Just a thought Frankie. Could it be your CPU. I see you are running a older dual core E5300 CPU. I have heard several people mention needing a fairly robust CPU to feed the GPU.

---

@Hal & Mike: Thx, I will try to find out more about this settings.

@maverrick: Maybe. But I just run vLHC on one core and keep the other one free.



Looks like the older driver didn't change anything for me:
http://setiathome.berkeley.edu/workunit.php?wuid=1562246103

GTX 460, once again 2.5 times faster then my card :-(




PS: Is there any testing tool for astropulse, like SetiPerformance for MB?

---

@Hal & Mike: Thx, I will try to find out more about this settings.

@maverrick: Maybe. But I just run vLHC on one core and keep the other one free.



Looks like the older driver didn't change anything for me:
http://setiathome.berkeley.edu/workunit.php?wuid=1562246103

GTX 460, once again 2.5 times faster then my card :-(




PS: Is there any testing tool for astropulse, like SetiPerformance for MB?

Of course the 460 is faster it has 7 CU`s whilst the 750 has just 4.

Since i wrote the optimization tipps in the OpenCL readme you can just use the values i`ve suggested.

There is offline bench tool at Lunatics called Knabench.

---

With each crime and every kindness we birth our future.

GTX 460, once again 2.5 times faster then my card :-(

And uses 3 times the power.

---

Grant
Darwin NT

@Hal & Mike: Thx, I will try to find out more about this settings.

@maverrick: Maybe. But I just run vLHC on one core and keep the other one free.



Looks like the older driver didn't change anything for me:
http://setiathome.berkeley.edu/workunit.php?wuid=1562246103

GTX 460, once again 2.5 times faster then my card :-(




PS: Is there any testing tool for astropulse, like SetiPerformance for MB?

Of course the 460 is faster it has 7 CU`s whilst the 750 has just 4.

Since i wrote the optimization tipps in the OpenCL readme you can just use the values i`ve suggested.

There is offline bench tool at Lunatics called Knabench.

It's also running stock apps on a very old version of BOINC so it's pretty safe to say that it's only running a single task at a time on that card. ;-)

BTW that card is also rated at 160W, more than double your 750's rated power use.

Cheers.

PS: Is there any testing tool for astropulse, like SetiPerformance for MB?

The same SetiPerformance can be used for AstroPulse
http://setiathome.berkeley.edu/forum_thread.php?id=73524

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Of course the 460 is faster it has 7 CU`s whilst the 750 has just 4.

Is there so much difference between overall performance and CUDA-performance?
This is from the latest edition of a big german hardware magazine (PC Games Hardware; pretty sure you know it, Mike):







According to this the GTX 750Ti should be more then 50% faster then the GTX 460

Since i wrote the optimization tipps in the OpenCL readme you can just use the values i`ve suggested.

I will crunch my last 4 APs with these settings and see if anything improves.

GTX 460, once again 2.5 times faster then my card :-(

And uses 3 times the power.

BTW that card is also rated at 160W, more than double your 750's rated power use.

Cheers.

Yes, but that's not the point. Those new cards should be faster (or at least on par) AND use less power.

PS: Is there any testing tool for astropulse, like SetiPerformance for MB?

The same SetiPerformance can be used for AstroPulse
http://setiathome.berkeley.edu/forum_thread.php?id=73524

Thx!



PS: I just found out that GPU-Z is using around 30% of my CPU. WTF??

---

Those new cards should be faster (or at least on par) AND use less power.

Why "should"?
Do you expect any new car to be faster than an old Ferrari? ;)
(and when you see some GPU looks like it is "2.5 times faster" always consider it is running one task)

PS: I just found out that GPU-Z is using around 30% of my CPU. WTF??

Use SIV instead (to monitor GPU load, temperature, MHz, ...)
http://rh-software.com/
 

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Is there so much difference between overall performance and CUDA-performance?

Yes, it is.


Model GeForce GTX 460 GeForce GTX 750
Manufacturer nVidia nVidia
Year July 2010 February 2014
Code Name GF104 GM107
Fab Process 40 nm 28 nm
Bus PCIe x16 PCIe 3.0 x16
Memory 768 MB 1024 MB
Core Speed 675 MHz 1020 MHz
Shader Speed 1350 MHz (N/A) MHz
Memory Speed 900 MHz (3600 MHz effective) 1250 MHz (5000 MHz effective)
Unified Shaders 336 512
Texture Mapping Units 56 32
Render Output Units 24 16
Bus Type GDDR5 GDDR5
Bus Width 192-bit 128-bit
DirectX Version DirectX 11 DirectX 11.0
OpenGL Version OpenGL 4.1 OpenGL 4.4
Power (Max TDP) 150 watts 55 watts
Shader Model 5.0 5.0
Bandwidth 86400 MB/sec 80000 MB/sec
Texel Rate 37800 Mtexels/sec 32640 Mtexels/sec
Pixel Rate 16200 Mpixels/sec 16320 Mpixels/sec

As you can see the 460 has 336 shaders the 750 has 512.
This explains good cuda performance.
The 460 has 24 ROPs the 750 only has 16.
Do you get the picture ?

---

With each crime and every kindness we birth our future.

Those new cards should be faster (or at least on par) AND use less power.

Why "should"?
Do you expect any new car to be faster than an old Ferrari? ;)
(and when you see some GPU looks like it is "2.5 times faster" always consider it is running one task)

PS: I just found out that GPU-Z is using around 30% of my CPU. WTF??

Use SIV instead (to monitor GPU load, temperature, MHz, ...)
http://rh-software.com/
 

Well, I would indeed expect a new Ferrari to be a bit faster and have a lot more (horse)power then an old Ferrari ;-)

English isn't my main language, so it's not always that easy to express what I wanna say. What I was trying to say is that the new cards are faster while using less power then their older counterparts. At least that's what tests and comparisons in magazines and on the web are suggesting. But, as I said, that's maybe just true for gaming performance, cuda performance may be different.



@Mike: The first 2 APs were running well with your settings, but when I came back from the shower I found my whole system frozen and stuck after running the second 2 APs for eleven minutes. Had to do a restart using the power button because nothing else worked. That didn't happen before, so I wonder if it could be a result of those settings I changed.

---

Those new cards should be faster (or at least on par) AND use less power.

Why "should"?
Do you expect any new car to be faster than an old Ferrari? ;)
(and when you see some GPU looks like it is "2.5 times faster" always consider it is running one task)

PS: I just found out that GPU-Z is using around 30% of my CPU. WTF??

Use SIV instead (to monitor GPU load, temperature, MHz, ...)
http://rh-software.com/
 

Well, I would indeed expect a new Ferrari to be a bit faster and have a lot more (horse)power then an old Ferrari ;-)

English isn't my main language, so it's not always that easy to express what I wanna say. What I was trying to say is that the new cards are faster while using less power then their older counterparts. At least that's what tests and comparisons in magazines and on the web are suggesting. But, as I said, that's maybe just true for gaming performance, cuda performance may be different.



@Mike: The first 2 APs were running well with your settings, but when I came back from the shower I found my whole system frozen and stuck after running the second 2 APs for eleven minutes. Had to do a restart using the power button because nothing else worked. That didn't happen before, so I wonder if it could be a result of those settings I changed.

Yes, it is possible.

Try ffa_block 6144 ffa_block_fetch 1536.
Even tough i have seen 750`s with much higher values.

And the 2 units you have finnished are looking good.

---

With each crime and every kindness we birth our future.

According to this the GTX 750Ti should be more then 50% faster then the GTX 460

We are crunching Seti WUs, not running synthetic benchmarks or playing games.
As has been mentioned several times in several threads- the GTX 750/750Ti is a new architecture, the current latest release for CUDA work has been optimised for older architectures. When a version optimised for the current architecture is released, then it will be able to perform up to it's full potential. As it stands, even with it's less than optimal crunching performance it is the most power efficient video card around for crunching Seti.

---

Grant
Darwin NT

@Mike: Thx, I will try those settings on the next patch of APs.

@Grant: Yeah, I said several times that cuda performance may differ from overall/game performance. And I don't think that a new cuda version would change much. Maybe a few percent, but it won't make my card crunch twice as fast or so.....

---