Dual Processor vs Dual Core

[PCC-Jon]

Summary:
It has always been a frequent question -- "Will I benefit from multiple processors?" With the growing popularity of dual core processors, the topic is more important than ever! Will multiple processors or a dual core processor be beneficial to you, and what are the differences between them? These are the questions this article will attempt to lay to rest.

View the entire article: http://www.pugetsystems.com/articles.php?id=23

Post your discussion here!

[fellix]

Have considered turning off HyperThreading on the Xeon setup for more accurate comarison with the P-D system?

[PCC-Jon]

That would just decrease the performance of the Xeons, making them fall further behind. The point behind hyperthreading is to maximize the throughput of the CPU by reducing the possible bottleneck in and out of the CPU, if I understand correctly. Another indicator that Intel has a bottleneck of data I/O to the CPU.

[PCC-Jon]

BTW, it has been great to see the popularity of this article...from *just about* everywhere! As of 10am today:

Digg (http://digg.com/hardware/Dual_Processor_vs_Dual_Core): 23,849 visitors ,
theInquirer.net (http://www.theinquirer.net/?article=30609): 6,701 visitors ,
Warp2Search (http://www.warp2search.net/modules.php?name=News&file=article&sid=27990): 1,314 visitors ,
HardwareZone (http://www.hardwarezone.com/news/view.php?id=4149&cid=4): 781 visitors ,
del.ico.us (http://del.icio.us/popular/): 255 visitors ,
Slashdot: 0 visitors (story rejected)

[PCC-Wilson]

Hmm...I wonder why slashdot rejected the story? Too good for them? :thumbsdown

[PCC-Richard]

I just think that slashdot general posts more complex articles. This was written more for the layman.

[Geoff]

Isn't there a fundamental problem with the memory benches (and subsequently others) here in that the dual-core solution has two memory channels, both filled, but the dual-processor solution has *four* memory channels, only two of which are being used? The dual-processor should rock compared to the dual-core in this category at least.

[PCC-William]

Geoff, your are correct! Several readers have pointed this out to me since we launched the article, but unfortunately the benchmarks we have available are all based on systems that have long since been shipped out to the customer. That particular build was intended, as the customer requested, to have 'growing room' for extra RAM. I failed to notice this as I was selecting benchmarks to use for the article.

I will be checking to see if there are any other similar systems we have benchmarks on file for, but as I recall that was the only dual-dual-core-opteron in that speed range that we have done. If I am able to come up with some better comparisons, and if they differ significantly (as I suspect), then I will see if we can update the article accordingly.

[Rhinoman]

In the comparsion both the single processor and dual core are at the same speed(3.0 Ghz). The problem being that what you get in Dual-Core is slower then single XEON (2.8 Ghz vrs 3.8 Ghz). In fact we have clients asking if they should by Dual Processor system at 3.8 GHZ or Dual - Dual-Core system at 2.8 GHZ? What are these comparison like?

[PCC-Jon]

Well, since his article showed that dual-core is comparable in speed to dual-CPU, that would mean that a dual-CPU system at a full GHZ faster would be....well....faster! Seems like this article shows that it is not a difficult comparison, but rather a very straight forward one.

[mharr]

However, it is important to note that if you are running a single program and it is not "multi-threaded", you will not see a benefit from more than one CPU or core.

Is this true? I am under the impression that a user will see some performance increase from a dual-processor system, even if a single primary application does not use or is not designed to use multiple processors, at least on Windows NT-class machines (NT/2000/XP/2003).

While the application is using just one of the processors, the Windows OS will perform its OS "stuff" on the other "unused", leaving less task-switching and interference for the primary application. That could, for a heavy application (database, CAD, game) have a noticable performance increase.

I remember testing this back in the day of Pentium Pro dual processor servers (200 mhz), using SQL Server. Although we limited SQL Server to 1 processor (due to licensing, buying only a 1 processor license), we saw 15-20% performance gain on a dual processor system. Later, testing a web server (IIS on Win2k) on hyperthreaded machine, we saw 10-15% performance increase (IIS default web site at that time could only run on on process, one processor).

So I think that just having an extra processor to let Windows offload its overhead onto, will make noticable difference in heavy applications.

Mark

[PCC-William]

Rhinoman,

It looks like you are talking about the dual-core Xeons compared to a 2x single-core Xeon. Yes, in that case, the dual-core Xeons are not available in as high of speeds as the single core are. However, the comparison I made was between two Xeons and a dual-core Pentium 4. The reason I went with this comparison was that most people don't seem to be very interested in two dual-core Xeons (they tend to run very hot and are quite expensive). In fact, to my knowledge, we have yet to sell a dual-core Xeon. However, anyone who might have been thinking about a two cpu Xeon setup would be interested (or so I hoped) to know that a dual-core Pentium with the same clock speed beat out a pair of Xeons. Sadly, I can't really provide benchmarks to directly answer the question you brought up, since we have had no systems come through with dual-core-Xeons.


mharr,

There will be some benefit from a second CPU or core in Windows, as you describe, so long as something else is going on in the background that needs CPU attention. There are cases where this seems to not be true, however. Gaming is a good example: in many benchmarks dual-core CPUs seem to do no better (occassionally even slightly worse) than single-cores with the same speed. Soon, however, gmaes will begin to be designed for multiple cores and we will surely see a large performance boost. Also, if you had something CPU intesive running in the background there could be a noticable difference. Really, it comes down to how you plan to use your system - which is great, because here at Puget you can get a computer custom-made to meet your needs!

[Rhinoman]

Well, since his article showed that dual-core is comparable in speed to dual-CPU, that would mean that a dual-CPU system at a full GHZ faster would be....well....faster! Seems like this article shows that it is not a difficult comparison, but rather a very straight forward one.

That was not my intend question. The comparison I am looking for is Dual - DUAL-Core 2.8 Ghz compared to Dual-Processor 3.8 GHz. The PROBLEM is you can not order Dual-Cores at the same speed as Xeon single processor. And I am looking for comparision of DUAL-DUAL Core to Dual Processor.

[PCC-Jon]

Sorry I misunderstood. I would still tend to take a simplistic view -- add up the GHz!

Dual Dual-Core 2.8GHz -> 11.2 "effective" GHz
Dual 3.8HGz -> 7.6 "effective" GHz

It isn't as simple as that, but (and correct me if I'm wrong William!) our benchmarks seem to predict that to be pretty darn close.

Of course, the CPU needs to be the bottleneck of whatever you're doing for the dual dual-cores to help, and as said above, you also need to make sure your applications are multi-threaded for best benefit.

[Rhinoman]

Sorry I misunderstood. I would still tend to take a simplistic view -- add up the GHz!

Dual Dual-Core 2.8GHz -> 11.2 "effective" GHz
Dual 3.8HGz -> 7.6 "effective" GHz

It isn't as simple as that, but (and correct me if I'm wrong William!) our benchmarks seem to predict that to be pretty darn close.

Of course, the CPU needs to be the bottleneck of whatever you're doing for the dual dual-cores to help, and as said above, you also need to make sure your applications are multi-threaded for best benefit.

That is the problem! We have application that are multi-thread but also are very CPU intensive. Multi-threading helps you if you can split up the problem. But the shear GHz speed helps to finish the problem. We have
some clients compare speeds with Dual-Core and the problem comes down to the speed of the CPU.

[PCC-William]

Rhinoman,

Not to change the subject too much, but have you considered AMD processors? We have sold several dual socket dual-core Opteron rigs, and they perform marvelously well. And they come in almost the exact same clock speeds as the single-core line, so it is easy to look at benchmarks and see a direct comparison. Also, they scale up more easily than Xeons - you can get 4 or 8 socket dual-core configurations if you really need the extra horsepower. Just make sure that, if you go this way, you have 2 sticks of RAM for every physical cpu-socket (to ensure you get the most bandwidth available via dual-channel memory).

[cercle]

:confused:AMD,they use memory controler,and also real ! four core .Each core is connected to other's memory.Intel use a core of two plus another ,and no memory controler integrated to core.Until next version ,memory controler integrated,I look forward to a real four core,each able to use its total amount of memory,...20XX?:cool: