What is the difference between dual-CPU dual-core?

[jbubrisk]

So here are my questions... Thanks in advance...

• What are the main performance differences between a dual-CPU system and a dual-core system? Advantages/disadvantages as well as general architecture?

• What is the point of having an add-in physics card? Wouldn't it be more cost effective and universal to use dual-core CPU's (and or dual-CPU) systems to do physics?

• Naturally, the physics cards are in their early stages, and don't offer much advantage now, but why not design a standard for making use of the dual-core systems in the first place? They are becoming more and more widespread as the prices fall.

• Off-topic question: Has anyone actually bought a physics card? Why would you at this point?

Thats all for now, maybe more questions later. Thanks again,
John

 

[tommo123]

afaik, dual core should be better due to less distance between both cores to comunicate together i think

an add on physics card takes that load off your CPU and adds more realism afaik (better explosions/effects and what not)

i expect nvidia/ati to add a physx support option in their high end cards in the future tbh, especially as amd and ati merge.

 

[jbubrisk]

The thing is, the "better" explosions and effects that the physics card produced actually slowdown the game. Not to mention the fact that the "better" effects are questionable. But do dual-core and dual-CPU systems accomplish the same tasks (More or less)? Or are they completely different? I also heard about the GFX companies adding physics support to their cards, but the article on Toms Hardware got me to thinking that the CPU (dual-core or whatever) would be better suited for this job. Or maybe its something thats limited to gaming, and therefore limits who develops it (what does Intel want with gaming physics?)?

 

[imported_inspire]

Physics cards slow down your overall framerate, yes. But, not really to a point that you will visibly notice (if you're running a system that's high enough to warrant a Physics Card in the first place...)

 

[CycloWizard]

I think the idea of a physics card is similar to that of a video card. You use a video card instead of placing this load on the CPU because the GPU is specifically designed to perform graphics operations. Since it is specifically geared towards these operations, it performs them much more efficiently than can the CPU. This also frees up CPU cycles to deal with the other facets of the game. Similarly, a physics card is geared towards performing the operations necessary to determine the physics of the game. It should do so efficiently and allow the CPU to focus on other tasks. In the long run, when CPUs are fast enough, graphics and physics cards will probably be obsolete because the loss of efficiency by running the code through the CPU will be negligible since the CPU is just that fast. This probably won't happen any time soon, but in the long run it could be possible.

 

[Calin]

Dual processors use for communication a long link. In the best way (Opteron), it is a hypertransport link that is routed through:
from core to processor pins; from processor pins to socket; from socket to mainboard; from mainboard to the other socket; from the other socket to processor pins; from processor pins to core. This means the link is slower (in speed AND bandwidth), must be amplified to compensate for the length and loss of signal, is limited in width by the traces on the mainboard.
The worst case would be the Athlon MP, in which the connection goes from processor to chipset to processor - Pentium4 communicate thru FSB directly, with no involvement from chipset (except requesting the chipset to wait until communication ends).
Early dual processors from Intel are two processors in a single package - the same limitation apply.
As for dual cores, you could use the hypertransport as in Opteron - just that you don't need amplification of signals. Or you could use a wider, faster connection - as you are not limited by mainboard traces. This way, communication between dual cores (AMD) is twice as fast (in latency) compared to dual processors on a chip (Intel solution - some Xeons of P4 family).

 

[Lord Banshee]

I think the physics card did some mistakes when it featured it to dev's.
1) They should have told devs don't make more physics with the card but just use the card... The problem is that the games that use this physics card are already GPU limited not CPU limited so adding more things for the GPU to render was a no no!
2) They should also have focused on non-effects based physics like collision detection and what not. Let me shot what every i want, and move what ever i want (like real life).

But CPUs are very very general purpose, they are designed to "a lot of different" things. GPUs or any dedicated card is designed to do "one" thing as fast a possible.

3) I also notice that "softimage" a 3D animation software took on this ageia(sp?). This was a smart move and i hope ageia goes after all the 3D Animation software, Particle effects(not GPU driven) and collision detection always takes a lot of CPU power, the effects and detail you do in these programs are nothing you see in games, but in movies so speeding this process up should have many studio money in time and cost of less render farms.

I don't.. it is late this is my rant on this so called Physics Hardware

 

[Calin]

What is the point of an add-in physics card?

what is the point of an math coprocessor? After all, the integer processor is able to do floating point math.

the idea of the add-in physics card is to create an architecture specialised to do JUST the calculations needed for physics. If you say physics use matrix (three by three) , then your architecture is made to make matrix multiplication in one step. The architecture could be so specialized that you would need the same step to multiply two integers.
This is the reason T&L calculation moved from processor to graphic card - the chip in the graphic card was optimised to do T&L, and it soon overpowered the main processor in this task (thanks to speed, memory bandwidth and so on) that now even a quad core with three cores dedicated to T&L could not keep up.

So, the idea of the physics cards is to have a specialised unit that could make specialised operations very fast - as compared to a microprocessor.
By example, a 3x3 matrix multiplication is comprised of 9 multiplications and an addition of the 9 results. If only for physics, you could use 32-bits floats (or maybe even 24-bits floats). The math coprocessor will do at most one floating point multiplication per clock - this could do one matrix multiplication per clock.
Multiplying a matrix by a value is a 9-step operation - and would take 9 steps for the coprocessor. This could be done in a single step on a physics engine.

Now, it might be possible to move the physics calculation to the T&L unit on a graphic card - if the calculations needed fit what the T&L unit can do well, you no longer need a physics. If not, you would be better downloading it on a second core.

Now, would a second core prove to be faster than a dedicated physics card? If so, use a dual core (which will help you not only in games). If not, no. Everybody keeps to its side of the story, and it is difficult to get impartial results

 

[icarus4586]

In the Athlon 64 X2 and Core Duo / Core 2 Duo, the two cores communicate basically through the L2 cache interface, which is very fast. Older Intel dual-core processors (Pentium D and it's derivatives) and 2 separate physical processors communicate over the FSB, which is slower. So as a rule, a dual core processor is faster than a single core processor.

 

[sdemerch]

Originally posted by: jbubrisk

• What are the main performance differences between a dual-CPU system and a dual-core system? Advantages/disadvantages as well as general architecture?

Current options out there do not touch on all possible configurations. You can find a reasonable start by reading this article.

In short, the Pentium D was basically equivalent to dual procs (with a lower class system architecture), the Athlon 64 X-2 has true chiplevel core-core communication (essentially L2 to L2), and the Core2 has a shared L2 cache which is another architecture. Read the article.

You should also note, that dual procs with AMD Opterons means dual memory interfaces so you have twice the theoretical peak bandwidth to RAM. For some workloads this could greatly boost performance (double the possible RAM and with a well compartmentalized algorithm little inter-processor communication would be needed). For others it could damage performance (CPU 2 has to contact CPU 1 to ask CPU 1 to get data from RAM and pass it to CPU 2).

 

[Fenixgoon]

Originally posted by: inspire
Physics cards slow down your overall framerate, yes. But, not really to a point that you will visibly notice (if you're running a system that's high enough to warrant a Physics Card in the first place...)

see anandtech's review of the PPU... framerates dropped like a rock

 

[ForumMaster]

a dual cpu is two sockets, two pieces of silison and seperate cores. a daul core CPU such as the pentium D is simply two regular CPU's on one piece of silicon. "true" or "native" dual core CPU's are two CPU's integerated into one. the advantage is less latency, a good opurtunity to share cache and requiers a cheaper mobo.

 

[Calin]

Originally posted by: Fenixgoon

Originally posted by: inspire
Physics cards slow down your overall framerate, yes. But, not really to a point that you will visibly notice (if you're running a system that's high enough to warrant a Physics Card in the first place...)

see anandtech's review of the PPU... framerates dropped like a rock

But the number of "pieces" resulting from the explosions went up by a non-trivial amount. Overall, I think the PPU had a positive effect (just that the added complexity brought the framerate down like a rock)

 

[borealiss]

What are the main performance differences between a dual-CPU system and a dual-core system? Advantages/disadvantages as well as general architecture?

This is really dependent on the inherent architecture of the processor and core arrangement. Dual cpu's will be better if they are 2 opterons and the system layout is such that each cpu has memory behind it. In a NUMA aware OS running database applications where memory bandwidth was the limiting factor, this arrangement would beat a dual core solution with a single memory controller.

A single small multithreaded application that causes a lot of snoops of caches will run faster on a dualcore amd cpu because cache coherency can be arbitrated through the system request interface (SRI). if this was done over hypertransport there is signifcantly more overhead.

What is the point of having an add-in physics card? Wouldn't it be more cost effective and universal to use dual-core CPU's (and or dual-CPU) systems to do physics?

This really depends on how well a universal physics API is engineered as well as the amount of physics in a game. It's going to depend on the application at hand.

Naturally, the physics cards are in their early stages, and don't offer much advantage now, but why not design a standard for making use of the dual-core systems in the first place? They are becoming more and more widespread as the prices fall.

There probably won't be a need for a standard. Naturally when cpu's become powerful enough to handle the load physics processors become obsolete. Just like hardware dvd decoding. The only case where it won't is if physics gets pushed to the limits like graphics.