/t
5.4.9
2068/3818
machine name: Boinc 3

AMD 64 1800mhz
single core
no h/t
5.4.11
1713/3198
machine name: Boinc 6

Intel P4 2400mhz
single core
no h/t
5.4.9
1000/3709
machine name: Boinc 7

Intel P4 3.4ghz
dual core
no h/t
5.4.11
1715/2610
machine name: Boinc 8

Intel Celeron 2.4ghz
single core
no h/t
5.4.11
814/1523
machine name: Sues Dell

more in a bit

---

Thanks Mikey, I can't tell from what you're posting, are you seeing a large variation(more than 30% dhrystone) during multiple runs of the benchmark? I'm only seeing a single run, and can't tell from this.

I suspect it's just dual cores, and maybe just AMD, but I don't have enough intels to check.

Thanks Mikey, I can't tell from what you're posting, are you seeing a large variation(more than 30% dhrystone) during multiple runs of the benchmark? I'm only seeing a single run, and can't tell from this.

I suspect it's just dual cores, and maybe just AMD, but I don't have enough intels to check.

No I am only running each benchmark once.

---

AMD XP 2166mhz
single core
no h/t
5.4.9
2006/3375
machine name: Donald

Intel Celeron 2500mhz
single core
no h/t
5.4.9
1304/2619
machine name: Donnie

Intel Celeron 900mhz
single core
no h/t
5.4.11
16372/44466 ???????????
machine name: justtesting

Intel PIII 733mhz
single core
no h/t
5.4.11
554/971
machine name: micron

AMD XP 1800mhz
sincle core
no h/t
5.4.9
1696/2856
machine name: Seti 1

Intel P4 1600mhz
single core
no h/t
5.4.9
803/1585
machine name: Seti 10

AMD XP 2000mhz
single core
no h/t
5.4.9
1893/3182
machine name: Seti 3

Intel PIII 1133mhz
single core
no h/t
5.4.2
1019/1833

AMD XP 2066mhz
single core
no h/t
5.4.9
1873/3154
machine name: Seti 5

Intel P4 1600mhz
single core
no h/t
5.4.9
813/1569
machine name: Seti 6

AMD XP 1700mhz
single core
no h/t
5.4.9
1380/2320
machine name: Seti 8

AMD 64 1800mhz
single core
no h/t
5.4.2
1730/3229
machine name: Seti 9

AMD XP 2000mhz
single core
no h/t
5.4.9
1869/3147
machine name: Soyo

---

AMD64 2800
single core
no h/t
5.8.4
1644/3079
1677/3097
1697/3076
1680/3098
1673/3096
1660/3108
1673/3101
1673/3108
1668/3122
1680/3081

no significant variation on this single core AMD either.

AMD X2 4200
Dual Core
No H/T
Boinc 5.5.0

4653/15139
4653/15139
Exit/Restart Boinc
4649/15219
4655/10584
4653/10577
4655/10574
4655/7871
4655/15253
4655/15230

Weird.....I'll do my X2 3800 next.

Edit....forgot which rig I was doing first...LOL. The one above is the X2 4200. Here's the X2 3800.

AMD X2 3800
Dual Core
No H/T
Boinc 5.5.0

4512/14462
4511/14611
4510/7914
4510/10292
4510/7789

I stopped there and reported that last benchmark so it's on the "Computer Summary Page" for the X2 3800.

Weird.

Bingo, Thanks, It appears the X2's don't play well with Boincs benchmark.

Any core 2 peeps wanna test it??

Bingo, Thanks, It appears the X2's don't play well with Boincs benchmark.

Any core 2 peeps wanna test it??

I've got a sick E6300 in for testing at the moment (intermittent hard disk problems). Running HDD test at the moment (about an hour to go) - if I can get a stable boot after that, I'll repeat your tests (bare BOINC install, no projects) and report - probably not till after maintenance.