a core as a physical computational unit as opposed to a hardware thread as a virtual computational unit;

The description reflects todays meaning of "physical CPUs" (that is, CPU cores).
Ten years ago, people would have distinguished between hyper-thrading and multiple processors.
Twenty-five years ago, a computer tower was sometimes misnamed as "CPU".
About 40 years ago, the term CPU or central processing unit would have not referred to a microchip,
but a huge separate processing box containing the ALU.

Kerr Avon wrote:

I was just looking at the latest Steam Hardware and Software Survey report at https://www.techspot.com/amp/news/70022-quad- … s- […]
Show full quote

I was just looking at the latest Steam Hardware and Software Survey report at https://www.techspot.com/amp/news/70022-quad- … s-10-reign.html, and it lists the 'physical' CPUs used by Steam users. But to me the list seems to represent the cores of the CPUs used by gamers, not the number of separate CPUs used by any given user, so surely the list should say "CPU Cores", as no matter how many cores a single CPU has, it's still only one physical CPU? Is the "Physical CPUs" heading wrong, or do some people really refer to each separate core as a physical CPU?

When looking at the table, one can see that the single and triple cores are very very low and dual and quad are almost 100% when combined.
This is most likely due to them being cores and not physical CPUs.

A Q6600 is not 4 CPUs, but a quad core CPU. Steam probably left it this way, as it's not hard to understand what the table actually meant for the vast portion of users using this table...even though imo it's incorrect when designating cores as physical CPUs.

Otoh (and this is where Jo22 was getting at I think), a core is in fact a physical core, and not (say) "emulated" like hyperthreading cores are?

Personally I would've prefered them calling it "CPU cores" instead of just CPUs.

edit: I do find it odd that there's so few hexa and octo cores?

edit2: Actually...as I'm thinking about this, I'm starting to have some doubts about what I mentioned, it does seem logical to call the separate cores a "CPU", as it basically is multiple dies on a single chip.
Central Processing Unit...

But still...what would Steam call the real physical chips in the above table? I still think it's easier to distinguish between Dual-core dual-CPU and quad-core single-CPU, as "core" in this context also refers to the die area which makes up the transistors of the CPU.

We all know that whoever prepared the table need to be educated on CPU terms.

It is more likely that users running steam use single cpu's with multiple cores, then using dual xeon cpu's. Not that there aren't any users working with those setups, but to me it is more likely that the defenition would have been cpu cores.
I think that with the release of amd's ryzen 5 and 7 we will see the number of 6 and 8 cores cpu's rise in the futher.
But since most games still prefer faster cores versus multiple cores, we have to wait until developers start using the processing power of multiple cpu's cores.

Vmware uses the terminology of sockets vs cores. Could probably suggest Valve add a new category but would probably take them a decade to add it.

i wonder how they would interpret my bulldozer then, looks like 6, but really really its just 3

mrau wrote:

i wonder how they would interpret my bulldozer then, looks like 6, but really really its just 3

The jury is still out on that one: https://arstechnica.com/gadgets/2015/11/amd-s … zer-core-count/

Tetrium wrote:

edit2: Actually...as I'm thinking about this, I'm starting to have some doubts about what I mentioned, it does seem logical to call the separate cores a "CPU", as it basically is multiple dies on a single chip.
Central Processing Unit...

Both Intel and AMD call the entire chip a single CPU.
For example, on Intel's ARK, one of the specs is the 'Max CPU configuration'. Take a Xeon with 8 cores and 8 threads that can be run in a dual-socket system, and Max CPU reads 2: https://ark.intel.com/products/52580/Intel-Xe … 6-GTs-Intel-QPI

The problem probably stems from the fact that the first multi-core CPUs were literally two CPU dies put in a single package, on a single socket. So all existing software saw it as two physical CPUs. You had to perform special detection to see what kind of CPU it really was. Same with HyperThreading, legacy software didn't know what HT was, so if it just called up the amount of cores in the usual way, it got the logical core count, not the physical core count, not knowing that there was a difference.
Steam probably never changed their original code, so it still counts cores as CPUs. So they're simply wrong.

Scali wrote:

mrau wrote:

i wonder how they would interpret my bulldozer then, looks like 6, but really really its just 3

The jury is still out on that one: https://arstechnica.com/gadgets/2015/11/amd-s … zer-core-count/

this is from 2015, i would assume amd won that easily - this is something that was well documented in the computer press before i bought mine, i just wasn't that interested in max multimedia/game performance;

mrau wrote:

this is from 2015, i would assume amd won that easily

Lawsuits tend to take years. I couldn't find anything beyond the initial lawsuit news from 2015, so I assume the jury is still out on that one, as I said. No verdict reached yet.

Edit: I found this response from AMD: https://www.theinquirer.net/inquirer/news/243 … -claims-lawsuit

An AMD spokeswoman told The INQUIRER: "We believe our marketing accurately reflects the capabilities of the 'Bulldozer' architecture which, when implemented in an 8 core AMD FX processor is capable of running 8 instructions concurrently."

If that is the level of their defense, they won't stand much of a chance. This is total nonsense of course.
x86 cores have been superscalar since the first Pentium, allowing them to run multiple instructions concurrently per core. Trying to claim that the core count is in any way related to the capability of instructions running concurrently is complete rubbish.
In fact, best-case, an 8-core Bulldozer would run 16 instructions concurrently (each module has 4 ALUs, each thread has a two-way superscalar integer pipeline). So AMD doesn't even understand their own architecture.

Maybe the marketing team wanted the message simplified so people got the gist of the argument ; not everything needs to be taken absolutely literally all of the time; ought to be reasonably obvious what they mean.

ratfink wrote:

Maybe the marketing team wanted the message simplified so people got the gist of the argument ; not everything needs to be taken absolutely literally all of the time; ought to be reasonably obvious what they mean.

This is not 'simplified', this is completely wrong and inaccurate.
In theory it is even possible to design a single-core CPU that executes 8 or more instructions concurrently. Cores and concurrent instructions have very little, if anything, to do with eachother, and the number of concurrent instructions is in no way any indication of how many 'real cores' a CPU may or may not have.
So as I say, if this is their line of defense, they're going to lose, since the argument is baseless.
By their logic, a Pentium classic is a three-core CPU, since it can execute 3 instructions concurrently. Does anyone agree that a Pentium classic is a three-core CPU?

Of couse not. Even the Zilog 80 was superscalar somehow. This AMD guy is probably just one of these marketing people.
Maybe he was referring to threads or whatsoever (even that wouldn't have been fully correct, though).
Somehow it makes me sad people are nolonger able to distinguish between dual processor, dual core and hyper-threading.

I mean, during the Windows XP era, that was one of the key differeneces between the different editions!
Windows XP Home supported a number of multiple cores, while only XP Professional supported 2 real, physical processors.

The generation gap: Windows on multicore

(And if we're going farther back in time, we'll see that even Windows NT 3.1 supported multi-processor machines.
The first one was based on multiple i386 processors. I once saw an article with a picture of it..)

http://www.os2museum.com/wp/nt-3-1-smp/
Re: The Systempro Project; Benchmarking (386 class)

Jo22 wrote:

(And if we're going farther back in time, we'll see that even Windows NT 3.1 supported multi-processor machines.

Yes, there were two different kernels in early NT. At the blue bootup screen you could see whether it loaded the single or multi-processor configuration.
See also: http://windowsitpro.com/windows/adding-proces … ling-windows-nt

Did SystemPro machines with multiple asymmetric 386s use a third kernel build with Windows NT 3.1 or was there some custom juju in the SMP kernel that allowed some offloading to the second CPU?

ratfink wrote:

Maybe the marketing team wanted the message simplified so people got the gist of the argument ; not everything needs to be taken absolutely literally all of the time; ought to be reasonably obvious what they mean.

Or maybe they expected to get too many customers asking them questions about what the difference is between "cores" and physical CPUs and just couldn't be bothered with it.

Or the app was made by someone who either doesn't know the difference or simply doesn't care.

Personally I think this is kinda unacceptable, it's plain stupidity.
If someone were to have to manually count the amount of traffic across a section of road, and 24h later he'd tell me "Oh, I counted bikes as cars too!!!" this is unacceptable to me. A bike is NOT a car! And a core is NOT a CPU!

And besides, if stuff gets dumbed down like that, it will only make the idiots more lazy and more successful as nobody will judge the crap anymore that these idiots reproduce (like how someone (or someone's) ended up mixing up CPUs and cores and this ended up with a caterer of millions of gamers..who will spread the stupidword further 😵 ).