I was wondering if it matter what % I set my processor to use vs what I get for work unit types. I am currently using 4 systems, 3 are set to 50% (all dual core) and 1 is set to 25% (8 cores).

1 is a laptop with integrated graphics (50%) - I currently have plenty of 6.03 enhanced for the cpu and 2 ap units - all in all what i would expect

1 desktop with an nvida card (50%), seems to get mainly all cuda units, with the occasional 6.03 cpu unit tossed in

1 desktop with an ati card (50%), seems to get just like the laptop, cpu unit and a couple ap cpu units.

1 desktop with an ati card (25%), this is the most advanced system and all I have is 6.03 cpu units.

I know the whole ATI/AMD thing is newer, are wu's rare or do I need to set my % higher to be able to get these units for the GPU's.

I have it set in my prefs to receive them.

---

In general BOINC will selectively get work for the most capable processor first.

This explains the fact that your "Nvidia PC" sees mostly CUDA work units.

ATI GPU need a lot of support from the CPU (unlike Nvidia one), so it is possible that by restricting your CPU usage you are all but preventing the GPU from working, hence only seeing CPU work and not GPU work.

Remove the CPU restrictions and you should see an increase in the amount of work the ATI GPUs see, and a general increase in the number of work units you get through

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

I was wondering if it matter what % I set my processor to use vs what I get for work unit types. I am currently using 4 systems, 3 are set to 50% (all dual core) and 1 is set to 25% (8 cores).

The % rule normally doesn't effect GPU tasks.
Using 50% is OK, 25% on a dual core probably not.

Try TThrottle (see below)
If temperature is an issue, just set the CPU and GPU temps.
Otherwise you can set a Max Cpu %, like 50%.

TThrottle uses a more smooth method.

But only works on Windows machines

---

TThrottle Control your temperatures. BoincTasks The best way to view BOINC. Anza Borrego Desert hiking.

In general BOINC will selectively get work for the most capable processor first.

This explains the fact that your "Nvidia PC" sees mostly CUDA work units.

ATI GPU need a lot of support from the CPU (unlike Nvidia one), so it is possible that by restricting your CPU usage you are all but preventing the GPU from working, hence only seeing CPU work and not GPU work.

Remove the CPU restrictions and you should see an increase in the amount of work the ATI GPUs see, and a general increase in the number of work units you get through

I think you nailed it. The 'nvidia PC' is an older XPS that I upgraded a number of things in, including the processor (Pentium D @3.6), but it is running a EVGA 9600GT card. I know that processor is no match for that card so it is just feeding the hungry card.

I just looked and it says the ATI/AMD system at my house received 2 AP unit for the card!

Are AP WU's the only ones that can be crunched on the ATI/AMD cards these days?

---

An Nvidia card presents a very low demand on the CPU. Basically so long as the motherboard will take the card and the operating system the drivers then the the card will work - I think the CPU just acts as an interface between the RAM/HD and the GPU (someone will be along soon to put me right on this...)

ATI cards on the other hand do need a lot of "nursing" from the CPU - hence all the talk of folks "reserving" one CPU core for feeding their ATI GPUs.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

ATI cards on the other hand do need a lot of "nursing" from the CPU - hence all the talk of folks "reserving" one CPU core for feeding their ATI GPUs.

A higher end NVIDIA card will use up to 15% of one virtual core, continuesly.
So they too need to reserve a CPU core.
This is needed because the "feeding" can take up the entire CPU core for even a couple of seconds
And the work done by the GPU outweighs the CPU work by far. So you want to keep the GPU going.

---

TThrottle Control your temperatures. BoincTasks The best way to view BOINC. Anza Borrego Desert hiking.

Thanks Fred.

---

With each crime and every kindness we birth our future.

Multibeam ATI apps only use a CPU a lot at the startup of a new work unit, under a minute, and then essentially nothing (< 3%) until it's done. Don't have any experience with Astropulse apps.

Didn't think I would see Green Vs Red prejudice here. "need a lot of "nursing" from the CPU", yea right.

---

"Life is just nature's way of keeping meat fresh." - The Doctor

the FX CPU's are a bit fussy about giving the GPU time while MB WU's are running. Mike has discovered that allowing the GPU to have 1 or 2 CPU cores massively increases the ATI/AMD Radeon GPU productivity

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

Wow, this is going well now!

The ATI desktop (i7-2600 @3.4 - 8 cores) that is set to 25% processor usage + any GPU tasks is currently crunching 2 wu's on the cpu and a 3rd AP unit on the GPU. Seti keeps feeding it the AP GPU units! I would be scared to see what would happen if I let it all go at 100%. Not sure the system could take it, or my wallet with the cost of electricity here!

---

Well back of the envelope calculations says your rig is doing MB 6.03 roughly at 45-55 credits per hour per core (ignoring the overflows). So once your pending settles down you should be getting around 2400 credits per day just from MB tasks with two cores. Get all eight cores going and it would place it over 9000 (insert DragonBall Z joke here). Of course I don't know how much CPU time is loss for an ATI/AMD AP app and what the credits/hour it generates on average.

---

"Life is just nature's way of keeping meat fresh." - The Doctor

IMHO both ATI and NVidia GPUs need CPU time, both high-end GPUs need about
the same CPU time.
But it's hard to compaire both GPUs and their Compute Language, cause they're
quite different.

As is the architecture of both GPUs.
Remember the problems with VLAR WUs on NVidia GPUs were ATI GPUs don't care.

---

IMHO both ATI and NVidia GPUs need CPU time, both high-end GPUs need about
the same CPU time.
But it's hard to compaire both GPUs and their Compute Language, cause they're
quite different.

As is the architecture of both GPUs.
Remember the problems with VLAR WUs on NVidia GPUs were ATI GPUs don't care.

Not all Nvidia GPUs have a problem with VLARs.

As far as i can tell since the latest fix vlar do not go to any type gpu
which is a great pity coz they put the least strain on the cpu in my rig
maybe someday we wil get a switch for vlar along with MB and AP
for any nutcase that wants them :¬)

As far as i can tell since the latest fix vlar do not go to any type gpu
which is a great pity coz they put the least strain on the cpu in my rig
maybe someday we wil get a switch for vlar along with MB and AP
for any nutcase that wants them :¬)

Well said. Some of us are not afraid
of the vlar boogeymen. ]:)