The future is/seems APU

[Nflight]

Dr Drezha; Along with your thoughts of using Distributed Computing I too have been eying the bug as well. While it is more non-combustible gas observance as I talk about CO2 monitoring. I feel a modeling equation of fire science derivatives is also possible. I too have a back ground in Fire Sciences among my many diverse aptitudes I follow in Science. Yes Engineering is my families trait I decided to avoid, Father is a mechanical eng, my brother has his MS in EE, and my sister followed with IE studies before she switched to Business. My Niece is in her freshman year for IE at the University of Minnesota, she is not happy with her Calculus courses though.

I had 25 years as a volunteer fire fighter before age and brain cells converted me to write grants for all those F.O.O.L. brethren looking to up their game with better equipment. I had the most unique an desired education experiences learning about Forensic study material. To watch and observe the fire act in slow motion is very fascinating to me.

With Today's society and the progressive emission explosion of CO2 from Power Plants we should also incorporate a convolution of our two sciences where by we evaluate the existing gases in our atmosphere. Depending on the proximity to emitters compared to the green pastures should show greater Oxygen opportunity. More O2 equals more fire hazard, more CO2 would diminish the fire danger would it not? Something to think about!!!

Now back to the original Title of this Forum post - Shall we incorporate APU's and CPU's or just one of these ?

Again Congratulations Drezha

[Dirk Broer]

Did you fire fighter guys already know that AMD has new 45 Watt tdp APUs?
You can use passive cooling for them, or use a <20 dB cooler so you almost can't hear it.
Less heat, less CO₂ too, don't you two have to agree?

[Nflight]

Dirk you know how to razz people just the right amount. I agree that lower tdp wattage is better then over powering and heating up the atmosphere sort of gibberish. Thanks for making me laugh!

[NeoGen]

I was shocked the other day when I saw that on AMD's AMA with Tom's Hardware the guy said that the optimal temperature for the Radeon R9's is 95 Celsius, and that all their boards are prepared to work at that temperature for the lifetime of the product.

Now I'm no fire engineer but would consider that a fire hazard! I can almost imagine the occasional loose Sata cable inside a case with one of those GPUs for some reason touching it and the plastic starting to melt and/or catching fire inside the case.

P.S. - Congratulations Dr. Drezha! That is one fine achievement, and you know you've reached the <0.01% of all the people in the WORLD that can reach that high. The sky is the limit!

[drezha]

Dirk - It's a major shame that the software isn't able to benefit from GPU calculations. I think there was a group in Germany looking at it.
Also, whilst the AMD chips are pretty good at TDP, they are unfortunately, significantly slower (I believe they're compiled with Intel compiler which could explain things - I've seen stories around the web that the Intel compilers cripple speeds on non Intel chips).

Some of the fire safety systems for data centres are pretty cool - Maybe crunchers should begin to consider how to protect all the crunchers they may have? ;)

[Dirk Broer]

I was shocked the other day when I saw that on AMD's AMA with Tom's Hardware the guy said that the optimal temperature for the Radeon R9's is 95 Celsius, and that all their boards are prepared to work at that temperature for the lifetime of the product.

It's perhaps for temperatures like these that if ever there was an advert for water-cooling your graphics card, these tempatures are it. Take this article about water cooling a nVidia GTX 690, using a EK-FC690 GTX
waterblock.

"After 30 minutes under full load for each setup, we recorded the peak temperatures, the results of which were pretty surprising. The reference cooler of the GeForce GTX 690 4GB has always struck us as appearing to be more than up to the job, however with a delta T of 63°C it was a huge amount warmer than our other setups.

For an air cooler, the Accelero Twin Turbo 690 put in a pretty impressive showing, managing to keep the GPU below a delta T of 35°C, that's 28°C cooler than the reference cooler - a fantastic result. What's more it remained incredibly quiet, even under load.

However, if ever there was an advert for water-cooling your graphics card, the EK-FC690 GTX waterblock is it. The delta T it recorded was nothing short of astonishing - just 15°C, with the temperature not rising more than 10°C from idle to load. This is a massive 48°C lower than the reference cooler, and again it's a fair bit quieter too. In contrast, the reference cooler span up to quite a din during testing"

For the R9 water cooling seems to be the way to go too.

[Dirk Broer]

Also, whilst the AMD chips are pretty good at TDP, they are unfortunately, significantly slower (I believe they're compiled with Intel compiler which could explain things - I've seen stories around the web that the Intel compilers cripple speeds on non Intel chips).

Actually the software people are using is mostly compiled using Intel compilers -whether games, bench marks or scientific applications- while it is a proven fact that the Intel compilers cripple non-Intel CPUs by excluding them from the more sophisticated instructions and forcing them to take the long road while computing, e.g. allowing no more recent innovations than SSE2. It was a big scam when Bulldozer CPUs did not seem able to perform AVX instructions, while technically they should be able to perform them. It was then that investigators into that matter found out that the Intel compiler used only would allow AVX instructions on Intel CPUs....

"The Intel compiler and several different Intel function libraries have suboptimal performance on AMD and VIA processors. The reason is that the compiler or library can make multiple versions of a piece of code, each optimized for a certain processor and instruction set, for example SSE2, SSE3, etc. The system includes a function that detects which type of CPU it is running on and chooses the optimal code path for that CPU. This is called a CPU dispatcher. However, the Intel CPU dispatcher does not only check which instruction set is supported by the CPU, it also checks the vendor ID string. If the vendor string is "GenuineIntel" then it uses the optimal code path. If the CPU is not from Intel then, in most cases, it will run the slowest possible version of the code, even if the CPU is fully compatible with a better version."
More....

So if you feel cheated by AMD for your poorly running FX CPU, you might actually be cheated by the very code you're running...

[drezha]

Yeah - this is partly why I'm pushing the company to invest in a Linux machine - I've seen speed ups of 30% just by using Linux and then I can also compile it myself without the Intel compiler for potential increases in speed.

[Dirk Broer]

I've seen speed ups of 30% just by using Linux

A quick question: What kind of CPUs saw the greatest speed increase using Linux as compared to the same CPUs performance under Windows?

[Jason1478963]

I would say the 30 percent is about what i'm seeing with the 8350 crunching on WCG using linux vs windows.