iXBT Labs - Computer Hardware in Detail

Platform

Video

Multimedia

Mobile

Other

Intel Pentium XE 955: Ultima Ratio of the Kings

December 27, 2005






It's quite obvious that Intel processors on Prescott core (and its ideological descendants) have problems with the clock growth. The assumption that dual-core processors will have at least the same problems or even worse looks logical: two cores get heated much more than one. But despite this situation, today Intel has announced a new processor in the extreme series (Intel Pentium eXtreme Edition 955) with 3.46 GHz clock. Is it high or low? On the one hand, it's not too high: just 260 MHz higher than in the previous Pentium eXtreme Edition 840. On the other hand, considering that power consumption and heating of the PXE 840 already disturbed computer users more than usual — even this growth is respectable.

But the problem is not only in the clock. Pentium XE 955 features practically ALL latest advances of this manufacturer: except for the clock, PXE 955 has a 1066-MHz bus as well as 2 (2!) MB L2 Cache per each (!) core. In fact, the resulting monster is sort of an extremum of NetBurst architecture in general. It's hard to figure out further possible roads of its development: well, at maximum, the clock will be raised by another 266 MHz, getting maximum close to the clock peak of single-core CPUs based on the same core. That is, considering the accented removal of Intel 4GHz CPU from roadmaps, Pentium 4 eXtreme Edition 955 will have only one heir even theoretically. If it will at all. Thus, today we either witness the performance peak, which NetBurst can provide in its perfect dual-core modification, or it's the very pre-peak. We shall carry out our today's tests relying on this assumption.

Bare facts

The "new" Presler core is a "double" variant of 90nm Prescott-2M core adapted for the 65nm process. In other words, this dual-core CPU actually hosts two fully-fledged single-core processors (like Intel's previous dual-core products). In case you were wondering, it looks like this:

Unlike Smithfield, here we have 2MB of L2 cache per core, which is natural, since the crystal size is reduced comparing to Smithfield due to the transition from 90nm to 65nm process. The extent of this reduction is still questionable, because we haven't found the exact crystal size in Intel's documentation available for us at the moment. But we do know the amount of transistors - 376 million, TDP (note: this is NOT peak energy consumption, but, more likely, a variation of typical) - 130W, and maximal (according to documentation) typical operating temperature - 68.6°C. As you can see, despite the transition to 65nm, the TDP of the new eXtreme Edition remained the same. It would be logical to assume that all 65nm process bonuses were consumed by increased core clock and FSB as well as additional L2 cache. According to specs, the minimal voltage of PXE 955 is 1.2V, while the maximal is 1.3375V.

Intel Pentium eXtreme Edition 955 features the following technologies:

  • Dual-core
  • Hyper-Threading
  • Intel Virtualization Technology
  • 1066MHz FSB
  • Execute Disable Bit
  • Extended Memory 64 Technology (EM64T)

All of them, excluding the 3rd, are already well-described. And Intel Virtualization Technology will be examined a bit later in a separate article.

Testing

Testbed configurations

  • Processors
    • Intel Pentium eXtreme Edition 955 (2 x 2 MB L2, 1066 MHz FSB, 2 x 3.46 GHz core)
    • AMD Athlon 64 FX-57 (1 MB L2, 2.8 GHz core)
    • AMD Athlon 64 FX-55 (1 MB L2, 2.6 GHz core)
    • AMD Athlon 64 X2 4800+ (2 x 1 MB L2, 2 x 2.4 GHz core)
    • Intel Pentium eXtreme Edition 840 (2 x 1 MB L2, 800 MHz FSB, 2 x 3.2 GHz core)
    • Intel Pentium 4 670 (2 MB L2, 800 MHz FSB, 3.8 GHz core)
    • Intel Pentium 4 eXtreme Edition 3.73 GHz (2 MB L2, 1066 MHz FSB)

  • Motherboards
    • Intel D975XBX Desktop Board (Intel 975X chipset)
    • Intel D955XBK Desktop Board (Intel 955X chipset)
    • ASUS A8N SLI Deluxe (NVIDIA nForce4 chipset)

  • Memory
    • 2x512 MB PC5400 (DDR2-533) DIMM 3-3-3-8 (Corsair)
    • 2x512 MB PC3200 (DDR-400) DIMM 2-2-2-5 (Corsair)

  • Video card: ATI Radeon X800 (256 MB)
  • HDD: Samsung SP1614C (SATA), 7200 rpm, 8 MB Cache
  • AC power adapter: FSP 550-60PLN (500-550W)
  • Windows XP Professional SP2, DirectX 9.0c
  • ATI CATALYST 5.4 (Display Driver 6.14.10.6525)

Test results

Diagrams with all test results (64 all in all) are published on a separate web page without comments, just as is. The article provides only summary diagrams that calculate the results of entire test groups into average scores. This approach appeases curiosity of the most inquisitive readers, who are against cutting down the number of test results published in our articles, and still makes the article less motley and graphics-intense.

SPECapc for 3ds max 6 + 3ds max 7.0

A complete set of diagrams

If we have a look at the detailed diagrams, we can see that Pentium XE 955 still manages to outperform the top dual-core processor from AMD (Athlon 64 X2 4800+) in the rendering test. But the advantage in the rendering test is not very high, its lag in the Interactive test is much higher. As a result, the dual-core processor from AMD still leads in the average score. But compared to Pentium XE 840, the 955 model managed to bring down the score considerably.

Maya 6.5

A complete set of diagrams

Dual-core processors do not look very good in SPEC for Maya (here is the reason: in this case, SPEC does not include the rendering performance test into its benchmark). But the increased clock (probably together with the high-performance bus) allowed PXE 955 to outscore its direct competitor nominally (Athlon 64 X2 4800+). However, Pentium XE 955 is outscored by the flagship dual-core from AMD in our rendering test just as nominally. On the whole we can say that both competitors are on a par,

Lightwave 8.2, rendering

Besides, Pentium XE 955 closed down the gap from A64 X2 4800+ to such a score that the processors look similar in performance. In case of Athlon 64 X2 4800+ vs. Pentium XE 840 comparison, we couldn't really say that...

SPECapc for SolidWorks 2003

A complete set of diagrams

The situation here is still bad. To all appearances, we witness the established "architectural preferences" of a given package. There is no point for Intel to fight with AMD in this case. Just think about it: if we use the simplest extrapolation formula, we'll get that in order to at least catch up with the Athlon 64 X2 4800+, a processor similar to Pentium XE 955 in architecture should have the core clock of... 3.8 GHz. That is to have the same clock as the top single-core processors. That's a weak hope... It's much easier to persuade SolidWorks to optimize its code for NetBurst architecture :).

Adobe Photoshop CS (8)

A complete set of diagrams

Adobe Photoshop is one of those programs, where even Pentium XE 840 looks quite decent compared to dual cores from AMD. The 955 model is certainly still faster. To all appearances, we should thank Adobe: I doubt very much that such a negation of the overall pattern (let's not ignore the obvious: AMD processors are better according to this pattern) is accidental. Adobe programmers probably studied well various Optimization Manuals for Pentium 4. On the other hand, it gives us food for thought: Intel's architecture may be not as bad as it seems, if you really put some effort into it. It just takes some skill to program for it and use all its features...

Adobe Acrobat 6.0

In this case developers obviously paid much less attention to optimizations for Intel: all processors from this company are steadily outperformed by AMD processors.

All-purpose data compression (archiving)

A complete set of diagrams

If we have a look at the detailed diagrams, we can see that even honest dual cores do not help Intel processors to get an advantage in the only archiver supporting multithreaded operations — 7-zip. But for objectivity's sake, we should mention that "multi" means not more than "dual" in this case.

Multimedia lossy compression (MP3/MPEG2-4)

A complete set of diagrams

One not quite correct (post factum) peculiarity of our method of testing multimedia lossy compression is hopefully known to our readers: Intel processors got a great advantage due to a single test: LAME with maximum quality settings. But we'd like to warn our readers from an excessively straightforward approach to the total score: if we have a look at the detailed diagrams, we'll see that Pentium XE 955 results in Canopus ProCoder 2 are quite good and reach the results of Athlon 64 X2 4800+. So, on the whole we can establish a fact that these processors are approximately on a par even in the results of other tests, not only due to LAME Encoder.

CPU RightMark 2004B

A complete set of diagrams

Four-threaded rendering (dual cores + Hyper-Threading) is a closing argument for the new Intel processor. On the one hand — obvious and expected, on the other hand — an absolute advantage is an absolute advantage, we cannot get away from it...

3D games and graphics visualization
in professional packages

DOOM 3

Far Cry

Painkiller

Unreal Tournament 2004

Total score in games

A complete set of diagrams

SPEC viewperf

A complete set of diagrams

What concerns 3D visualization, Intel processors just cannot win here no matter at what cost. No matter how high the clock is — they just cannot win. As all the four games and the benchmark for professional graphics packages stick to the same opinion of this architecture, we should draw a conclusion that the trouble is in the general architectural principles, it can't be helped...

Shy germs of multicore support

After the "mandatory program" we decided to test (purely experimentally) how well multi-core processors are supported by those programs that so pompously claim this support. We've taken two applications for this purpose: Quake 4 with the latest Patch 1.0.5 beta and DivX 6. As our main task was not to compare processors between each other, but just to find out whether multi-cores are of any help here, we limited ourselves just to a single dual-core processor — the main hero of this review, Intel Pentium eXtreme Edition 955.

Quake 4

We benchmarked it with setting autodetected by the game itself: High Quality, 800x600 @ 32-bit. As we know, Quake 4 incorporates a benchmark for measuring performance, but it does not contain built-in demos — you have to record them on your own. At first we took the demo, kindly provided by the editor of our video section Andrey Vorobiev, and decided that SMP support in Quake 4 is just a canard: this support enabled or disabled, the result was the same (the difference of 0.2 fps is obviously within the inaccuracy of measurements).

But after we read materials on other web sites, which used the demo available on the THG server, we decided to run it as well for conscience sake. Imagine our surprise when SMP support suddenly snapped into action!

Here are the facts: as no game settings are stored in a demo file, we can draw a conclusion that there exist certain scenes, where SMP support helps to raise fps, and there are scenes, where we cannot expect any gain. Here we enter the domain of assumptions (quite logical though, judging from what we know about game engines): judging from the absolute fps values, THG demo is much easier for a video card. That's probably why the SMP effect is noticeable. Our demo loads a video card much heavier, so the speed of rendering a scene by a processor (processors) has nearly no effect on the resulting fps. If our assumptions are true, we get a classic dilemma, even for single-core processors: a fast processor for games is a good thing, of course; but a fast video card may provide a much more noticeable effect.

DivX 6 Pro

Of course, we didn't find a way to disable one core. So "1 core" on the diagrams does not mean that a core is physically disabled, but that affinity mask is used (any process in Windows XP can be assigned what processors to use). The result is evident: DivX 6 really supports SMP, but in comparison with Quake 4 (in case of THG demo) its effect is much less pronounced. We can also see clearly that virtual multiprocessing (single core + Hyper Threading) provides a much lower effect than sterling dual cores. Enabling HT on two physical cores yields nothing at all. That's quite expectable, though: load distribution among processors demonstrates that using all dual-core capacities in DivX 6 encoding is out of the question:




In this case, full load on all CPUs would give 100% CPU Usage (according to Windows Task Manager). Load on a single processor only would give 25% CPU Usage (as there are four processors all in all). As you can see, the real value is 43%. That is a tad higher than 100% load on a single CPU, but even lower than 100% on two processors. CPU 2-4 graphs also indicate that they do something on the background of the main processor — but not very actively...

Conclusion

Here is our verdict: Intel nearly managed to catch up with the dual-core processor from AMD. At the cost of a hugely enlarged chip (1066-MHz bus + 2 MB Cache per each core + increased frequency) — but it still managed to do it. Indeed, Intel Pentium eXtreme Edition 955 is currently approximately on a par (in total score) with AMD Athlon 64 X2 4800+ in performance. Of course, it's a tad slower (again in the total score in all applications) — but the approximate parity is still there. Let's put it like this: in the majority of cases the difference is so small that you will hardly notice it. Can we feel optimistic about this fact?.. We cannot really say. It's clear that both manufacturers have reached maximum clocks, they practically cannot be increased by both sides. Thus, a purely formal leadership will be granted to a manufacturer who will manage to overclock its processor a tad higher than the competitor... but it will not change the general situation. OK, Company X has ONE processor, which is faster than the top processor from Company Y. So what? What about the other models?

Pricing acts the most important role in such a situation. And here we can see another confrontation appearing: AMD, stressing relatively (compared to Intel, of course) high performance of its dual-core solutions, is not in a hurry to drop prices. Intel admits the relatively low performance of its dual-core processors — but it offers them at much more democratic prices. We've got a strange parity — those who need top performance buy AMD processors (but the results of today's tests show that Intel can also make an offer to such people now). On the other hand, those who stake on dual cores, but don't have enough financial resources, have to buy Intel processors, as AMD does not offer anything in this price sector.

Mass vs. technical attractiveness? Looking back to the history of computer electronics, we can just establish a fact that the outcome of such confrontations was often not so obvious as we could expect...



Stanislav Garmatiuk (nawhi@ixbt.com)
December 27, 2005.


Write a comment below. No registration needed!


Article navigation:



blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook


[an error occurred while processing this directive]

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.