www.digit-life.com/articles2/cpu/intel-conroe-2-13-ghz.html
" (Instead of Introduction) Rumours about wondrous performance of the new desktop processors from Intel have been circulating for a long time already. From time to time they were confirmed by various express tests. But the structure of those tests and their numbers did not allow to evaluate how cool the new processors were in any article we had read. So when you browse the results, you should take into account the following: Core 2 Duo E6400 is represented by an engineering sample, which characteristics may differ a little from the release model, and works with an i965-based motherboard. Core 2 Duo E6700 is represented by a release model and works with an i975X-based motherboard. Of course, we could have waited until we got all production-line samples and published an "ideal" article. But as nobody offered us new Intel processors on Conroe core, we decided not to be so fastidious. Quick-look article seemed more important than conceptual completeness.
What concerns the latter, it has been noted many times that they are direct descendants of the 6-gen core of Intel processors (P6), which had been designed prior to Pentium Pro and had lived to Pentium III. Thus, on one hand, Core 2 is a step back, as it sort of cancels many architectural solutions that appeared in processor cores of the later generations (the famous Intel NetBurst architecture of the seventh generation). On the other hand, we cannot say that Core 2 is the direct heir to Intel P6 because of too many modifications. The key differences of the new architecture from the closest parent -- Intel Core Duo (Yonah) come down to the following: * Improved decoder, expanded to 4 decoders of x86 macro-ops (3 in the closest counterparts, Intel Pentium M / Core Duo). TDP indexes "T", "L", and "U" are reserved for mobile and Ultra Low Voltage processors of the Core 2 series. We have also heard that the 4xx0 and 6xx0 series will include desktop processors, while the 5xx0 and 7xx0 series -- mobile models. Those microarchitectural details, we could find out with our tests of the engineering sample, can be found in the traditional detailed analysis of the platform in RightMark Memory Analyzer, written by Dmitri Besedin. And the official detailed description of the microarchitecture of the new processor will hardly be available before the official announcement.
" means "per each core" ** -- in AMD processors it's frequency of the memory controller bus *** -- it's measured differently in Intel and AMD processors, so a direct comparison is not correct **** -- engineering sample, its characteristics may differ from a production-line sample Software 1 Windows XP Professional x64 Edition SP1.
Of course, Core 2 Duo E6400 is not as impressive as E6700 -- but don't forget that its clock is lower by 25%. The E6700 tears all its opponents to pieces, having outperformed its nearest competitor (AMD Athlon 64 FX-62) by 27% . And the FX series from AMD has been traditionally opposed to the XE series from Intel. The only extreme model on Conroe core is not represented here, the E6700 is just a top CPU from a usual desktop series!
On one hand, we can congratulate Intel: its new 213 GHz processor outperforms the old top 373 GHz model. On the other hand, only Core 2 Duo E6700 can stand up to AMD Athlon 64 FX-60/62. So, CAD/CAE packages are not the most favourite applications of the new core so far. The X6800 will be better, but in this situation AMD can at least hope for catching up with its main competitor, the gap is not too large.
Out of doubt, Intel has made a great progress, compared to NetBurst CPU results. Compared to top Athlon 64 FX processors, the new core also looks good. But AMD can evidently try to wrestle, it has some chance to catch up with Intel (most likely after the upgrade to the 65-nm process technology)...
The result is not very reassuring, but let's not forget that CPU RightMark is well optimized for all known CPUs. Core 2 Duo does not belong to this group yet, of course.
Quite a rare case: having demonstrated an excellent result versus the opposing top models, the junior Core 2 Duo... is outperformed by the top Pentium eXtreme Edition based on the NetBurst architecture, everyone already gave up hope in! An excellent reminder to all those people who criticized Pentium 4: yep, this processor is not good for all tasks, but it does not mean that is was bad for all tasks. However, results of the E6700 put everything into place: the Intel platform leads in total score (better by 20% than the top processor from AMD) It's will be much more difficult to close this gap. In order just to catch up with Core 2 Duo E6700, AMD will have to overclock Athlon 64 FX to about 34 GHz...
The junior Conroe is outperformed by Athlon 64 FX-62 by 5%. But if we take into account that its clock is lower by 31% and have a look at the result of the E6700, it becomes clear that AMD K8 has no chances to catch up with the new processors from Intel in this test, just like in the previous one.
Judging by what we know about this class of programs and its preferences, the laurel is shared by the Conroe-based processor and Intel chipsets, their memory controllers, to be more exact.
But I repeat that even the E6400 demonstrates a good result for its clock frequency (to say nothing of the E6700, its column is too long to miss :). To all appearances, the situation will be as follows in such cases: top representatives of the Intel Core 2 line will outscore top processors from the main competitor. But AMD will certainly try to close the gap after some time (about six months). In our opinion, the old K8 architecture can still kick, if the gap does not exceed 15% (as in this case).
This non-standard (for our test procedure) diagram illustrates an evident parameter: efficiency per frequency. These values are just a total score divided by the clock of corresponding processors (in GHz). Here you can see the quintessence of Conroe's triumphal hymn. While the efficiency of Athlon 64 FX is higher by 58% versus Pentium eXtreme Edition 965, Conroe's efficiency is higher by 23% versus Athlon 64 FX! It seemed impossible to create something more efficient than the AMD K8 core, but Intel engineers managed to do it. Conclusion On the whole, we can say that the new processors from Intel turned out even more impressive than we expected. Looking at the results, we don't want to sing the praises, congratulate Intel with the victory, laud the new core, etc. We'd rather look through the diagrams once again to understand what the world giant of CPU building has unleashed. Conroe and its successors are here for a long time, that's obvious. The efficiency per GHz it demonstrated is not just admirable -- there is the reverse of the medal: such impressive results are not obtained easily. So we cannot expect anything cardinally new for at least three or four years. We'll see Conroes with increased clocks and probably with larger caches. What can we say about the architecture looking at the test results? Conroe offers relatively low results in 3D modeling and CAD/CAE packages. Of course, its results are still excellent for its clocks. So we partially confirmed the hypothesis that the classical FPU in the new core... no, it's not "bad", but it's less perfect than the other units. The manufacturer will probably make up for this problem with higher clocks. The situation in 3D modeling and bitmap graphics can be improved by increasing the number of cores -- these algorithms can be easily distributed between multiple cores. Excellent results are demonstrated by the new core in data archiving, web-server test, OCR, and games. Having analyzed this collection of tests, we can establish a fact that Intel did well for memory and cache operations and well as for "chaotic code" that used to stupefy the long-piped Pentium 4 In fact, the company defeated AMD on its own field with its favourite weapon. Proceeding from the above-said, I'll risk a global conclusion: from the microarchitectural point of view, the new core will be much simpler for programmers than the previous generation (Pentium 4 / NetBurst). You shouldn't worry much about optimizations (except for distrib...
|
http://csua.com/feed/