Lot's of new chips from AMD this week

[Schmide]

Lost circuits power

Does a pretty good job. They're a little slow to catch up with new processors.

 

[Idontcare]

Originally posted by: drizek
why doesn't anand just try and burn up a cpu as much as possible to get a real TDP? You can also just use a killawatt or hook up an ammeter to your motherboard power input.

The question here is "why?"...what would you, or any other reader, do with the data such a test would generate?

I'm all for more characterization data but I like it to be actionable data...and NIST quality TDP data table for a single sample CPU (each CPU will have its own unique TDP based on its own intrinsic temperature dependent power-consumption profile) would not enable me, as an end-user, to take any new action or alter any ongoing action in regards to how I operate my CPU.

But I do not claim to be special, my needs are not all encompassing, perhaps you can make use of the TDP dataset from a single CPU? Help us understand how this information would be actionable to you as an end-user.

 

[ShawnD1]

Originally posted by: Idontcare

Originally posted by: drizek
why doesn't anand just try and burn up a cpu as much as possible to get a real TDP? You can also just use a killawatt or hook up an ammeter to your motherboard power input.

The question here is "why?"...what would you, or any other reader, do with the data such a test would generate?

I'm probably misunderstanding the question, but total power would be nice to know just so I have some idea of how hot my computer room is going to get if this thing is crunching proteins or playing games. Anandtech already does this.

If someone wants to know CPU power itself instead of total system power, it's probably time to take your Seroquel and chill out.

 

[Idontcare]

Originally posted by: ShawnD1

Originally posted by: Idontcare

Originally posted by: drizek
why doesn't anand just try and burn up a cpu as much as possible to get a real TDP? You can also just use a killawatt or hook up an ammeter to your motherboard power input.

The question here is "why?"...what would you, or any other reader, do with the data such a test would generate?

I'm probably misunderstanding the question, but total power would be nice to know just so I have some idea of how hot my computer room is going to get if this thing is crunching proteins or playing games. Anandtech already does this.

If someone wants to know CPU power itself instead of total system power, it's probably time to take your Seroquel and chill out.

Yeah, TDP is a property of just the CPU itself, not the system's power consumption at the wall. One is a useful thing to know (as you point out) the other not so much (as I was trying to point out).

 

[dmens]

AMD TDP is if every functional unit on the cpu was maxed out which never happens ever. There is no intel equivalent.

I'm still amazed by how widespread this notion is. Go pick up a top bin chip from both intel and amd, throw together an assembly loop that maxes out the scheduler slots, and i can guarantee you both chips will exceed their stated TDP at nominal voltages.

the real interesting question is, how far does that power draw during such a torture test exceed the TDP for each different CPU family. That is the real indication of TDP "fudge" that both companies practice.

 

[drizek]

Originally posted by: Schmide
Lost circuits power

Does a pretty good job. They're a little slow to catch up with new processors.

Thanks, thats a really great site. My X3 720 apparently uses half the idle power as my old 5200+.

Tests like this are important because they give us actual numbers instead of putting CPUs into brackets. FOr example, how can a 2.8ghz CPU have the same TDP as a 2.6ghz CPU? Doing real world tests shows us the true cost of the extra 200mhz. It is also important to me becuase I have a first generation 780G board that doesn't like 125W CPUs, so I try to keep it down. Most importantly, I like to have a system as close to silent as possible, and the lower the power consumption the quieter the PC. You need a smaller CPU heatsink, slower case fan, slower PSU fan, etc.

Of course, I also want to know the same data for motherboards, ram and video cards, and I try to get at it as much as possible, but it really isn't out there. I see all sorts of benchmarks of low end high efficiency CPUs put in huge motherboards with crossfire video cards and 12gigs of ram and then displaying power consumption "at the wall". That information is absolutely useless for anyone wanting to build a reasonable media center/internet PC.

 

[DrMrLordX]

Originally posted by: Idontcare

Originally posted by: drizek
why doesn't anand just try and burn up a cpu as much as possible to get a real TDP? You can also just use a killawatt or hook up an ammeter to your motherboard power input.

The question here is "why?"...what would you, or any other reader, do with the data such a test would generate?

Point would be to try and estimate maximum thermal dissipation from the CPU for cooling purposes. All CPU cooling solutions have a practical limit when it comes to the amount of heat they can effectively dissipate before failing to do their job properly (and in the case of TEC cooling, such a failure can be catastrophic).

For a lot of us that's not a big issue since we just get the best cooling we can afford/can tolerate and tweak speeds and voltages until the chip operates within a desired temperature range, but for the more technically-minded overclocker who is willing to go the extra mile, knowing how to estimate thermals from the chip can be very useful. That being said, such individuals represent a tiny fraction of PC users out there.

 

[drizek]

Another advantage is that it allows you to compare results from different sites. For example, you might be able to compare the results of an intel cpu on AT to an amd cpu on xbit. Sure, it wont be exact, but you can at least get an idea. It is also usefulto compare results across reviews from the same site if you want to compare a new fangled cpu to the old one you have in your system. You cant do a direct comparison by, say, running a c2q in an X78 motherboard a few years down the line.

As for the comment about having only one sample, of course that is a problem. It is unfortunate that they dont ship at least a handful of retail cpus for reviewrs, but thats the way it is. It never stops review sites from giving overclocking results with just one sample(and a cherry picked sample at that, in many cases).

 

[Soleron]

Originally posted by: hans007
intel tdp = for cooling designs.

amd acp = for cooling designs.

these are just for basically cooling design classes, not everything in the same amd acp or intel tdp "class" will actually put out that heat (which is the maximum for that class).

AMD TDP is if every functional unit on the cpu was maxed out which never happens ever. There is no intel equivalent.

Yeah. ACP is in line with Intel's TDP. Intel's own specs admit TDP is not the maximum possible power use.