Once Intel makes Sense … The Competition has Reasons to get Frightened
How long has it been since Intel could prove this point? Well, Intel showed its competence when they proved last year that they could produce an excellent processor even out of the 7-year old P6-design with the release of ‘Coppermine’ to successfully compete against archenemy AMD and their Athlon CPU. However, since Coppermine’s release people were either suffering in silence or loudly complaining about Intel’s lack of sense when it came down to a platform for the latest Pentium III processor.
Sometimes Intel has strange ways of doing business. Instead of giving the poor public what it wants, Intel decided that it will bless us, or rather try to force us, into using platforms that are supposed to work with Rambus’ RDRAM memory. Besides the fact that the performance of this memory was never able to live up to the largest hype campaign in the history of computers, it is also super expensive and therefore neither worth consideration, nor worth the money you hopefully never paid for it. Basically Intel launched a very good processor and garnished it with chipsets that require memory that nobody wants and that nobody can afford.
Even Intel Couln’t Get Away Without SDRAM Support
In a second step Intel finally smelled disaster and added SDRAM-support to their i820 and i840 chipsets. Unfortunately the way of this SDRAM-support was like adding racecar abilities to a truck. Instead of giving the i820 and i840 chipsets native SDRAM support, Intel created a chip called ‘MTH’ for i820 and ‘MRH-S’ for i840. Those poor chips can talk to SDRAM, but they are doomed to communicate with the main chipset over the Rambus protocol, which is destroying any performance advantage that SDRAM has over RDRAM. It makes those ‘i820/i840-plus-SDRAM’-systems expensive but slow. It was not easy to overlook that Intel hadn’t designed ‘MTH’ and ‘MRH-S’ full heartedly and so it didn’t take long and the first bugs were found in each. First it was clear that neither of them would work with ECC SDRAM reliably and finally Intel found the ‘system reboot/hang’-bug (Tom’s Blurb – MTH – The Story Isn’t Over Yet …, The Impact of Intel’s MTH-Issue on the IT-Business, Tom’s Blurb – All Owners of Systems With Intel’s i820 Chipset That Don’t Use RDRAM Yet Will Now Get It For Free From Intel!) in the ‘MTH’ chip, stopped the production of it and started a huge recall/refund campaign for the poor users that were crazy enough to buy this sorry combination of a doomed chipset and a badly designed add-on chip.
The sweet FreeBSD deamon was used just as a nice devil figure, without any intension to connect Rambus and FreeBSD.
Last year when all the Rambus hype campaign started, which basically consisted of statements like ‘The earth is flat – RDRAM is good’ or ‘Politicians are noble – RDRAM outperforms any other memory type‘, Intel’s competitors in the platform area started to favor PC133 SDRAM and later DDR SDRAM over the crazy Rambus-idea. Today we know how successful this made them. Particularly VIA took great advantage of Intel’s RDRAM aberration. Last year around this time Intel claimed that PC133 SDRAM as well as DDR SDRAM was unreliable and not performing well enough, which would be why Intel will never support either of the two alternative memory types. Later Intel finally realized that stubbornly riding the Rambus train was like a trip directly to hell. Thus suddenly you could find PC133 and DDR SDRAM on Intel roadmaps. The first product that resulted out of Intel’s late insight is the highly anticipated and soon to-be-released ‘i815’ or ‘Solano’ chipset.
The Solano Story
We know for quite a while that Solano is in the works. However, the question might be asked why it took Intel so long to produce this PC133 SDRAM chipset. Several answers are possible and you can pick one. Maybe it is because Intel was so focused onto Rambus chipsets, that they didn’t spend any resource on developments of SDRAM chipsets for quite a while. Maybe Intel sacked all the engineers that brought us the highly successful 440BX-chipset, the last performance SDRAM chipset that came out of Satan Clara in 1998. Maybe those engineers had forgotten everything about SDRAM or suffered from internal ‘RDRAM is the holy grail’-brainwash sessions, also leading to the ‘MTH-disaster’. Maybe, and this seems most likely to me, Intel just wasn’t sure if they should really release a SDRAM chipset, which would mark a step away from the hyped ‘Rambus-blessing’. Whatever the reason was, it is very hard to understand why a company that produced the fastest and best SDRAM-chipset that ever existed in 1998 would take more than six months to create a successor. Politics are the only explanation.
Now Solano is close to its release, but the above-mentioned politics are clouding its fate already. You can hear that the i815 chipset will only be produced in ‘limited quantities’, it will only be shipped to OEMs and it’s produced in the same technology as FLASH-memory, thus limiting production resources, because FLASH-memory is in extremely high demand and its production is already sold out for the next few years.
We have reviewed an early (A1) version of ‘i815’ or ‘Solano’, but even if you should be pleased with what we found you should not get your hopes up as to being able to buy an i815-platform anytime soon. Intel cannot risk the loss of all the money it has blown into Rambus and the RDRAM-hype by supplying the public with large quantities of Solano, for this could be the (well-deserved?) death of Rambus.
Solano’s Specs
First of all I’d like to point out that this article is not ‘supported’ by Intel (surprise surprise!). Solano is still under under non-disclosure and so we don’t have any official data sheets at hands. Therefore we are depending on other sources of information as well as the very motherboard we’ve got for testing.
Basically, i815 is a beefed up ‘BX133’ for the new millennium with ‘optionally’ integrated 3D. Its most important spec is of course the support of PC133 SDRAM. Now this is no half-hearted SDRAM support as found in i820/i840, but real native SDRAM support. “RDRAM – no thanks!” is the motto of i815 and it’s a pleasure to realize that. As opposed to VIA, Intel knows how to make signal paths short. Solano is probably able to clock the processor front side bus (FSB) asynchronously to the memory bus. A recent article at ‘Electronic Buyer’s News’ suggests that. We can still not give you a definite statement however, since we are lacking hard data on this issue. Our test motherboard was definitely NOT able to do this and the PLL used on the board was an ICS-type and not an IMI.
Integrated 3D Graphics
Intel is trying to make a point and wants to prove that i815 is only for the ‘value’ segment and thus no competitor to the wonderful RDRAM-chipsets i820/i840 by integrating a 3D graphics solution into Solano. However, different to the i810 chipset, which forces you to use its integrated graphics, i815 turns off its internal 3D graphics as soon as you plug an AGP-card into the system.
Most of us won’t care much about this added ‘value segment aliby’ feature, but it might come in handy for the OEMs that try and produce cheap Pentium III platforms. We, the performance consumers, will have to pay for these integrated graphics, regardless if we want it or not.
ICH/ICH2 Goodies
The other goodies are all the stuff we’re used to and some more. i815 offers AGP4x support and it’s using the ‘Intel Hub Architecture’ with its advantage of a data path of 266 MB/s bandwidth between ‘North’ and ‘South’ bridge vs. the usual connection via the PCI -bus, which is limited to 133 MB/s.
Depending if Solano comes with the well-known ‘ICH’ ‘I/O Controller Hub’ or with ‘ICH2’ you get even more features. Solano with ICH offers ATA66-support, AC97-support, full ACPI-support (for the ones who care) and one USB hub. ICH2 adds ATA100-support, another USB-hub, a six channel AC97-codec and semi-integrated LAN as long as you add a chip called ‘PHY’ to it.
Feature Overview
We tried to create a short overview table to compare Solano with its current competitors.
Chipset | Intel 440BX | Intel 815 | Intel 820 | Intel 840 | VIA Apollo Pro 133A |
Launch | April 1998 | June 2000 | November 1999 | October 1999 | November 1999 |
CPU Platform | Slot 1/Socket 370 | Socket 370 | Slot 1/Socket 370 | Slot 1/Socket 370 | Slot 1/Socket 370 |
Front Side Bus Clock (FSB) | 66/100/133* MHz | 66/100/133 MHz | 66**/100/133 MHz | 100**/133 MHz | 66/100/133 MHz |
Memory Clock | 66/100/133 MHz | 66/100/133 MHz | 66**/100/133 MHz | 100**/133 MHz | 66/100/133 MHz |
Asynchronous Memory Clock? | No | Probably | No | No | Yes |
33 MHz PCI at 133 MHz FSB? | Yes | Yes | Yes | Yes | Yes |
66 MHz AGP at 133 MHz FS1B? | No | Yes | Yes | Yes | Yes |
Maximal DIMM/RIMM Slots | 4 | 3 | 2 | 4 | 4 |
Maximal Memory | 1024 MB | 512 MB | 1024 MB | 2048 MB | 1024 MB |
VC SDRAM Support | No | No | No | No | Yes |
RDRAM Support (Rambus) | No | No | Yes | Yes | No |
Number of Rambus Channels | N/A | N/A | 1 | 2 | N/A |
ATA Standard | ATA33 | ATA66/100*** | ATA66/100*** | ATA66/100*** | ATA66 |
USB-Ports | 2 | 4 | 2 | 2 | 2 |
max. number of PCI-Masters | 6 | 6 | 6 | 6 | 6 |
integrated 3D Graphics | No | Yes | No | No | No |
AGP Speed | 2X | 4X | 4X | 4X | 4X |
ACPI-Features | Partly | Yes | Yes | Yes | Yes |
* overclocked
** not officially supported
*** ATA100 Support with ICH2 (makes 815/820/840 to 815E/820E/840E)
Performance Expectations
My regular readers know that I’ve already said a lot about this, but I hope they won’t mind that I will repeat myself. All of us performance-hungry PC-users that loved Intel’s BX-chipset for all of its qualities felt very deserted when Intel left the ‘right path’ and started to shove RDRAM-platforms up our tail end. Besides the fact that RDRAM-platforms never lived up to the claims of Intel and Rambus, besides the fact that i820/i840 have a history of failures and disappointments, and besides the fact that the educated PC enthusiast feels at least annoyed if not disgusted by all the hype that Rambus is trying to spread, so besides all of that, Rambus memory is even super expensive, blowing up your system costs significantly. It feels as if you’re being ‘pre-punished’ for making the wrong choice. Maybe it could be seen as a warning as well, saying “Well, we made RDRAM unreasonably expensive, so be aware that you will have been had if you should be crazy enough to buy it!”
The educated performance user had two choices so far. Either he would go for VIA’s Apollo Pro 133A chipset, which offers good performance for a very good price, or he would take the path of a real ‘outlaw’, and run his 133 MHz FSB Coppermine on an overclocked BX platform. The latter includes the risk that the 3D-card might not like the out-of-spec 89 MHz AGP-clock, but once it runs there’s nothing faster than ‘BX133’, as you can read in my article ‘Showdown at 133 MHz FSB – Part 2‘.
‘Solano’ or ‘i815’ is supposed to end this dilemma. It should offer at least the same performance of ‘BX133’ without the overclocked AGP. Therefore we should expect results close to BX133 and better results than platforms with VIA’s Apollo Pro 133A. I almost forgot to mention our anti-favorites i820 and i840. Of course we should expect that i815 outperforms them too, since BX is able to do this already.
The Test Platform
Before you get too enthusiastic about the benchmark results I have to let you know that our platform is not the final product. We received our i815-motherboard from a company that does not want to be named, as you can imagine. This motherboard came with Socket370, 3 DIMM slots, 1 AGP slot, 5 PCI slots and one AMR slot. The board was using the new ‘ICH2’ chip as ‘south bridge’, giving us two USB-hubs or four USB-ports, but ATA100 was not enabled. The chipset stepping of this motherboard was ‘A1’. All steppings with ‘A’ in it are pre-production or beta samples, which normally don’t come with all the performance features enabled. Therefore I suggest that the final i815 will perform better. However, the performance as well as the stability of our test platform was already very good. Equipped with new Intel INF-files we could install Windows98 without a problem and the board never failed even once.
Here you can see the Win98 system manager with all the funky i815 devices.
Due to time constraints we decided against testing of the integrated 3D graphic of i815. Firstly the drivers are early stage, secondly nobody expects great performance and finally only minorities of you will even consider using it. We will supply you with test results of i815’s integrated 3D-solution at a later stage.
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 from 5.08 to 5.22. Anyway, there are no reasons to worry about the test setup that we used for our i815 testing.
Platform Information | |
Graphics card for all tests | NVIDIA GeForce 256 120MHz Core, 300MHz DDR-RAM 32MB |
Hard Drive for all tests | Seagate Barracuda ATA ST320430A |
CPU for all tests | Intel Pentium III 1GHz, 133 MHz FSB |
Intel i815 Chipset Pre-Release Stepping A1 |
|
Motherboard | No Information. |
Memory | 128 MB, Wichmann WorkX MXM128 PC133 SDRAM CAS2 |
IDE Interface | onboard |
Network | 3Com 3C905B-TX |
VIA Apollo Pro 133A Chipset |
|
Motherboard | Asus P3V4X, ACPI BIOS 1002 final, March 2000 |
Memory | 128 MB, Enhanced Memory Systems PC133 HSDRAM CAS2 |
IDE Interface | Promise Ultra66 PCI card |
Network | 3Com 3C905B-TX |
Intel 440 BX Chipset |
|
Motherboard | Asus P3B-F, ACPI BIOS 1005 beta 01, March 2000 |
Memory | 128 MB, Enhanced Memory Systems PC133 HSDRAM CAS2 |
IDE Interface | Promise Ultra66 PCI card |
Network | 3Com 3C905B-TX |
Intel 820 Chipset |
|
Motherboard | Asus P3C-L, ACPI BIOS 1020 beta 05, March 2000 |
Memory | 128 MB, Samsung PC800 RDRAM, RDRAM clock adjusted in BIOS |
IDE Interface | onboard |
Network | Onboard i82559 |
Intel 840 Chipset |
|
Motherboard | OR840, special unreleased BIOS |
Memory | 128 MB, Samsung PC800 RDRAM 128 MB, Samsung PC700 RDRAM, running as PC600 RDRAM |
IDE Interface | onboard |
Network | Onboard i82559 |
Driver Information | |
Graphics Driver | NVIDIA 4.12.01.0508 |
viagart.vxd for VIA Chipsets | 4in1 4.17 (download here!) AGP-driver 3.56 |
ATA Driver | Promise Ultra66 driver rev. 1.43 Intel Ultra ATA BM driver v5.00.038 |
Environment Settings | |
OS Versions | Windows 98 SE 4.10.2222 A Screen Resolution 1024x768x16x85 Screen Resolution 1280x1024x32x85 for SPECviewperf |
DirectX Version | 7.0 |
Quake 2 | Version 3.20 command line = +set cd_nocd 1 +set s_initsound 0 Crusher demo, 640x480x16 |
Quake 3 Arena | Retail Version command line = +set cd_nocd 1 +set s_initsound 0 Graphics detail set to ‘Normal’, 640x480x16 Benchmark using ‘Q3DEMO1’ |
Expendable | Downloadable Demo Version command line = -timedemo 640x480x16 |
Unreal Tournament | Ver. 4.05b high quality textures, medium quality skins, no tweaks 640x480x16 Benchmark using ‘UTBench’. |
Office Application Performance under Windows98SE
From the Giga Battle or the Showdown at 133 MHz FSB you might remember the astonishing performance that BX133 is able to produce in Sysmark2000. Pre-release Solano is already able to leave everything except of BX133 and i840 w/Pc800 RDRAM behind it. You can see that i820 has already lost and I wouldn’t be surprised if i815 would smoke i840 as well once it is in full operation mode.
3D Gaming Performance – Quake 3 Arena
I was a tad bit disappointed with Solano’s score here, but let’s not forget that this is ‘A1’-stepping! Solano was not quite able to surpass VIA’s Apollo Pro 133A, but it’s at least on-par with i840. Again, i820 doesn’t look too good at all even against this beta- i815!
3D Gaming Performance – Quake 2
Scoring second place in Quake2 is certainly a commendable performance of the beta-i815. Even i840 can’t quite reach Solano and i820 is lagging behind.
3D Gaming Performance – Expendable
In Expendable Solano can surpass VIA’s PC133 chipset and is again neck on neck with i840. You can see how bad i820 looks here, so I don’t need to mention it once more.
3D Gaming Performance – Unreal Tournament
In Unreal Tournament our beta-Solano can almost reach BX133, as we will expect it from the final production silicon anyhow. Even i840 is left behind, not mentioning i820.
Professional OpenGL Performance – SPECviewperf 6.1.1
Awadvs-03 runs into GeForce’s fill rate limit with any platform, so we don’t learn exactly much from this chart.
In DesignReview the beta-i815 shows that it still needs quite a bit of work. Its place slightly behind i820 is a disgrace and can not be explained with anything else than the early status of our Solano platform.
Professional OpenGL Performance – SPECviewperf 6.1.1, Continued
The same as above holds true for Data Explorer. The early Solano scores rather bad.
In Lightscape Solano can surpass BX133 for the first time, taking advantage of its AGP4x-support. Still the performance needs improvement, because i820 is almost as fast as our early i815.
CDRS is another AGP4x-lover, so that i815 can surpass BX133 once more. Still the performance needs improvement. The high scores of i820 are only explainable with some strange GeForce driver issues.
Conclusion
There’s no doubt, our early i815-sample shows that this chipset will most likely be what i820 should have been. Although the early stage of the Solano-platform could not quite fulfill all of our expectations, it is showing that Intel is on the right track. We expected BX133-performance and that’s what Solano should give us at least. Nevertheless, even our beta-Solano was already able to beat i820 in most of the consumer benchmarks. This should be proof enough that i820 and its RDRAM-support are absolutely obsolete in the consumer performance segment. Even i840 will have a hard time against Solano and that should finally raise the question for the basic justification of RDRAM altogether.
Solano will be released on the first day of Computex2000, which will be June 5, 2000. Once the i815 chipset is official we will have to see if the consumer out there will be able to purchase i815 motherboards. Intel is not planning to supply any reasonable quantities to motherboard makers that sell via retail, even for OEM customers there shall be only ‘limited quantities’. Intel’s Solano may turn out to be a great product. It could threaten VIA’s Apollo Pro 133A sales, but we might never see that happen. Intel has decided to choke us with Rambus and i815 could destroy all of Intel’s plans. That is not supposed to happen. Therefore most of you might only have the chance of dreaming of a Solano motherboard.
The good performance of our early i815 motherboard sample should raise one question. Why didn’t Intel release a chipset like this with Coppermine in Fall 1999? Was it really necessary that all those Rambus/i820 disasters had to happen? Is Intel not interested into what their customers want? What kind of policy is it to patronize the customer and force him to buy a product that he doesn’t want? The world would do just fine without Rambus. In fact the world might do a lot better even. Today there are hundreds of thousands of disgruntled i820 customers who wasted their money for RDRAM and a platform that doesn’t cut it. Those people will never get their money back, and Rambus is using it to finance one questionable marketing campaign after the other. Right now SDRAM is offering the best performance and the best price. The future will probably also have better solutions than a product that has to be shoved down a customer’s throat while milking him really badly at the same time!
Finally I’ve got a little quiz for you once you’ve digested this article. How do you think the performance would be if Solano would support DDR SDRAM instead of PC133? I hear that DDR SDRAM is ‘just’ able to compete with RDRAM … is that supposed to be comedy or the product of someone who simply doesn’t want to see the truth right before his eyes?
Follow-up by reading the article ‘Tom’s Blurb: Intel’s ‘Almador’ Chipset, Solano News and the MTH-Debacle‘.