Intel Pentium 4 2.2GHz and 2.0GHz Northwood Processors Review
|
Old computers second life | |
Retro philosophy | |
Why? Una storia personale Retrocomputing e non solo Roadmap |
comparison with the predecessor and assessment of prospects
The problem of optimal selection of computer components, unfortunately, does not have a clear mathematical solution: there are too many factors. Usually, this method is used: a price list is taken, extreme positions are thrown out of it (total crap and too steep), and from the remaining ones, a product is selected with the largest increase in trademark recognition and quantitative swelling of pseudo-technical characteristics for price increase. The decision is not the worst, let’s not even pay attention to the fact that 256 MB of memory onboard the GeForce2 MX is a bit too much. Let’s think better about this: why do we need a new processor at all?
Indeed, why? No, there are, of course, tasks that even an N-GHz Pentium M will load in the highest registers, but they are terribly far from the people, and the dances of blue-orange men in your information display devices are not designed for their solvers. No, there is still a difference even when working in Word, it’s just harder and harder to measure it it’s not for nothing that each new generation of benchmarks gives the final result in points an order of magnitude lower, which means there is room to grow, and there are 10 users, 5 points strain much more than 1050.
Games Yes, there are still games, as always, but here everything is too neglected. Intel vainly demonstrates 3D demos at presentations, calculated and shown by the forces of the central processor alone. It is much more likely that some next GeForce, which has already practically integrated the sound, will at the same time calculate the physics in the game, and they will agree with the hard drive: they will organize a new protocol for direct access to graphics memory, and why then they will need a processor with system memory these remnants of the past, erroneously considered the engine of progress?
And the poor, useless processors will keep shrinking and accelerating until one (beautiful?) day they scatter a handful of photons. Well, to this day, however, is still far away. For now, let’s see what the new Pentium 4 will please us with — you look, and the light will shine
The new Pentium 4 «Northwood» core
To begin with, the hero himself and his predecessor:
Intel Pentium 4 «Northwood», 2.2 GHz , Socket 478
Intel Pentium 4 «Willamette», 2 GHz, Socket 478
Pay attention to the bottom line of the marking: the numbers «512» at the end just indicate the only architectural difference between the new processor and the old one — 512 KB L2 cache instead of 256 KB. However, the implementation of this difference became possible only due to the transition of Intel factories to 0.13-micron production technology using copper connections. As a result, a 200 mm wafer can hold twice as many dies as a 0.18 micron wafer. (In the future, with the transition to 300 mm wafers, this number will increase by another 2.5 times.) Well, the result of the new technical process will obviously be a higher permissible frequency of processors and lower heat dissipation (49.8W TDP for the 2.2GHz version of Northwood versus 75W for the 2GHz Willamette). It is curious to note that the processor on the new core contains 55 million transistors (the area is 146 mm 2 ), of which about 40% (the Willamette has 42 million transistors, and only the L2 volume has changed) falls on the cache not bad?!
Northwood 2.2 GHz running at 2 GHz
Willamette 2 GHz
478 were already designed for new processors, their current VRM supports such a direct one, so in the worst case, you will have to reflash the BIOS of the board. It remains to be mentioned that the processor with a frequency of 2 GHz will be the last in the Willamette line, while the Northwood line will start with a 2 GHz one, so to distinguish between extreme copies, Intel again has to resort to the “A” index: the younger Northwood will be officially called the Intel Pentium 4 2 A GHz.
Performance Study
Test Bench:
- Intel Pentium 4 «Willamette», 2 GHz, Socket 478
- Intel Pentium 4 «Northwood», 2.2 GHz, Socket 478
- Motherboard (Intel 850)
- 2×128 MB PC800 RDRAM RIMM Samsung
- ASUS 8200 GeForce3
- IBM IC35L040AVER07-0, 7200 rpm, 2 MB cache, 40 GB
- Windows 2000 Professional SP2
- NVIDIA Detonator v22.50 (Vsync=Off)
- RazorLame 1.1.4 + Lame v3.89 codec
- VirtualDub 1.4.7 + DivX v4.02 codec
- WinZip 8.0
- WinAce 2.11
- eTestingLabs Content Creation Winstone 2001
- eTestingLabs Business Winstone 2001
- BAPCo & MadOnion SYSmark 2001 Internet Content Creation
- BAPCo & MadOnion SYSmark 2001 Office Productivity
Secondly, the existing Willamette 2 GHz was bus overclocked to 2. 2 GHz (110 MHz x 20). This allows us to clarify the «academic» question, which is better: more cache or a faster bus, as well as to make some estimates for the future, because it’s no secret that soon the Pentium 4 bus frequency (though already Northwood) will increase to 133 MHz (533 MHz Quad -Pumped).
We also note that those who wish to compare the performance of the new processor market flagship from Intel with the current fastest model from AMD can compare the results of today’s testing with the results of comparing Willamette 2 GHz and Athlon XP 1900+. As soon as AMD announces its 2000+ model, we will definitely face Northwood in a fair fight.
Test results:
MP3 encoding with Lame, as we have seen many times, is not too sensitive to memory bandwidth. In this test, we can observe some (3-4%) gain from the increased cache size in Northwood and good (~10%) scalability of the test in terms of processor frequency. The effect of tire acceleration seems to be there, but the absolute difference in the readings is too small, let’s wait to draw conclusions.
Here, even before the actual results of testing the processor, the benefit from using the new version of the DivX codec, optimized for SSE2, attracts attention. 30-40%! Even comments are superfluous here!
As for the comparison of processors on the old and new cores, we can note several interesting points. Northwood gains 7-9% of the advantage due to a «clean» increase in frequency by 10%, but the advantage of the bus-accelerated Willamette is more significant: 10-15%. Northwood’s increased cache has some, not too significant, effect, and this effect is offset by an increase in the bus frequency (which overclocks, among other things, the L2 cache). By the way, pay attention to the fact that when SSE2 is enabled, the difference manifests itself more clearly in all cases: when the processor frequency is increased, when the bus is accelerated, and when the cache size is doubled.
The results of WinZip resemble the results of MP3 encoding both in absolute numbers and in the nature of the dependence. Again, we note that a difference of a couple of seconds does not make it possible to draw far-reaching conclusions.
And the results of WinAce, as expected, resemble the results of MPEG4 encoding both tests aggressively work with memory, but when archiving WinAce, the dictionary is frequently accessed, which provides a larger (up to 8%) gain, other things being equal to the processor on the Northwood core . Obviously, if the entire 4-MB dictionary could be pushed into the cache, the speedup would be quite significant. In terms of processor frequency, the test does not scale very well, here the speed of working with memory has a greater influence.
The picture in both “semi-synthetic” tests is generally similar, the only oddity is that at a frequency of 2.2 GHz, the “honest” Northwood outperforms the bus-accelerated Willamette more than at a frequency of 2 GHz under equal conditions.
The description of the scenes used for 3DStudio MAX can be found in our introduction to testing professional video cards, and the processor comparison result we are interested in now is as follows: regardless of the complexity of the scene calculation by the processor and the complexity of drawing the scene by the graphics accelerator, the nature of the calculations is the same and accelerates calculation of all scenes is the same. In numbers, it looks like this: by ~9% with an increase in the frequency of the processor, by ~ 3% due to a larger cache, the bus acceleration has almost no effect on the result. As you can see, this test, which we have repeatedly emphasized, is purely computational, and it scales exactly according to the processor frequency, slightly depending on the memory speed and other factors.
Things are more interesting with SPECviewperf, primarily due to the variety of subtest dependencies shown. AWadvs-04 does not bring any surprises, steadily resting the scene rendering speed on the speed of the video accelerator. But DX-06 It would seem nothing strange: system performance increases slightly when moving from 2 to 2.2 GHz; if at the same time the tire accelerates it grows more strongly. Only bad luck: there is a very noticeable negative effect from «extra» cache. Surprisingly, but true: increasing the cache size in this test significantly reduces the speed of the exchange with memory. At the moment we have no clear explanations, it remains only to state: if you have devoted your life to working with IBM Data Explorer Northwood is not for you.
The current testing allows us to identify the difference between these two tests, which usually show similar results dependencies. In DRV-07, as you can clearly see, the system almost does not get a performance boost from a simple increase in the processor frequency, while the bus frequency (most likely, the memory-AGP connection) has a decent effect on the result. But that 8% gain is nothing compared to the 30% gain of the 512K L2 Northwood over the 256K L2 Willamette. Let’s postpone consideration of this artifact for a while and turn our attention to Light-04. The picture is really similar to the previous one, only without Northwood’s mysterious take-off, and in its absence, the overclocked AGP bus of Willamette 2.2 GHz brings it to the lead in full accordance with the data of our past tests: the performance of the video accelerator first of all.
Yeah, and that takeoff again! Well, now, relying on the same experience, we can say with sufficient confidence: the reason is RDRAM. We have repeatedly noted that Expendable is a game with a rather «chaotic» algorithm, which «crashes» the long Pentium 4 pipeline and, as you can see, greatly strains the latency RDRAM with its «memory jumps». Northwood’s increased cache significantly smooths out this effect, and the result is obvious. However, it should be taken into account that if, for example, DDR SDRAM was used, this artifact would manifest itself less noticeably, so only Expendable fans (if there are still any left the test has been asking for a dump for a long time) and Rambus will be able to experience the magical acceleration. simultaneously. It is also worth noting that overclocking the bus in this test, unlike SPECviewperf, speeds up not the video accelerator (you can’t speed up something that doesn’t exist), but the processor-memory link.
All more or less modern toys demonstrate the same picture: Northwood has a 5% advantage due to the larger cache, which smooths out to 1-2% when the Willamette bus is overclocked. All this, of course, is true only for those tests and resolutions where the overall system performance depends on the processor.
Conclusions
Summing up, the first step is to determine the position from which Northwood will be evaluated. Fortunately, today we are not on the verge of some kind of revolutionary turn, which inevitably entails a change in the chipset and type of system memory. There is a smooth evolutionary process in which we are assigned only the role of an observer, and Intel will do the rest. So let’s relax and enjoy.
Fortunately, there are reasons to be satisfied. The new processor has only one advantage over the old one, but it really works and brings very good dividends. On one side of the scale, however, we have a penny superiority in encoding with DivX v4.02 and a negative gain in DX-06, but on the other hand, we have an improvement in work with RDRAM reaching up to 30%. On average, the scales show + 5-8%. And this is considering that in purely processor tasks there can be more (and why else do you take the top model?).
However, let me remind you once again that it is not very interesting or smart to directly compare the new and old cores: they will not intersect in the Pentium 4 line anyway (more precisely, they will intersect only once, and the comparative indicators are in front of your eyes), so there is no choice will not be. You can only put a notch in your memory: Intel did not disappoint, the new processor is no worse than the old one.
But what about the prospects? They also appear in a rosy light: in tests for the interaction of the processor and memory, a 10-15% gain is obtained for a 10% bus acceleration. We will not take into account the advantage in graphics-oriented applications, since the AGP and PCI frequencies of boards with a future 133 MHz processor will obviously be “normal”, but Pentium 4 and DDR266 synchronously operating at 133 MHz this is very enticing and very «efficient» as the Willamette 2.2 GHz would suggest.
It remains only to understand: why do you need a new processor?
90,000 Intel Pentium 4 Northwood in Kemerovo: 500-TOVARS: Free delivery, Discount-16% [Go]
Partnership Program of
Kemerovo
Catalog
Catalogs of 9000
Clothes and shoes
Building materials
Building materials
Textiles and leather
Textiles and leather
Health and beauty
Health and beauty
Children’s goods
Children’s products
Electrical engineering
Electrical engineering
products and drinks
products and drinks
House and garden
House and garden 9000
Agriculture
Agriculture
Industry
Industry
All Categories0003
Processor Intel Pentium 4 2. 26 GHZ HT 512KB 533 MHZ SL6RY Northwood MPGA-478 OEM, 2.26 GHz (533) OEM (without culer) Similar products
processor Intel Pentium 4 2, 4 GHZ HT 512KB 800 MHZ SL6WF Northwood (Kwema, 2,
07 4 907 4 907 4 908 400 )
DETAILSMore prices and similar products
Processor Intel Celeron ( Pentium 4 ) 2.0 GHZ 128KB 400 MHZ SL6VR Northwood MPGA-478 OEM, 2.0 GHz (400) OEM version (without cullar) Prices and similar goods
1111
335812-001 processor HP [ Intel ] Pentium IV HT 2400MHz (512/800/1.525V) SOCKET478 NORTHOOD 9000
9000
9000 9000
9000 9000 9000 9000 9000 9000 9000 9000 9000 9000 9000
207 Processor
Intel Celeron ( Pentium 4 ) 2, 4 GHZ 128KB 400 MHZ SL6VU Northwood MPGA-478 OEM, 2, 4 GGC (400) OM (400) OM (400)
Read more prices and similar goods
processor Intel Pentium 4 2. 4GHZ/512/800, Northwood OEM: intel 9001 9,000 9000 9,000 9000 9000 9000 9000 9000
208
4 ) 2.6 GHZ 128KB 400 MHZ SL6W5 Northwood MPGA-478 OEM, 2.6 GHz (400) OEM version (without cooler)
more details
In Pentium 4 2.4GHZ/512/800, Northwood OEM (without cooler) Manufacturer: Intel , line
more than price
processor interlt0208 4 — M 1.70 GHZ, 512k Cache, 400 MHZ FSB
In the price of prices and similar goods
processor SL6CH MOBILE Pentium 4 — M 1.70 GHZ, 512KA, 512K, 512KA CAC
In the store prices and similar goods
processor Intel SL6CG Mobile Pentium 4 — M 1.60 GHZ, 512K CACHE, 400 MHZ FSB
In the price of prices
9,000 9000 9000
003
In the price of prices and similar goods
processor Intel SL6CJ Mobile Pentium 4 — M 1.
ASUS 060 CD-ROM0002 Software: