Another Gift to Overclockers: AMD Sempron 3100+ on E Core Revision

by Ilya Gavrichenkov
04/04/2005 | 06:06 PM

Together with the Athlon 64 processor family, which is getting stronger and stronger in the processor market, AMD continues the successful development of its budget Sempron solutions as well. AMD is striving for unified processor architecture that is why they are going to discontinue all processors on the older K7 architecture and focus their efforts on the CPUs with the next generation K8 cores. We have already mentioned a few times that Sempron processors for Socket A form-factor are about to disappear from the market completely, and the entire Sempron family should be moved to processor cores on K8 architecture and Socket 754 form-factor. Now we are ready to say even more: the new 90nm “E” core revision, which has already successfully entered the market in the Athlon 64 processors, will soon come to the Sempron family, too.

On April 4 AMD starts shipping Sempron processors designed for Socket 754 and based on the new E core stepping. And it means that Athlon 64 is going to be not the only one who acquires SSE3 instructions support: from now on AMD’s budget processor will also boast this feature as well as a few other innovations, which we have already discussed in detail in our article called AMD Athlon 64 3800+ CPU: E3 Processor Core aka Venice at the Door.

I would like to stress that the budget processors, just like their high-performance counterparts are very popular in the market today. The thing is that they very often boast excellent frequency potential despite their comparatively low price. So they become a great choice for better value but still fast overclocker’s systems. That is why we were so excited to take a closer look at the new Sempron processors based on the E core stepping. As we have just seen from our detailed review of Athlon 64 3800+, this processor core can provide Sempron CPUs with very important advantage, namely higher frequency potential and hence more efficient overclocking results.

Of course, we couldn’t disregard the changes that happened to the Sempron processor family and when we were lucky to get our hands on the new Sempron 3100+ for Socket 754, we sat down to testing it.

So, let’s find out what benefits the new processors core brought to economical overclockers and hardware enthusiasts.

AMD Sempron 3100+ Version 3: Closer Look

This is what the new Sempron 3100+ with the E core stepping looks like:

The last two letters of the marking, “BO”, indicate that this is a new core, from the same family as the just reviewed Venice. As you remember, the same processors based on the D core revision were marked with “BA” letters at the end. The rest of the marking abbreviation states that this CPU features 256KB L2 cache, 1.4V Vcore and bears the maximum processor case temperature of 69^oC. In other words, everything remained the same as by the older Sempron processors based on CG and D core revisions.

And this is what the CPU-Z utility says about the newcomer:

As we see, CPU-Z doesn’t know yet about the new Sempron processor core that is why the information it reads about the core is absolutely wrong. It apparently mistakes the core of our Sempron 3100+ CPU for Paris. All other data are correct. In particular, it indicates correctly the 1.8GHz clock frequency of the processor, 256KB L2 cache and recognizes SSE3 instructions support.

Since the CPU-Z didn’t recognize the CPU core correctly we decided to resort to one more diagnostic utility called CrystalCPUID. Luckily, this tool did recognize the Sempron 3100+ on the E core revision the right way.

It is true that the CPU we are reviewing today is based on the core codenamed Palermo. AMD selected this particular name for the E core stepping, which is a modification of the new Venice and San Diego cores used in Athlon 64 processors. Here I would like to point at some confusion with the codenames for the desktop Sempron processors designed for Socket 754 form-factor. The first CG core designed for these processors and manufactured with 130nm production technology was called Paris. According to AMD’s official roadmap, the next core was supposed to be Palermo, and the finer 90nm production technology was intended to be its major distinguishing feature from the predecessor. However, AMD’s roadmap doesn’t distinguish between the D and E core revisions. At the same time, these two core revisions are very different from one another, even though they are both manufactured with 90nm process. E core revision supports SSE3 instructions and boasts slightly upgraded memory controller compared with the memory controller used in D core revision. Nevertheless, the roadmap calls both cores Palermo, despite the above mentioned significant differences.

Therefore, you should understand that even though Sempron processors on Palermo core have already been around for quite a while (you can check out our review of the Sempron 2600+ CPU dating back to March 4, 2005), the Sempron 3100+ CPU we are going to talk about today is a different product. Our today’s processor is based on the E core stepping (Venice analog), while the CPUs that have been selling before April 4, 2005 are all based on D core stepping (Winchester analog).

So, if you are going to shop for some Sempron processors, you should be very careful and pay special attention to their marking, because only CPUs on the E core revision with SDAXXXXAIO3BO and SDAXXXXAIO2BO OPN will offer you SSE3 support and other nice advantages.

But let’s return to our story about the main features of the new Sempron 3100+ based on the E core revision. Like its predecessors, it supports Cool’n’Quiet technology. In other words, when the processor shifts to power-saving mode, its working frequency drops down to 1GHz and its Vcore – to 1.1V.

As for the AMD64 technology support, Sempron processors based on the E revision of Palermo core do not have it, either. I have to stress that this is a very disappointing drawback of all budget CPUs from AMD, because the 64-bit version of the Windows XP operating system is about to be released this month already (for more details about this OS read our article called Windows XP Professional x64 Edition Preview: AMD64 and Intel Extended Memory 64 Technology). Besides, the competing Intel Celeron D processor family is going to acquire EM64T 64-bit extensions pretty soon. However, it is still too early to expect that 64-bit operating system will enrich the processor functionality or increase their performance. There is hardly any software, which would work in the native 64-bit mode. So, the absence of AMD64 technology support in AMD Sempron processors is very unlikely to upset any PC users within at least the next year.

Testbed and Methods

Since we have already studied the performance of Socket 754 Sempron processors against the competitor’s solutions in our previous reviews (take a look at the article called AMD Sempron 2600+ for Socket 754: K8 Architecture Made More Affordable, for instance), we will refrain from repeating all those results today. We will focus on the performance analysis of the new Sempron 3100+ based on E revision of 90nm Palermo core against the older Sempron 3100+ based on the 130nm Paris core. This will allow us to evaluate the effect from the enhanced memory controller of the new processor core, as well as the performance improvement resulting from the SSE3 instructions support.

For our performance tests we assembled the following testbed:

• CPUs:
  • AMD Sempron 3100+ (Socket 754, 1.8GHz, 256KB L2, Palermo);
  • AMD Sempron 3100+ (Socket 754, 1.8GHz, 256KB L2, Paris).
• Mainboard: DFI LANPARTY UT nF3 250GB (Socket 754, NVIDIA nForce3 250GB).
• Memory: 1024MB DDR400 SDRAM (Corsair CMX512-3200XLPRO, 2 x 512MB, 2-2-2-10).
• Graphics card: PowerColor RADEON X800 XT (AGP 8x, 500MHz/500MHz).
• HDD: Western Digital Raptor WD740GD (SATA150).
• OS: Microsoft Windows XP SP2.

Performance

Here are the benchmark results:

The obtained results are pretty logical. The new Sempron 3100+ is not that much faster than its older analog. In fact, we didn’t expect anything else, because otherwise, AMD would have simply increased the processor rating of the already existing CPUs on Palermo core.

The average advantage of the E core revision over the CG core revision doesn’t go beyond 1-2%. This is exactly the performance improvement to expect from the enhancements made to the new processor memory controller. Although, in some cases the performance advantage could be even bigger than that. For example in games, the performance of the CPU based on the new Palermo core could grow up by 5%. It looks like AMD intended all memory controller optimization to first of all affect the gaming performance of its CPUs, which has always been a trump of K8 architecture.

Note that against the background of the performance gain provided by the memory controller enhancements, the performance improvement cause by the SSE3 instructions support is almost null. In fact, the new command support in the Sempron 3100+ processor remains practically unnoticed: benchmarks using SSE3 do not stand out in any way among all other tests. Nevertheless, note that these are the benchmarks that do use SSE3 instructions: SuperPi test, Doom III game, Mainconcept MPEG Encoder and DivX codecs, and ScienceMark benchmark. However, the performance improvement we observe in all these tests can be also considered the contribution of the enhanced memory controller. In other words, the SSE3 instructions support implemented in Sempron 3100+ processor is more of a marketing move, rather than a feature that should affect the processor performance.

Anyway, we wouldn’t complain about the SSE3 instructions and their implementation in AMD processors. There are still too few applications out there that actually use these instructions. There should be new tasks coming out where SSE3 instructions will have much bigger influence on the performance. Especially, since it turns into a worthy stimulus toward active implementation of SSE3 instructions for the software developers. Once AMD Athlon 64 and Sempron processors acquired SSE3 instructions support, we can state that all CPUs currently available in the market are fully SSE3 compatible.

Overclocking

As we have just seen, the performance of Sempron processors on the new Palermo core hasn’t changed that much. However, besides the SSE3 instructions support and enhanced memory controller, the new core should also boast higher frequency potential. At least the E core steeping for high-performance processors aka Venice can work at frequencies over 2.8GHz without any specific extra cooling (for more details see our article called AMD Athlon 64 3800+ CPU: E3 Processor Core aka Venice at the Door). This fact gives us hope that E revision of Palermo core will also show similar results.

So Sempron processors on E core revision can turn out an excellent choice for an overclocker. Selling at affordable price, these CPUs can theoretically show very good level of performance, if you manage to overclock them successfully and reach high working frequencies.

The overclocking statistics for Sempron cores based on CG and D core revisions indicates that these processors have been producing up to 2.5-2.6GHz stable working frequency so far. And in this case the performance grew up to the level of mainstream CPUs (for more details see our articles called AMD Sempron 2600+ for Socket 754: K8 Architecture Made More Affordable and AMD Sempron 3100+ Overclocking). Keeping in mind that E core revision should boast pretty good frequency potential, we expect that the budget system owners will be able to clock their platforms up to the level of high-performance one without much effort.

So, let’s finish our theoretical discussions and move on to some practical experiments. First of all we would like to point out that successful Sempron 3100+ overclocking requires the use of high-quality mainboard, which can work stably with high frequencies of the clock generator. The maximum clock frequency multiplier of the Sempron 3100+ processor is 9x and it can be adjusted only into a smaller value. So, if we are aiming at breaking the 2.6GHz barrier, we will have to increase the clock generator frequency beyond 288MHz. Far not all the mainboard allow that. Our recommendations about the mainboards for Sempron overclocking experiments remain the same: the only chipset that suits for that is NVIDIA nForce3, and as for the brand name, you should definitely go for the product from a well-known company, which has earned its reputation of the overclocking-friendly solutions maker. As for us, we stick with DFI LANPARTY UT nF3 250Gb mainboard for our overclocking experiments with Socket 754 processors, because this board is known to be one of the best overclocking-friendly products for this socket type. This mainboard allows increasing the clock generator frequency to 300MHz and up without any problems (see our article called AMD Sempron 2600+ for Socket 754: K8 Architecture Made More Affordable for more details about this mainboard).

So, our testbed for overclocking experiments was configured as follows:

• DFI LANPARTY UT nF3 250Gb (BIOS N32LD128.BIN) mainboard;
• AVC Z7U7414001 cooler;
• Corsair CMX1024-4400C25 memory (1024MB, DDR550 SDRAM, 2.5-4-4-8);
• PowerColor RADEON X800 XT (AGP 8x, 500MHz/500MHz) graphics card;
• Western Digital Raptor WD740GD (SATA150) HDD;
• Microsoft Windows XP SP2 OS.

We will start our overclocking experiments on Sempron 3100+ processor based on the new 90nm Palermo core with raising the clock generator frequency without raising the processor Vcore. This way we managed to increase the clock generator frequency up to 286MHz in no time at all. In other words, keeping in mind that the clock multiplier equaled 9x, we managed to increase the processor actual working frequency from the nominal 1,800MHz to 2,574MHz. I should say that this almost free 43% frequency increase above the nominal value is a great achievement so far.

However, this is far not all the new Sempron processor core is capable of. As we know, 90nm AMD cores are very sensitive to any changes of the processor core voltage, thus reaching much higher clock rates with higher Vcore. That is why we increased the processor Vcore by 13% above the nominal for our further experiments: up to 1.58V. In this case the clock generator frequency raised to 298MHz and the system was still working stably. Unfortunately, we failed to hit the beautiful value of 300MHz: the system would boot-up but would never go through the stability tests. However, as soon as we dropped the clock generator frequency to 298MHz, all problems vanished: all stability tests including Prime95 and S&M were completed successfully. The CPU temperature in this case didn’t exceed 55^oC.

So, with the processor Vcore increased by 13%, we managed to reach 2,682MHz clock rate, which is 49% above the nominal value.

Is it a lot or not? On the one hand, this frequency is definitely higher than the nominal frequency of any AMD processors, which allows us to expect the overclocked Sempron 3100+ on the new Palermo core to perform real well. However, on the other hand, we managed to overclock Athlon 64 processor on the Venice core (and this core is Palermo’s direct relative) to much better heights reaching 2.8GHz bar. Although we have already pointed out during our previous Socket 754 Sempron processors tests that these CPUs overclock a little worse than their counterparts from Athlon 64 product family. Maybe AMD uses lower quality dies for its budget solutions, especially since there is special equipment on Fab30 that supports APM (Automated Precision Manufacturing) technology, i.e. detects die features at early production stages.

Nevertheless, we should hardly complain about the obtained result. 2.68GHz core clock is higher than the nominal frequencies of the Athlon 64 processors and higher than the frequencies our previous Sempron processors on CG and D core revisions managed to achieve. So, now let’s find out how greatly the performance of the overclocked Sempron 3100+ on the new Palermo core improved compared with the results we have just obtained for the CPU working at its nominal frequencies.

Performance of Overclocked Sempron 3100+

Of course, the Sempron processor with 256KB L2 cache overclocked to 2.68GHz could easily compete against the fastest Athlon 64 and Pentium 4 processors. In order to evaluate correctly the level of performance an overclocker can squeeze out of the Sempron 3100+ processor on E core revision, we decided to run a separate test session where we will compare the performance of the overclocked Sempron on Palermo core with that of the high-end CPUs from both: AMD and Intel.

So, we assembled a few testbeds with the following configurations:

• CPUs:
  • AMD Athlon 64 4000+ (Socket 939, 2.4GHz, 1024KB L2, ClawHammer);
  • AMD Athlon 64 3800+ (Socket 939, 2.4GHz, 512KB L2, Venice);
  • AMD Athlon 64 3500+ (Socket 939, 2.2GHz, 512KB L2, Venice);
  • Overclocked AMD Sempron 3100+ (Socket 754, 2.68GHz, 256KB L2, Palermo, revision E);
  • Intel Pentium 4 660 (LGA775, 3.6GHz, 2MB L2);
  • Intel Pentium 4 570 (LGA775, 3.8GHz, 1MB L2);
  • Intel Pentium 4 570 (LGA775, 3.6GHz, 1MB L2).
• Mainboards:
  • ASUS P5AD2-E Premium (LGA775, i925XE Express);
  • DFI DFI NF4 Ultra-D (Socket 939, NVIDIA nForce4 Ultra);
  • DFI LANPARTY UT nF3 250Gb (Socket 754, NVIDIA nForce3 Ultra).
• Memory:
  • Corsair CMX512-3200XLPRO, 2 x 512MB, DDR400 SDRAM, 2-2-2-10;
  • Corsair CMX512-4400C25, 2 x 512MB, DDR550 SDRAM, 2.5-4-4-8;
  • OCZ PC2 4300, 2 x 512MB, DDR2-533 SDRAM, 3-3-3-8.
• Graphics cards:
  • PowerColor RADEON X800 XT (PCI-E x16).
  • PowerColor RADEON X800 XT (AGP 8x).
• HDD: Maxtor MaXLine III 250GB (SATA150).
• OS: Microsoft Windows XP SP2.

Trying to squeeze the maximum performance out of a system built around overclocked Sempron 3100+, we synchronized the clock generator frequency with the memory working frequency. And I have to admit that we did it. Corsair CMX512-4400C25 memory modules worked just fine at 298MHz (596MHz DDR) with 2.5-4-4-8 timings when we increased their voltage to 2.8V.

Note that despite the single-channel mode of the Sempron 3100+ memory controller, the memory bus bandwidth grows up significantly during overclocking.

Now let’s take a look at the results of our performance tests.

Synthetic Benchmarks

Games

Audio and Video Encoding

Computational Benchmarks

Applications Benchmarks

Well, the Sempron 3100+ on the new E core revision of Palermo core overclocked to 2.68GHz is quite impressive. It feels quite at home among the top Athlon 64 and Pentium 4 CPU models. In fact, it would be fair enough to state that the overclocked new Sempron 3100+ processor can on average reach the performance level of Athlon 64 3800+ and even Athlon 64 4000+.

However, we shouldn’t disregard the dependence of overclocked Sempron 3100+ performance on the type of applications it is running. Foe example in games, this processor works about as fast as Athlon 64 3800+ leaving behind all Intel solutions. It copes with audio and video encoding faster than any of the Athlon 64 CPUs, but can’t catch up with the top rivals from Intel. The same situation can be observed during final rendering, too. In computational tasks the overclocked Sempron sets new performance records defeating all Athlon 64 and Pentium 4 CPUs. And during image processing, the Sempron processor working at 2.68GHz is at the level of Athlon 64 3800+.

However, if we take into account the price of this CPU, which is now set around $110-$120, then its overclocking performance will definitely be worth the money you pay for it and even more.

Conclusion

Together with the transfer of all Athlon 64 processors to a new Venice core supporting SSE3 instructions and boasting higher frequency potential, AMD did the same thing to its Sempron processor family for Socket 754 systems. The budget Sempron processors acquire Revision E of the Palermo core, which also supports SSE3 instructions and enhanced memory controller, just like Venice. However, these Palermo core modifications do not cause any significant performance improvement. According to our tests, the performance advantage of the Sempron processors based on the new revision E of Palermo core over the predecessors is not higher than 1%.

However, the major advantages of the new core are different. Our investigation revealed that the new Palermo core boasts higher frequency potential compared with what CG and D core revisions of previous Sempron processors had. In particular, we managed to overclock our Sempron 3100+ processor on the new core up to the frequencies approaching 2.7GHz. As a result, the performance of this processor turned out beyond all our bravest expectations: overclocked Sempron 3100+ processor on the E core revision was running neck and neck with the top-end CPUs from Intel and AMD.

So, the relatively inexpensive Sempron processors on the new E core revision of Palermo will become even more desired by the economical overclockers. Processors like that are excellent basis for a high-performance system at a minimum cost.

However, there is still one fly in this ointment. Unfortunately, even with the new core the Sempron processors didn’t acquire AMD64 64-bit extensions support. This way Sempron processor family remains 32-bit only, which can be a pretty frustrating fact for some of the potential buyers.