Review of Final i815/Solano Chipset – THG.RU

Introduction

This review comes right from our Computex Testing Lab in Taipei. We were lucky to receive two motherboards with final i815/Solano silicon, and both boards came from very well known motherboard makers. Due to some time constraint here, I will keep this review a bit shorter than what you are used to, but I’d recommend having a read of the Solano Preview Article from a few weeks ago if you want the full background information on Intel’s upcoming Solano chipset.

Rambus? No, Thanks!

Basically, i815 is Intel’s first chipset with support of PC133 SDRAM and therefore Intel’s first alternative to the unsuccessful chipsets i820 and i840 that require the highly unpopular and expensive RDRAM. Less than a year ago, Intel stated that RDRAM is the future and that they won’t ever supply a chipset with support for PC133 SDRAM. The bad image of Rambus and RDRAM, its rather questionable performance, the superior performance of the good old BX-chipset at 133 MHz as well as several major bugs that were found in i820/i840 forced Intel to turn away from their stubborn RDRAM-only policy and so the plans for i815/Solano were finally dragged out of the drawer. It seems that Intel is finally starting to turn away from Rambus to save their credibility and popularity.

A Worthy Successor of 440BX …?

All the cracks out there know about the excellent performance supplied by Intel’s old 440BX chipset running at 133 MHz FSB. The only problem with this configuration is the fact that BX133-systems are forced to run the AGP beyond spec at 89 MHz, which is not healthy for all AGP 3D cards. Now i815/Solano should finally offer an official solution for people who want a platform for Coppermine Pentium III processors that are using 133 MHz FSB and PC133 SDRAM.

The expectations for i815/Solano are high. Intel’s more than two-year-old 440BX chipset proved how excellent the combination of 133 MHZ FSB plus PC133 SDRAM is able to perform and i815/Solano should be at least as fast as its elder cousin. I made a little table to demonstrate why:

You can see that i815 looks pretty strong compared to 440BX. The only disadvantage that i815/Solano has versus 440BX is the memory support of only 512 MB. It’s certainly not far fetched to expect that Solano should perform either as good or even better than ‘BX133’, because most i815-specs are identical if not better than BX133, as in case of the ATA standard and the AGP spec.

Ugly Politics

We should not forget that nowadays not performance, but politics are actually ruling the IT world. It’s not the better product that wins the market, but the one that’s pushed harder. Otherwise the Rambus-stock would be at some well-deserved $5 and nobody would buy RDRAM systems. It’s not exactly in Intel’s interest to make i815 perform way better than their RDRAM chipsets i820/i840, even though well informed people know from BX133 how well it could perform. Intel needs i815 to offer its disgruntled chipset customers an alternative to RDRAM platforms, but the Rambus-chipsets still need to be pushed and therefore Intel could not let i815 perform significantly better than the much more expensive Rambus solutions. Basically you can say that if i815 should perform as well as or even better than BX133, there was no sensible justification for i820/i840 and RDRAM anymore at all. So Intel had to make a choice. Either i815 would be designed to perform as great as 440BX, which would completely destroy the RDRAM hype, or i815 would be deliberately slowed down, which would upset only the well-informed minority and the overclockers of this world, the other people wouldn’t know about this cheesy issue and Intel could still ‘prove’ that Rambus is better than SDRAM. It’s not too surprising that Intel went for the latter. I can tell you already now that i815/Solano will still not be able to reach BX133 performance in the majority of benchmarks, but at least it’s beating all the rest, including even Intel’s dual-Rambus-channel 840 chipset, in most benchmarks.

The Specs of i815/Solano

First of all, there are ‘two Solanos’, the i815/Solano and the i815E/Solano2. Both chipsets are using the same ‘north bridge’, called ‘GMCH’ = Graphics Memory Controller Hub. The difference lies in the ‘south bridge’. The i815 is using the well-known ‘ICH’ = I/O Controller Hub that we already know from i810, i810E, 820 and 840. ICH supports ATA66, AC97-stuff and 2 USB ports. The i815E comes equipped with ‘ICH2’, which adds ATA100, 4 USB-ports and ‘Advanced Communication & Network Riser’ CNR-slots, which are a beefed up AC97 plus networking. The ‘ICH2’ can be found in new i820 and i840 boards too, which makes the name of those chipsets change to ‘i820E’ and ‘i840E’ as well.

Besides those south bridge features, the ‘GMCH’ supports processor front side bus clocks of 66, 100 and 133 MHz and the memory clock for the supported SDRAM can be chosen between 100 and 133 MHz as well. Therefore i815 is a platform that can host Celeron processors as well as the latest Coppermine Pentium III processors with 133 MHz FSB. At the same time you can use PC100 as well as PC133 SDRAM. You can also see that i815 does mark an improvement over ZX/BX and i810 for Celeron owners. You can run your Celeron at 66 MHz FSB, but the memory will run at 100 MHz at the same time, thus improving performance.

Most of you already know that i815/Solano includes integrated 3D graphics, which we have not tested yet. In the 3D gaming field the expectations aren’t high for this integrated solution, but it’s just fine for office applications. People who want to use a high-end 3D-accelerator of their choice can do that by plugging it into the AGP-slot. As soon as that is done, i815 turns its internal graphics off.

Benchmarking Setup

We ran the tests in the same configuration as we used for ‘The Giga Battle – Part 2‘. This way we were able to compare our i815 results with the scores of BX, i820, i840 and VIA Apollo Pro 133A without having to re-run the whole batch of benchmarks. Therefore we used the driver release 5.08 for the GeForce graphics card. I still checked if there’s a significant difference to the scores under the latest rev. 5.22 drivers and could only find that the Q3-results dropped a bit from 5.08 to 5.22. Hence we feel that there’s even a lot of sense in preferring 5.08 over 5.22.

We used the Asus CUSL2 motherboard for this test. We have already reported on this board from Computex in Tom’s Computex 2000 News – Part 2. The platform was very stable and easy to install, as we are used to from Intel-based platforms.

Benchmark Scores – Office Application Benchmark

The final i815-system could win a few points over our previous test sample, but there’s still quite a gap to BX133. Still i815 can leave most of the competition well behind, only i840 is able to score higher results.

3D Gaming Benchmarks

Again the final i815-silicon could gain a few frames over the previous sample, but it’s still a 7 fps behind the BX133-system. Still we should appreciate that Solano leaves all the other systems in the dust.

In Quake2 Solano is able to get rather close to BX133. Still it is unable to reach the performance of its elder cousin, while leaving the rest of the competition behind, including the Rambus-bunch.

3D Gaming Benchmarks, Continued

Good old Expendable shows the same once more. BX133 still rules, Solano comes in second and the Rambus chipsets i820 as well as even 840 are the losers.

Finally Solano is able to take the lead over BX133 by a tiny margin. This is obviously due to its AGP4x advantage. The rest of the competition looks rather bad against the two SDRAM-cousins from Intel.

OpenGL Workstation Performance – SPECviewperf 6.1.1

We are already used to this picture. Advanced Visualizer shows the same scores across the board, since the frame rate is limited by the fill rate of GeForce rather than anything else.

Design Review cares more about memory performance than AGP-bandwidth. This is why BX133 is still ahead of Solano and i840 can score points as the fastest platform.

OpenGL Workstation Performance – SPECviewperf 6.1.1, Continued

Data Explorer depends on memory bandwidth even more, which leaves BX133 ahead of Solano once more and the RDRAM-systems are able to look good for the first time.

Lightscape wants a high AGP throughput, but still a lot of memory bandwidth too. The AGP4x advantage of Solano is able to show, leading to a higher score than BX133. For the second time the RDRAM-platforms are able to look good, but Solano is still able to beat i820.

ProCDRS shows very similar behaviour as Lightscape. Only i820 looks overly good, probably due to a driver issue of GeForce.

The Integrated Graphics of i815 [Updated]

I guess nobody is surprised to see that the integrated 3D solution of i815/Solano is not exactly a great performer. However, I did not expect the bad results in Sysmark. You should think that Intel is at least able to design the integrated graphics with a decent 2D-accelerator. This does not seem to be the case. Hence I suggest using the integrated graphics for low-budget solutions only. The advantage is that such a system can easily be upgraded by adding a dedicated AGP 3D card. The disadvantage for performance users is the fact that they’ve got to pay for the integrated 3D decelerator, even though they might never use it.

Conclusion

I would have loved to congratulate Intel for the development of i815, but the fact that BX133 is still faster in the majority of benchmarks leaves me unsatisfied. I still don’t see any sensible reason why Solano should not be able to reach BX133-scores or even outperform its two-year-old predecessor. Solano’s ‘underperformance’ is raising some serious questions. Was Intel unable to make i815 as fast as 440BX or did Intel simply not want to make it too fast? I am sure that we won’t ever get an honest answer to this question, so I leave it up to you to answer this question for yourself.

After I pointed out my disapproval of i815’s performance in comparison with BX133, I still have to praise it for leaving its Rambus-brothers behind in the majority of benchmarks. It may be that the overclockers amongst us are disappointed about Solano, but all the ones out there, who were afraid of running BX beyond spec, have now found their perfect solution. Solano is faster and cheaper than i820 or i840 and it beats VIA’s Apollo Pro133A PC133 chipset too, so it deserves to be called ‘Best Coppermine Solution’, taking this title away from the Apollo Pro133A. The question now is how expensive i815-platforms will be. After all you have to pay for an integrated graphics solution, regardless if you want to use it or not. Motherboards with VIA’s Apollo Pro 133A will remain the most cost effective choice, but i815-platforms will be able to deliver superior performance at a price that is significantly lower than i820 and i840. Should i815 ship in volume it will kill i820 sales completely. i840-sales have never been good in the first place and I wouldn’t know why that should change. This leaves a happy smile on my face while I see my hands waving bye-bye to Rambus. Intel’s 815 chipset is a good product and it makes me feel very sorry for all owners of i820-systems. Well guys, that’s what you get if you believe in marketing hype!

The conclusions would not be complete without me pointing out that VIA will very soon release the successor of Apollo Pro 133A. This new chipset will support DDR266 SDRAM, which will most likely outperform Solano. Keep this in mind if you are planning to buy a new system within the next few months. Solano is good right now; it will be old news once the DDR266 systems become available.