Processor Power

[martin007]

Hi.


The performance of a processor depends on the clock frequency (Hz) and the instructions per cycle.
It also depends on the number of core and if they work in series or in parallel.

I understand that depending on whether they work in series or in parallel, they perform some functions better than others...


The question is how to compare different processors, with different cores.

[Burgerman]

Benchmarks. A enchmark is a measuring tool.

The best CPU benchmark as used by nearly everyone in the industry is CineBench R23.
Download it for free, run it on your PC and lets see the score?

Heres my R23 multicore CPU score. It gives single core data too. But theres little that uses single cores in todays market. Maybe the odd game. Many CPU run a single core faster than multicores. My overclocked (by 40% multicore) CPU needs the water cooling, uses 16 cores 32 threads. And is complete overkill.

The problem is that modern CPUs cannot be rated in this way. They go as fast as the cooling system, and motherboard current (Amps) VRM capability allow. So the same CPU score depends on cooling, power and how you configure the Bios, as well as the actual mothrboard used. Which is why I can get a 40% higher frequency than stock...

I get over 32,000 scores continually with no throttling or even high temperatures. Most with same CPU will only get 24,000.

[martin007]

What I want is to rate a processor without physically having it.

[Burgerman]

Realistically you cant. Again then you could look at AVERAGE benchmarks.
Those are uploaded by users.
There isnt another way because all modern CPUs depend on the cooling, and the motherboard specifics to determine the actual speed.

Variation is not only in the various cooling and motherboard VRM power but in the aplication and the need too

If you were building a workstation, you would want a CPU with high single or maybe 4 core performance as this is what you will mostly be doing.
If you were building a PC for say solid modeling, or movie editing, or a server for a lot of users in a business, you want a high core count, as these sorts of aplications use as many cores as you can give them. So a relatively low frequency CPU with high core count will be much faster. Games on the other hand tend to use 2 or 3 cores, and so high frequency, and high memory bandwidth is more important/essential or the modern GPU will be "throttled" (bottlenecked) by the CPU capability at least in medium or low resolution usage where frame rate is high. At 4K gaming, the CPU is less important, as the GPU will be the bottleneck. So again before you decide on a CPU you need to know what it is to be used for. Or you cannot compare them. There isnt a single number that allows that.

Its like evaluating all cars with a number. Without thinking about the usage, number of passengers etc. Is a big heavy 4x4 or a two wheel drive sports car going to be number 1? Different uses need different cars. BOTH may be great at different purposes.

If in doubt, overkill everything! £££££ Fast CPU, big cache, high memory bandwidth, loads of cores and massive overclock capability and then add an overkill GPU as this too is used for CPU calculations, and gaming, and so build a money no object monster PC like I did. Its total overkill, and can do anything. At a price.

[martin007]

When you say workstation, do you mean a multipurpose personal computer?

[Burgerman]

Take this image. This really needs a powerful computer to work on. https://www.wheelchairdriver.com/board/ ... &mode=view
If I run DXOs "deep learning" noise reduction software on this, and I did, it takes a rediculous 9 minutes to output the finished image. On my 4 CORE laptop. If I do the exact same thing on my 16 core 5.2Ghz desctop PC it takes around 5 seconds... It does it so much faster that the fans dont even get time to spin up.

So If I batch process 100 photos on my laptop, a typical photo session like yesterday. It takes (or would take) 900 minutes! If I do it on my desktop PC which I did, it takes 500 seconds. Or 8.3 mins.

Its not just the CPU though. Its the memory quantity, as well as memory bus speed, CPU cache, and the super fast NVMe generation 4 drives. The PC is immeasurably faster in every way. The CPU in the PC doesent even do the hard work. The GPU and the programs "deep learning and artificiel intelligence" runs on the faster GPU in parallel. So again even this depends on all the other components not just CPU.

The R23 benchmark ONLY tests CPU capability.

[Burgerman]

Then theres another problem even when using benchmarks...

When I bought my AMD 16 core 32 thread CPU i noticed that every review was with the motherboard manufacturers DEFAULT setting with all overclocking set to OFF. As specified by AMD to 3rd party motherboad vendors. This, on a decently cooled, high power system, like my motherboard (220Amps VRM, yes CPUs take real Amps) means the advertised performance figures, as well as the Youtube tests you will see, have around 40% of the potential performance IGNORED and left on the table! Why? Because it takes too much cooling for most systems to manage. And more Amps than they can give. It DOUBLES the normal watts (TDP To 250 watts)

Whereas the INTEL CPUs and matching motherboards have much of this enabled as a default setting as stock. So they LOOK BETTER on tests, and on paper. With decent motherboard and cooling without "overclock" settings. But theres very little left to manually overclock later. Headroom all used up! On the AMD systems at this time, you can literally change one setting in the bios that allows factory overclock settings. Not the best way but 1 click... But then they overtake the intel offering at the time of the testing. So even the motherboard manufacturers recommended default setting has to be considered. It can make a 40% difference in performance! Thats a HUGE difference when talking about CPUs. And the recommended motherboard defaults vary between manufacturers.

[Burgerman]

When you say workstation, do you mean a multipurpose personal computer?

Yes. But this depends of WHAT you will do with it.

If I built a do anything computer it COSTS! Many cores (16 to 64), tons of memory like 128GB fast single thread performance, expensive 3090 graphics card with good calculation capability and large fast drives. You dont get change from 5K. I built that! So it can run and do anything. Fast.

If I built a gaming computer it needs fast frequency 4 to 6 core CPU, maybe 16GB memory, and an expensive graphics card - depends on resolution. A 4K screen/game needs a 3090. A lower resolution like 1080p needs 1/4 the power... So a 3060, at 1/4 the cost. And it needs a small cheap boot NVMe drive and a 2GB SSD for games storage.

If I built a workstation, for say Office, or surfing, etc almost anything at all less than 10 years old, and 8GB RAM any half decent old CPU and an SSD will be fine. Anything else is a waste of money. Non of this akes muh to run.

If I had specific needs I would vary this depending.
A server doesent need expensive graphics cards, it needs multi core CPU, lots of RAM, lots of storage, fast connectivity etc.

[martin007]

How many gigahertz is the processor in your laptop?
What processor is it?

[Burgerman]

Its a 1.90ghz 4 core, 8 logical thread (it is a hyperthreading CPU).
But again its just not this simple. Right now its running at just 400mhz... At lowered voltage. Because I am not stressing the CPU. So it throttles to match load.
When I do run something that needs more CPU speed it ramps up to 3.2Ghz if its a single core thread, or around 3Ghz on 2 cores, 2.8Ghz on all cores at once. Again this all depends on the System bios written by Dell, and on cooling. And available power. So if I turn the laptop fans to auto they will get noisy if its busy and it stays at 3ghz plus all core. If on battery I set that to silent then the CPU throttles and I get whatever frequency keeps the cpu CORE temp below 100C.

So the stated 1.9Ghz is totally meaningless!

It all depends on what Dell allowed it to do, in the Bios. They can allow and control its voltages, max current, frequencies, temperatures etc and they configure all this to match their cooling and the motherboard power characteristics. So 2 different brand laptops will likely perform completely differently with the SAME CPU depending on each laptop manufacturers implementation, max allowed TDP etc.

[martin007]

Suppose these two processors with the same hardware.

Intel Core i5-650 3,20 GHz
Cores 2
Threads 4

> https://technical.city/en/cpu/Core-i5-650


AMD PRO A6-8570 3.5 GHz
Cores 2
Threads 2

> https://technical.city/en/cpu/Athlon-64-X2-4200-plus


Intel processor is faster than AMD, but...
Any idea how to score (accurately) those two processors?

[Burgerman]

Compare in similar systems with benchmarks.

No other way.

Even that doesent really help. One may be really fast at say compressing a huge zip file but slower at compiling some executable code. The other may excell at rapid compiling, but be crap at zipping a huge file.

Also depends on the particular motherboards max TDP and other settings. Which vary by manufacturer.

So only testing when using for your intended purpose can really tell you the answer. Unless one os just massively faster at everything (and so will be more expensive).

[martin007]

OK.

[Burgerman]

Look up on here for e.g.

https://www.cpubenchmark.net/

[martin007]

I write it down.
Thanks.