We have not yet touched on the new Celeron based on Conroe-L core in any of our articles, although the subject has already become ripe for discussion. In this article the above-said shortcoming is going to be corrected. We shall acquaint you with performance characteristics of the new low-end Celeron models (which seem like a logical starting point). Unfortunately, none of the AMD Semprons have been examined using our new testing method yet (which is an oversight that we shall soon take care of). Therefore, it was somewhat difficult to choose worthy opponents in the same category. That is why we decided to try an unusual approach, which we thought was rather peculiar. We are not going to examine how Intel's new low-end model measures up against the modern low-end CPUs from its main competitor. Instead, we shall compare it to quite old processors that have been nowadays driven from their medium price-range position into a low-cost section of the market. In our view, such a comparison would be rather educating.
Hardware and Software
* — "2 x ..." means per core
A Necessary Foreword to the Diagrams
We present the results of our tests in two unique ways due to our testing method. First of all, in the summary diagrams of this article all data types are reduced to one – relative integer points. What we mean by relative integer points is that the performance of all processors is relative to that of one processor, which in this case is the Intel Core 2 Duo E4300. The performance of the Core 2 Duo E4300 is rated at 100 points. Second, detailed results regarding the tests are available in a Microsoft Excel table. In the article itself only summary diagrams are given which are grouped together by benchmark classes. Nevertheless, when required, we will occasionally draw your attention to the expanded results which are located in the Excel file.
3D Modeling Suites
It can be easily noticed that the old AMD models are only represented by Athlon 64 X2 3800+ for Socket 939. Alas, it currently is the least powerful AMD processor of the ones we have already tested. That is the sole reason why it was the only one chosen. As you may see, this fact didn't keep it from turning out to be the fastest one in the test. :)
Overall the distribution is quite irregular. The least of the Celerons (420) has even lost to Pentium 4 521, which has a single core too and not such a high clock rate (for a NetBurst architecture). However, let us not forget that, in contrast to the new Celeron, Pentium 4 can operate like a double-core processor due to Hyper-Threading.
It is quite interesting to compare Celeron 420 and Pentium E2140. Being equal in core and bus frequencies, the latter has exactly twice the L2-cache and the number of cores. Conventionally speaking (simplistically!), we can think of Pentium E2140 as a dual system using two Celeron 420s (although in reality this is not likely to ever become possible). We do not observe a doubling of performance, though the gain is significant: 36%.
Architectural preferences of the suites being used become apparent in these test results. The second core seems to be completely ignored (there is no other reasonable explanation). Particularly indicative are the better results of Celeron 440 in comparison to Athlon 64 X2 3800+. Both processors have equal clock rates. If we assume that the second core was indeed not used, then the cache size, for the purposes of this test, is the same (3800+ has 512 KB per core). The processors are practically in a tie, although technically Celeron is the winner.
Digital Photo Processing
The results are a consequence of Adobe Photoshop being optimized for a multi-core architecture as well as... NetBurst. It is not hard to notice that none of the single-core processors have won over any of the dual-core processors in this test (including Pentium D 805, which is not at all powerful judging from many other tests).
As a whole, we again notice that software is optimized for multi-core processors. Only the Celeron 440 has barely managed to outperform the least powerful dual-core Intel processor based on outdated architecture.
Multi-core AMD processors are traditional favorites in this test. The significant lead of Athlon 64 X2 3800+ follows the pattern observed earlier. New single-core models (Celeron) have turned out to be faster than the old Pentium 4 521. Nevertheless, even the least powerful of the dual-core processors was too much for them to beat. Pentium E2140 has outperformed Celeron 420 by 38% - recall that we have already seen a similar value for this pair somewhere above.
Clock rate and dual-core architecture rule the ball in this test, apparently in equal proportions.
A very interesting test. Considering that all tested processors were roughly equal in bandwidth of the CPU bus and memory speed, and the fact that data compression software is extra sensitive to this particular parameter, the results should have been quite close. (The only exception is Pentium D 805 with the FSB frequency of only 533 MHz QP.) It is exactly what has happened here. If we take a look at the details, we see that Pentium D 805 has ended up almost tying Celeron 430. Moreover, such is the case in both subtests despite the fact that one benchmark application (7-Zip) supports multi-core processors, while Celeron 430 doesn't have a second core. Probably, Celeron was able to compensate for the lack of the extra core with a higher bandwidth of the CPU bus. Whereas Pentium D 915 has most likely won due to its L2 cache size, which, after all, amounts to a total of 4 megabytes...
The fact that Athlon 64 X2 3800+ has scored low is an indirect confirmation of a long-speculated hypothesis that ABBYY FineReader's code is just not working well with AMD's processors. Most likely, it cannot identify that they have certain sets of additional instructions. Pentium E2140 has produced shocking results. Is it possible that having two cores can actually hinder the batch recognition process?! Very strange software, indeed. However, it has become popular in its application area, so we just have to live with it...
An old test that doesn't support multi-core functionality in any of its subtests. The test is traditionally slightly biased against the NetBurst architecture and, as it turns out, still favors Conroe core even in a most cut-down version over the K8.
Once again, we observe the situation where dual-core CPUs are way ahead of the single-core ones regardless of architecture.
Comparing Celeron 420 vs. Pentium E2140 we see a clear example of how modern games support dual-core design, and support it quite well.
The final diagrams have come out to be much more interesting than we could have expected in the beginning. First of all, the overall score in the professional-use software category for Pentium D 805, which was originally included in the testing "just for fun" (intended as a "dummy"), is higher than what all three of the low-end Celerons based on a much newer core have produced. Of course, one may note that there are also Celerons 530/540/550. On the other hand, they are not as readily available. They exist as parts of complete systems, while Pentium D 805 is far from being the only dual-core model in the series based on NetBurst architecture.
Secondly, this is one of the few articles where the overall scores for professional and household/amateur software differ so greatly. In the "amateur category" Celeron 440 has almost tied the old Athlon 64 X2 3800+ and has just barely lost to a full-fledged dual-core Pentium E2140! At that, this Celeron costs significantly less.
In regards to the common overall score, due to the aforementioned bias differences between professional and amateur/household software the chart looks like a "mess". Nevertheless, fourth place for such a low-end CPU as Celeron 440 doesn't look so bad.
Supposed Power Consumption
We do not see anything surprising on these diagrams. Of course, if we cut a half off of Conroe core, which by itself is low-consuming, then it would consume even less. However, this reduction, as it turns out, is not great...
As a whole, we think that the idea of comparing the old middle-end with the new low-end models has proven to be quite fruitful. The results are interesting and have quite tangible practical value for the owners of systems based on Pentium 4 and Pentium D. The test results are also of some interest to the owners of Athlon 64 X2-based systems, though limited due to low coverage. Besides that, from a pure professional standpoint, as testers we can not help noticing how in some cases the results were quite unexpected.
Despite all that, we found nothing sensational. Celeron 4xx is not going to become the "killer" of low-end Athlon 64 X2 models. It is clear from looking at the results of the X2 3800+ alone. Neither will the new Celeron be a rival for Pentium E2xxx. Too many of the modern applications are capable of taking advantage of the second core, which can become a decisive argument. As to the competition between Celeron 4xx and Pentium D, indeed it is taking place. However, there is no certainty in it. Too much depends on specific software bias.
Yet, let us not get carried away, shall we? After all we are considering a low-end CPU! In its class the new Celeron is a very successful processor, without any doubt. We can make such a claim even now without any comparison to Sempron. Simply do not expect of it anything more than what it is worth. :)
Testbed memory modules provided by
Corsair Memory Russia
Stanislav Garmatyuk (email@example.com)
November 19, 2007
Write a comment below. No registration needed!