How do I compare CPU speed versus number of cores for performance?

[Yousuf Khan]

That's amazing!

Well, at least with those set of benchmarks that they were using.
Another site, Passmark, uses a different set of benchmarks and rates the
processors this way:

http://www.cpubenchmark.net/
AMD A6-1450: 1653, http://is.gd/qiJCUx
Intel Celeron 1007u: 1450, http://is.gd/MegWjZ

Still pretty close, but the A6 is slightly ahead. Doubtful if it will be
noticeably faster though. You should check out various rating websites
to make your decision.

Yousuf Khan

 

[Darklight]

She's a typical user.
Nothing special.
She sometimes has multiple browser windows up.
And, she might have an Excel spreadsheet or Word document up.
That's pretty much it.
Very basic user.

My personal view is the quad core will remain cooler hence the life of cpu
will be longer. A quad core cpu the cores will do less work because of
sharing between the cores. the dual core might be faster in areas but each
core will do more work and hence be hotter. My friend has a Intel dual core
cpu media laptop and i was running an app called dvdstyler on it and the
temp of the cpu was between 95-98C at idle it was about 45C.

My fx-8350 8 core at idle 27-30C BF4 55-60C cooler H80i but that is
irelervant.
I have seen an intel cpu quad core with h80i cooler gaming running at about
90C+. The intel might be better at gaming. but the heat is off putting.

But hay in the end it's your choice and you will have to live with the
consequences. Heat is not your friend.

 

[Paul]

My personal view is the quad core will remain cooler hence the life of cpu
will be longer. A quad core cpu the cores will do less work because of
sharing between the cores. the dual core might be faster in areas but each
core will do more work and hence be hotter. My friend has a Intel dual core
cpu media laptop and i was running an app called dvdstyler on it and the
temp of the cpu was between 95-98C at idle it was about 45C.

My fx-8350 8 core at idle 27-30C BF4 55-60C cooler H80i but that is
irelervant.
I have seen an intel cpu quad core with h80i cooler gaming running at about
90C+. The intel might be better at gaming. but the heat is off putting.

But hay in the end it's your choice and you will have to live with the
consequences. Heat is not your friend.

With mobile gear, your best indicators of power consumption,
are runtime in hours, and battery size in watt-hours. Those
give you some idea whether one design uses more power than the
other. Extracting TDP numbers from the table, is relatively
meaningless. For example the 8W (or 9W depending on which
slide deck you look at) of the AMD Temash, rises to 14W if
Turbo is enabled in the design. We can't get any reliable
information what's in this particular product, in terms of
whether Turbo is enabled or not. Turbo does not change the
Passmark number, because the processor still runs at 1GHz
when all four cores are fully loaded. It only makes single
threaded (SuperPI) performance a little closer to the Celeron.

It should also be pointed out, that AMD attempted to change
the definition of TDP, in a bid to make their numbers look
better. We can't even trust what a "TDP" means any more.
(The AMD and Intel definitions could differ.)
Thermal Design Power is a number used to aid in designing
a heatsink for the processor. It's not a thorough or precise
indication of power usage. My 65W TDP Core2 Duo E4700, used
36W running Prime95 and 13W at idle, just to give you some
idea how much "TDP" deviates from "reality". My 65W E8400
weighed in at 43W under the same conditions (and is
designed in the next process generation). The idle power
was about the same. If you were to say "oh, that processor
is a 65W processor", you'd be wrong. We don't really know
what it is, until measuring it. TDP does not equal
real power. The definition is a lot more complicated,
and involves "marketing" as much as "engineering".

Finding a laptop/tablet review site, one that uses the same
battery life test suite for each, will give some idea which
of them sips power. Since we can't trust the processor
companies when it comes to characterization, we have to
rely on other metrics to judge them. Some people don't
care about battery life, while for others battery life
is everything. In which case, a review site is where
such a person should be. And a review site that compares
apples to apples.

When it comes to the AMD Turbo, one web site was claiming
that the Temash Turbo feature could be turned on, on a
laptop, because it "had a fan". Then, I visit another site,
which is actually reviewing a laptop with Temash in it,
where Turbo is enabled. There is no fan on the unit.
I had to find a third site, with a "picture gallery", and
a photo of the bottom of the unit, to verify there
were absolutely no vents on the thing.

A lot of these sites, would rather repeat the dribble
they find in the slide deck, than clamp an ammeter
onto the power lead feeding the processor, and actually
characterize it for themselves.

Given that the sites are so untrustworthy, about
all we can hope for is a battery life test. Where
the same test suite is run on competing units.

Paul

 

[Paul]

A problem in determining both what actually takes place with "multicore"
vs sheer processor speed is and how the processor OEM implemented the
multicore integration.
Multicore processors can function as more or less separate CPUs sharing
some resources, or an alternative is to have a higher level of
integration that allows them to pass off "work" back and forth, in ways
that minimizes use of external (slower) resources.
Naturally, to confuse matters more, the OEM's may not really fully
define how this works, and how to best take advantage!
Although, if you happen to be a large bulk purchaser, you may have more
access to support and the details than others.
I've seen (in the past) software implementations and hardware as well.
Sometimes, due to the general PC environment and differences in
hardware, the implementations worked very well on only a limited number
of the various processors/CPU's of the time.
Others used a software implementation, when the hardware might have done
things better. The exact opposite was also true.

When all is said and done, the test programs can only show a small part
of the relative performance, so multiple programs and in application
behavior is needed to really compare performance.

For desktop systems, OEMs play a minimal role in how the hardware
works. They just follow the instructions, and produce a cookie cutter
design. It's what I'd do if building one. Why ? Because there
are no useful "hardware spigots" to offer improvements. The functions
of the silicon chip are so fixed, there's no room for innovation.

For example, if you see Anandtech review six different LGA1150
motherboards, the performance is the same within a couple percent.
And the only difference, might be one of them cheats by setting the
BCLK a megahertz higher than they should (so they'll win the web
site reviews when they happen). Doing a little "clock cheating"
is the only difference.

At the other end of the computing universe, the OEM (HP/SGI/Sun) add
custom chips to interconnect the processors and maintain cache coherency.
Their efforts succeed to varying degrees, in which case a person
shopping for an HPC (High Performance Computing) system has to be very
careful about what they're buying. (I.E. Check the benchmarks.) Some of
these systems are so seamless, you can install a bog standard Windows OS,
and "fool" the OS into thinking the computer is just an over-endowed desktop.
At least, until the limitations of the OS prevent Task Manager from drawing
the performance graph properly.

(Some portion of an HPC system given to Stephen Hawking... It's a
computer holding racks and racks of CPUs, and some sort of high speed
coherent network to connect them together.)

http://techpubs.sgi.com/library/tpl...b=bks&fname=/SGI_EndUser/UV_2000_UG/ch03.html

So an OEM who designs custom silicon, to build compute complexes with
very high socket counts, their efforts do make a difference. For
companies like Dell or Acer or Gateway, much of their stuff is
cookie cutter design. All the same. All perform the same (when the
clock frequencies are the same). Those desktop systems, there is
no real value added to them. It's up to the OS to figure out
the internal architecture, and make the best of it.

*******

Once we get past that stage, then it's a matter of OS design. And
how threads work. I just found a nice article here, that gives a
flavor of how it works. I wasn't aware that the kernel of the OS,
threaded at this level.

http://www.cs.uic.edu/~jbell/CourseNotes/OperatingSystems/4_Threads.html

*******

Even though there are claims that Windows 8 is "NUMA aware",
I still hear stories about certain utilities not
running that well on big systems. For example, if you run
7ZIP on a dual socket G34 system (total 32 cores), it doesn't work
very well. And that can happen, because when you malloc in a thread,
the memory allocated on the processor, is over on the wrong
socket (a distance away from the core doing the computing).
There are few web sites with the skill or interest in testing
stuff like that.

(A web page about NUMA - typically seen on 2/4/8 socket
server motherboards, and still smaller than HPC systems
with racks and racks of blades.)

http://www.qdpma.com/systemarchitecture/NUMA.html

Here, a thread of 7ZIP compression effort, could ask
for a memory allocation, and the memory might come from
a socket other than the socket where the computing is done.
Being "NUMA aware" means not making mistakes like that.
The four links between Opterons in this figure are termed
"coherent links" because they maintain cache coherency,
so all the processors can figure out where the most
recent copy of some memory is. It might be contained in
the L1/L2/L3 of one of the other processors. And a coherency
protocol, is all part of keeping track.

http://www.qdpma.com/SystemArchitecture_files/Arch_Opteron.gif

Paul

 

[Darklight]

Given that the sites are so untrustworthy, about
all we can hope for is a battery life test. Where
the same test suite is run on competing units.

Paul

Concerning to above i remember the gadget show on channel five done such a
test. And come to think of it, it is a very important point.

 

[Paul]

Don't know which "townie" you live in, but up here in Canada, 4WD is
actually useful inside the town. Snow, ice, freezing rain, etc.

Yousuf Khan

Actually, tires are useful. I'd slightly prefer excellent
winter tires, to 4WD and crappy tires. The problem is
finding just the right kind of tires.

I think it's interesting, that you can make tires, where
they have what looks like a winter tire tread, and it
doesn't *work* like a winter tire tread. Probably has
something to do with stepping on someone's patent. Only
my Michelin winter tires were good.

http://en.wikipedia.org/wiki/Siping_(rubber)

Paul

 

[Gene E. Bloch]

Don't know which "townie" you live in, but up here in Canada, 4WD is
actually useful inside the town. Snow, ice, freezing rain, etc.

"Townie" is slang for town dwellers, so he presumably means people who
live where there's no real reason to have 4WD.

Quoting Wolf:
"It's like townies buying a 4WD truck with a tow-package but they have
no trailer to hitch up to it, nor do they ever go into the back country
where 4WD might help them get out of a swampy hole in what some years
ago passed for a road.... ;=)"

The young son across the street has such a thing, a big diesel-powered
truck with super-size tires, and never a splatter of mud on it. It's so
large and loud that I gave it a name, even though I am not a car namer.
I call it Grendel.

 

[Ken Blake]

On 31/01/2014 2:36 PM, Gene E. Bloch wrote:

Yeah, I know, I was being sarcastic.

Considering the quotation marks around "townie," I thought you might
have been, but to tell the truth I wasn't sure.

 

[Gene E. Bloch]

Yeah, I know, I was being sarcastic.

OK, you get at *least* 12 points for fooling me

 

[Gene E. Bloch]

On Sat, 01 Feb 2014 09:21:20 -0500, Yousuf Khan

Considering the quotation marks around "townie," I thought you might
have been, but to tell the truth I wasn't sure.

That's part of Yousuf's 12 point bonus. I thought he meant he wasn't
sure what Wolf meant by the word

 

[G. Morgan]

How would you compare dual-core 1.5GHz versus quad-core 1.0GHz for performance?
(Windows 8, touchscreen)

My sister asked me whether it's better to get the ASUS X200CA-HCL12050 or the
Dell Inspiron i3135-3750slv inexpensive $300 laptops.

Looking at the specs, the main difference seems to be:
Asus: dual core, 1.5GHz
Dell: quad core, 1.0GHz

Which helps in performance more, additional cores? Or faster speeds?
Note: I realize both help - but which is MORE important, cores or speeds?

For multitasking, I would go for more cores. The 500MHz difference in
clock cycles would be barely noticeable in speed, but if you have
multiple programs running you can take advantage of the load balancing
across 4 CPU's instead of two. It will be "faster" overall if you have
more cores, even if the individual cores do not have the higher amount of
clock cycles in the two-core configuration.

It boils down to what you mostly do with the computer. For gaming, I'd
rather have two cores (if the game is able to use both at the same time)
running at 3GHz than four running at 2GHz.

For general purpose use, I'd take the four cores so a media center (for
instance) can get it's own core and office applications will not bog-down
(steal CPU cycles) from the media center that you may want to run 24/7.