Performance per watt

[2006]

I see the administrator has now added a performance per dollar section. What would also be good is a performance per watt section too.

[passmark]

Yes we have just recently included pricing (http://www.passmark.com/forum/showthread.php?t=2642) in the CPU benchmark chart (http://www.cpubenchmark.net/)s.

The power used by each CPU is harder to collect and keep up to date in an automated fashion however. While it would be nice to have I can't see us adding it in the short term.

[TruePurple]

I would be very much for this, registered to the forum to post just such a suggestion before I even saw this thread, it would help environmentalists, and would help people trying to stay within their PSU limits, not to mention providing some help to those who just want to save a few bucks in electricity.

[passmark]

Yes, it would be good for know PSU limits, and for heat considerrations.

But maybe not so helpful for environmentalists. The figures that are available tend to be the maximum watts the CPU might use when under full load.

CPUs are very rarely under full load. Especially new CPUs with 6 or 12 threads. So the more interesting number is what the CPU uses when idle or low load (the situation most CPUs are in 90% of the time). New CPUs throttle their speed and shut down sections of the chip as required. So a CPU with a very high maximum current draw, might look much more reasonable when idle. Power usage while idle for a wide range of CPUs is hard data to collect.

[2006]

http://en.wikipedia.org/wiki/List_of_CPU_power_dissipation

I used to have a link that showed the idle times for the newer processors but I cant find it at the minute.

I disagree that todays computers rarely work at their maximum. Watching high definition movies on the PC is very popular and there is no such thing as a fast CPU when you are encoding videos, which is also popular among enthusiasts and somewhat less among the general public. Bad, bloated java applications will also account for a lot of CPU usage when morons go to flash-based websites to watch 1080p movies and so on.

This is why it would be good to have an energy efficiency section, for video cards too, as they are famous for being even more power hungry than the CPU. No pressure though, I understand you are doing this for free.

[passmark]

Watching high definition movies on the PC is very popular... This doesn't stress a new CPU. Playback of a high def 1080p movie only gets CPU usage of 12% to 25% depending on the scene. And this is on an older quad core. Hardly the 100% usage you are claiming for a new CPU.

So a Core i7-880 and Core i7-940 probalby use similar power most of the time. Despite the Thermal Design Power of 130W vs 95W.

Not to say we won't do it one day. Just saying the TDP isn't always the best indicator or actual power use.

Further you need to look at what is in the CPU. If the memory controller and video card are in the CPU (like some of the i5 models) then you are saving power on the motherboard by removing componets from the motherboard. And this won't be obvious by just looking at TDP of just the CPU.

Most of the idle power usage figures I have seen on the net are for a full system (i.e. measured at the mains power input, including MB, disk & video card) and not just for the CPU.

[2006]

Yes, but encoding videos is still very slow and not possible to do in real time for 1080p yet. Gaming is also really popular and I think plenty will be designed to take full advantage of the new i7's, not to mention that some very old games always take 100% CPU regardless how much it really needs, so yeah.

[passmark]

The old games only use 100% of 1 core of a CPU. So on a 12 thread CPU this is 8% usage overall. Far from full usage.

New games might use 2 to 4 cores, but not fully in general. But I agree this would be the most common scenario for maxing out a CPU.

Video encoding might hit 60% to 80% usage depending on how fast your hard drive is. But video encoding might make up <1% of overall PC usage.

So I'll stand by my claim that the average PC's CPU is near idle most of the time.

[TruePurple]

Well, for GPU anyway, which are usually bigger draws of electricity and sources of heat then CPU's, having a average electricity usage would be helpful. Idle draw too would be good.

I am not sure how you get your benchmarks, I think at least some of your GPU benchmarks are off, I will post a thread about that latter when I have the time and energy to compose it. But surely there must be a way to get at least general power draws for GPU's.

Many manufacturers provide that information,(granted, not always completely accurately) there are any number of benchmark type sites that do too. If you count on people submitting information with benchmarking performance and costs, why not do the same with electrical consumption? It won't be totally accurate, but neither is your cost/performance charts either. Its just something helpful to go on.

[passmark]

Our software & scripts collect all the current information. User's don't type in their CPU type and benchmark score, nor do they type in the price. I would hate to think how much inaccurate information we would get we we started asking user's how many watts their GPU & CPU draws. Most people don't even know what's a watt. And even the people who do know their milliwatts from their Joules won't have the equipment to take a measurement.

I don't know of any site that has comprehensive list of real world CPU / GPU power usage benchmarks. (As opposed to the TDP figures).

[2006]

I think one way to calculate idle time is to divide the TDP by half. I have a power measurement tool which confirms my PC at 95 watts under stress and 60-65 when idle, the extra 20 watts is probably due to the other components.

The old games only use 100% of 1 core of a CPU. So on a 12 thread CPU this is 8% usage overall. Far from full usage.

Not so, a 1996 game I have uses 100% on my Pentium D. I can always set the affinity to one core on task manager but what casual non-geek knows how to do that?

[TruePurple]

Before I can approach this any further, I need to know how you automatically collect performance numbers and prices. (also useful for evaluating the data in the most accurate light)

[passmark]

Benchmark numbers are collected from our PerformanceTest (http://www.passmark.com/products/pt.htm) software. Prices are collected from Amazon and Newegg at the moment.

[TruePurple]

Benchmark numbers are collected from our PerformanceTest (http://www.passmark.com/forum/../products/pt.htm) software.

That is not much of a explanation. So people who run this software, have their results automatically or by choice uploaded to your servers? A average of these numbers? But if its different random people and different random computers, how do you screen out the influence of the system X video card etc is plugged in to?

[passmark]

That is not much of a explanation...

I would suggest you actually try using the software, then it should be more obvious.

There are also notes about the graphs here (http://www.cpubenchmark.net/graph_notes.html).

how do you screen out the influence of the system X video card etc

The video card used does not effect the CPU performance.

[TruePurple]

Well, I would rather you explain it, my PC is very old (desperately need to upgrade, which is why I am researching parts for a new one, and why I am posting here.) What is the system requirements of your benchmark? Anyway, is it really that hard to explain with words compared to me having to download, install, use, then uninstall a program?

[passmark]

System requirements for PerformanceTest can be found on the PerformanceTest product page (http://www.passmark.com/products/pt.htm).

Well, I would rather you explain itNo point pasting in pages of text when I can just link to it, and you can read it.....

[TruePurple]

Is there a way to read these pages of text without installing the software?

[TruePurple]

Well the lack of answer on my last question wasn't appreciated, but now that is moot, moving on...

I downloaded and installed, no apparent answers. I ran tastes, two or three of which didn't work at all, one of which didn't close down right (do I have to worry about junk files from the test taking up HDD space from the improper close down?) So now I have test results I don't care about, no answers to my questions. I poked around menu options, no answers.

So in short, no answers, what am I suppose to be looking for anyway? (knowing what one is looking for usually helps in the finding)

BTW, do you have benchmarks for IGP? I tried to find a couple IGP and I could not. If you do not have them, why not?

[passmark]

Not sure what "pages of text" you are refering to.

Comments like, "it didn't work" are pointless. It would be better if you describe the problem in detail (hardware used, software used, Direct X10 support on your video card, steps to reproduce the problem, log files. etc..). Anything else just makes it look like you are trolling.

what am I suppose to be looking for anyway

We can't tell you what you are looking for. We really don't know. We suggested you try the software because you were asking how the software works.

do you have benchmarks for IGP?

I assume you are referring to integrated graphics. You can find a list video cards for which baselines files are available here,
http://www.videocardbenchmark.net/gpu_list.php
These cover the full range of discrete or integrated cards.

[TruePurple]

Comments like, "it didn't work" are pointless. It would be better if you describe the problem in detail (hardware used, software used, Direct X10 support on your video card, steps to reproduce the problem, log files. etc..).
As I have said, I don't care about getting any benchmarks, I only installed and ran it because you said it would answer my question, and I was reporting the results of doing so.

Not sure what "pages of text" you are refering to.

No point pasting in pages of text when I can just link to it, and you can read it...

We can't tell you what you are looking for.

No, what you want me to see. Surely you know what your talking about? If you don't, you can't ask me to know.

We suggested you try the software because you were asking how the software works.
You ("We"? Am I talking to more then one person using the same login?) suggested I use the software for my answer when I asked you how you gather and arrive at your benchmark numbers automatically without anyone needing to do anything. I didn't ask how the benchmark software worked.

I would suggest you actually try using the software, then it should be more obvious.

What do you want me to see when I "try using the software"? What should I be looking for when using the software?

Heh, must be a second person, since you talk like you just joined the conversation, and at that, didn't read any of the posts leading up to this point.

Regarding IGP (Integrated Graphics Processor, I thought that acronym was common knowledge among those well in the know with putting together PC's) You (or perhaps, "you guys"?) might consider making it so that there is markings next to IGP to make them visually easier to spot among the lists. At that, if you do include IGP, I wonder why my previous searches didn't turn up anything on two different ones I thought were pretty common.(I forget now though)

[passmark]

I don't care about getting any benchmarks...Then maybe you are in the wrong forum.

Your posts are adding nothing constructive and no longer have anything to do with the original topic of CPU performance per watt.

I am closing this topic.