Intel Pentium D Processors

[Coola]

alright, so im looking for an intel cpu for multitasking. I wanted to know what's the difference between the

Pentium D Processor 820 2.80GHz 800FSB

and

Pentium® D Processor 830 with Dual Core Technology (3GHz, 800FSB)

Is it worth $136 (According to Dell) to upgrade to a 3.0GHz from a 2.80GHz?

 

[Pete]

The difference is 200MHz per core, AFAIK. Worth it? IMO, no--that is, unless WMV HD videos require literally 3GHz, and only the 830 will do. I'd like to think the dual cores will be enough for that, though, even at just 2.8GHz.

You could probably put that $136 to a lot better use.

 

[Coola]

what about FSB? what is that

 

[wireframe]

what about FSB? what is that

The FSB (Front-Side Bus) is the bus through which the CPU communicates with the rest of the system (North & South bridges, memory). An FSB of 800 MHz translates to 6.4 GB/sec on the Pentium 4 platform and the more the merrier (it also turns out that running DDR-400/PC-3200 in dual channel equals 6.4 GB/sec, so that works out nicely).

I would definitely go for the 3.0 GHz model. Even if you have two CPUs/cores, it is always good to have better single thread performance when you need it. 3.0 GHz Pentium 4s have been around for ages now and they aren't exactly the pinnacle of performance. You will be bound to the same performance (roughly) of a Pentium 4 3.0E when running single thread applications. Of course, price and what you intend to use the computer for mostly will have to determine which you opt for.

 

[Coola]

just basically web surfing, lots and lots and LOTS of multitasking, minor (very minor) video editing, photoshop, flash mx stuff like that.

I just wanted to know if it was worth the money to put into it. Will 200mhz really make that big of a difference for the price?

I've decided that I'll be aiming to get a Dell Dimension 9100, though concerned about purchasing it now and then the next version of windows pops out. So i'm trying to build a system *now* that can run longhorn and other apps down the road.

Also - whats the top of the line intel cpu as of right now and when will the next one be available?

 

[epicstruggle]

you know, i hate pimping amd, but they do have the best product out there. Only problem is that youll have to go and get something other than a dell.

epic

 

[RussSchultz]

In our tests, the Pentium 840EE provided the best compile times in our environment that can spawn 8+ threads at once.

The AMD-4000+ was almost 50% slower.

Granted, we didn't test the X2 class, but no big name company sells those yet, and we didn't want to buy from a mom-and-pop.

 

[wireframe]

just basically web surfing, lots and lots and LOTS of multitasking, minor (very minor) video editing, photoshop, flash mx stuff like that.

Multitasking is one thing, SMP/Dual-Core is targetted at "compute intensive" multitasking. While it may make most things feel slightly smoother, it's really aimed at situations where one of your tasks requires 100% CPU usage. I am sure that if you run Photoshop and Flash MX (web design?) concurrently, you will see some benefit, but even this is more on the luxury-fringe of what SMP is really needed for.

Whatever you do, get lots of RAM. That helps even single core CPUs get multiple jobs done without too much hitching. Perhaps this would be a good reason to save $150 on the CPU and spend it on more memory.

I just wanted to know if it was worth the money to put into it. Will 200mhz really make that big of a difference for the price?

Well, it's a pretty bad deal if you are looking at single threaded performance $150/200MHz, but you need to double that MHz number when considering multi-threaded apps.

I've decided that I'll be aiming to get a Dell Dimension 9100, though concerned about purchasing it now and then the next version of windows pops out. So i'm trying to build a system *now* that can run longhorn and other apps down the road.

Perhaps you would be better served by waiting for the next Windows release before buying a new rig. This would also give you time to save up extra cash to really splurge on it (so 200MHz or $150 here and there are not your main concerns). There is something to be said for a balanced high-end system. Don't think a "fancy new CPU" will solve all your problems.

I don't think a system based on Pentium D 8xx will have any problems running Longhorn, but you may be missing out on a few features that may be introduced with that operating system. This is always a danger so either you buy without expectation or you upgrade your hardware in time with OS release cycles.

Also - whats the top of the line intel cpu as of right now and when will the next one be available?

The Pentium D 840 (3.2GHz) would be the top of the line dual core for desktops (there is also the Extreme Edition (Pentium XE) 840, but I fail to see what this offers over the D model at the moment. Is it really only HyperThreading?).

Like Epicstruggle said, you should at least look into what AMD offers. Maybe even consider a traditional SMP (AMD Opteron or Intel Xeon) board. These will require registered RAM, but you might get good/better performance for less money. I'm not going to say that Intel Pentium Ds are bad, but if you really want to pound on a dual-core/SMP system you would probably be better off with an Athlon X2 or Opteron. The Pentium D is NetBurst based (Pentium 4) and those processors suffer from memory access with just one core. They are relatively good at streaming operations like video encoding, but once you go more general the Athlon stomps all over them.

One of the main selling points of the Pentium 4 was Hyper-Threading, the ability to run multiple (two) threads concurrently on one core. With dual core this advantage is removed so any argument that a Pentium 4 multi-tasks better than an Athlon 64 is no longer important. The Pentium D dual core and the Athlon X2 both have two cores and no Hyper-Threading and in this scenario the Athlon makes the Pentium look rather weak by comparison.

Sorry to sound so negative about something you are obviously excited about buying, but I feel I would do you a disservice to at least not mention these points.

 

[wireframe]

In our tests, the Pentium 840EE provided the best compile times in our environment that can spawn 8+ threads at once.

The AMD-4000+ was almost 50% slower.

Granted, we didn't test the X2 class, but no big name company sells those yet, and we didn't want to buy from a mom-and-pop.

Hardly a fair test, but still interesting results. I really think you need to get hold of a X2 4800 (which is basically two 4000+) and test that some more. Would be interesting to see the results. As it stands, you are comparing a CPU with two cores and 4 virtual processors to a single core. If we forget about HyperThreading, you still have an enormous advantage in the dual core Pentium XE.

 

[Coola]

Waiting would be difficult but it would make sense. What's the supposed ship date for Longhorn, anyone know? The OS i was looking to get was Microsoft Media Center 05 which brings me to another question, will longhorn already have media center functionality out of the gate?

As for RAM the 9100 comes with 4 dimm slots, and I was intent on getting 1GB starting off and getting more if needed down the road.

No sweat on hounding the system, that's what i want to hear. I want to know that my investment is worthwile and not a complete waste suckered in by corporate marketing.

 

[RussSchultz]

Hardly a fair test, but still interesting results. I really think you need to get hold of a X2 4800 (which is basically two 4000+) and test that some more. Would be interesting to see the results. As it stands, you are comparing a CPU with two cores and 4 virtual processors to a single core. If we forget about HyperThreading, you still have an enormous advantage in the dual core Pentium XE.

Oh, I agree, it wasn't a terribly fair fight, but here's the rest of the numbers:

Machine #1:   Precision 670, Dual 3.2 Xeon processors (3.2 Ghz, 2MB Cache)
                        10K RPM drive

Machine #2:   Precision 380, single Intel 640 (3.2 GHz, 2MB cache)
                        10K RPM drive

Machine #3:   Precision 370, single intel 3.2 Ghz (I don’t know how much cache)
                        10K RPM drive, Hyperthreading turned off.
 
Machine #4:  Precision 360, single intel 3.0 Ghz (I dont know how much cache)
                        7200 RPM drive, hyperthreading turned on.

Machine #5:   Precision 380, Intel 840EE (I believe.  3.2Ghz, 1MB cache each processor, hyperthreading enabled)
            Single 10K RPM drive

Machine #6:   HP workstation, AMD Athlon64/4000+ (2.4Ghz, 1MB cache)
            Single 7200 RPM drive

Machine #7:   HP workstation, AMD Athlon64/4000+ (2.4Ghz, 1MB cache)
            Single 7200 RPM drive, single 10K RPM drive (used as data drive)

#1                    10:10
#2                    11:36
#3                    12:02
#4                    12:55
#5                    10:20 (2 threads)
                       7:44 (4 threads)
                       7:11 (8 threads)
                       8:26 (8 threads, hyperthreading turned off)
#6                    16:36 (2 threads)
                      15:45 (8 threads)
#7                    14:45 (8 threads)

Extrapolating from this (that a single core 3.2Ghz Intel 640 with hyperthreading turned off beat an Athlon64-4000 by a considerable amount), an Athlon X2 wouldn't have beat the 840EE, or likely even an 840.

Granted, on one of our other test suites (which is an older application and single threaded), the Athlon turned in the highest score by a smidgeon. Sadly our new environment is the one where the Athlon system gets handily spanked.

One major difference is the the multi-threaded application is also very IO bound, which might be the achilles heel of the particular system from HP when compared to the Intel chipset.

Of course, realise that these tests are very targetted toward our needs (highly parallel compilations) and that they are not indicative of real world performance.

 

[wireframe]

RussSchultz,

Those are some interesting numbers and I thank you for sharing them. Leaving the Intel v AMD fight alone for a while, it is really interesting to just look at the 840 XE versus the older Xeon. A cheaper, more compact package running at the same clocks, but a healthy increase in performance. This is the kind of stuff we all like to see, regardless of any praise we have for one particular brand or architecture over another. I wouldn't be surprised if the XE 840 edged out the X2 4800+ and they are at the same $1,000 price point.

I am not sure how I/O (assuming you mean hard disk here) would be a major player as it is evident that the Pentiums scale nicely using the same storage sub-systems.

Your results are a bit odd because it is more common to see Athlons beat out their Intel counterparts in compilation tests. Not sure what to make of that other than according to your numbers, the Pentium D XE 840 is a beast in this case. Is this C/C++ code and, if so, what compiler were you using? I would dare venture a guess on Intel's compiler.

I apologize to the thread owner if my interest in these single benchmarks have detracted from the overall purpose of this thread.

 

[RussSchultz]

I am not sure how I/O (assuming you mean hard disk here) would be a major player as it is evident that the Pentiums scale nicely using the same storage sub-systems.

Compilation is very IO bound. Read lots of files in, compile, write file out. My understanding is the Intel SATA IO subsystem is noticeably better than the ones that frequent AMD chipsets (NVIDIA, ATI, VIA, etc.)

Is this C/C++ code and, if so, what compiler were you using?

A mix of C and ASM code. The compiler is an product not from Intel that is cross compiling for a microcontroller that is not an Intel product.

Your results are a bit odd because it is more common to see Athlons beat out their Intel counterparts in compilation tests.

I was a bit surprised myself. However, our single threaded test (which used an older compiler for our previous product), the AMD performed quite admirably--some 30% faster than the 840EE. (Which isn't too surprising, considering the 840EE is only 3.2Ghz, and the AMD does have a better pipeline, etc.)

Quite honestly, I think it came down to the number of virtual processors that made the difference. In this case, hyperthreading gives us about 15% increase.

 

[wireframe]

I am not sure how I/O (assuming you mean hard disk here) would be a major player as it is evident that the Pentiums scale nicely using the same storage sub-systems.

Compilation is very IO bound. Read lots of files in, compile, write file out. My understanding is the Intel SATA IO subsystem is noticeably better than the ones that frequent AMD chipsets (NVIDIA, ATI, VIA, etc.)

I realize this, but if you look at your results between the XE running 4 and 8 threads, you can see it squeezing out almost 10% more performance. That doesn't sound like something that is horribly IO bound. Would you know of any good tests regarding the Intel v World in terms of IO sub-system performance? I keep reading this, but I have never seen it conclusively proven.

PS. For some reason I thought SCSI when I read 10K RPM and assumed they were using the same controller card. I think I should have understood this as 10K RPM SATA aka WD Raptor, right?

Is this C/C++ code and, if so, what compiler were you using?

A mix of C and ASM code. The compiler is an product not from Intel that is cross compiling for a microcontroller that is not an Intel product.

Your results are a bit odd because it is more common to see Athlons beat out their Intel counterparts in compilation tests.

I was a bit surprised myself. However, our single threaded test (which used an older compiler for our previous product), the AMD performed quite admirably--some 30% faster than the 840EE. (Which isn't too surprising, considering the 840EE is only 3.2Ghz, and the AMD does have a better pipeline, etc.)

Is it possible that this "proprietary" (?) compiler was written with NetBurst in mind? When you say the older tool hands the Athlon the win it seems that perhaps great strides were made to optimize the new one particularly for NetBurst (nothing wrong with that, just postulating).

Quite honestly, I think it came down to the number of virtual processors that made the difference. In this case, hyperthreading gives us about 15% increase.

I think you are right on that. I keep reading that HyperTHreading is just a marketing trick, but in my experience it works and works very well at times.

 

[RussSchultz]

I realize this, but if you look at your results between the XE running 4 and 8 threads, you can see it squeezing out almost 10% more performance. That doesn't sound like something that is horribly IO bound.

I'm thinking the IO issue is helped by aggregating IO requests from multiple threads that can be satisfied simultaneously. (For example, many files include the same header files). This would help keep all the processing units fed with data. Or, while one thread is blocking, another will be processing, etc.

Considering there are only 4 virtual processors, adding more threads shouldn't eek anything out--unless it was IO bound. (Also, when we move to RAID1, we get an additional 15% increase, so that bolsters that idea).

 

[suryad]

This is not the first test I have seen where the multicore athlon gets spanked by the multicore pentium D. Something funny is going on. I think the fact that the Pentiums have HT should be giving them a massive advantage.

 

[RussSchultz]

This is not the first test I have seen where the multicore athlon gets spanked by the multicore pentium D. Something funny is going on. I think the fact that the Pentiums have HT should be giving them a massive advantage.

Hold on. I didn't test a multicore Athlon, so we can't make that conclusion.

 

[BRiT]

Must agree that a test with the X2 series would be very nice to see. The X2 has some nice tweaks to it, as it reports it can support HyperThreading. Even if it only translates to running single process thread on each core, it should help out significantly.

 

[RussSchultz]

I couldn't convince IT to look too closely at the athlon products because Dell doesn't sell them.

 

[Coola]

I apologize to the thread owner if my interest in these single benchmarks have detracted from the overall purpose of this thread.

heh no problem, i always find these advanced tech talk always interesting, and it's good to know what benchmarks decide which is better.