Rise is an American company. It was founded in November 1993 and its main goal was x86 compatible processors. It wasn't that well-known till 1998. However, about a year ago the company changed its course so drastically that you could hardly disregard it. The company started showering mass media agencies with numerous press releases of very unexpected contents. Frankly speaking, Rise seemed to have totally changed its major to developing advertising strategies instead of hardware stuff.
And in the meanwhile Rise promised a x86-compatible Socket 7 processor intended for sub-$1000 PC and notebooks market. It is also known as mP6 processor. According to Rise, this processor should perform almost as fast as Intel Pentium II at the same frequencies. And no matter if it is a really cheap one. This great effect could be achieved due to superpipelined superscalar architecture, which lets three integer instructions, three MMX-instructions or two floating point operations run simultaneously. This is theoretically the best way to surpass Pentium II, which can boast only two integer pipelines, one single FPU block and two MMX blocks. So, we have to admit that AMD K7, which hasn't come out yet, is Rise's only worthy competitor.
Actually, all this seems quite achievable, especially taking into account Rise's skillful engineering team. Though this company doesn't have its own productive capacities, there are about 80 engineers, who used to work in other processor companies such as AMD, HP, IBM, LSI, Logic, MIPS, NEC, Sun, etc.
And now Rise mP6 is already available in retail despite all doubts and scepticism. Here is its specs list:
- Electrically compatible with Socket 7;
- Software compatible with Intel Pentium family and all x86 applications;
- 0.25 micron technology;
- The marking uses Pentium Rating (PR), which compares its performance to that of Intel Pentium MMX;
- The following models are produced now or may be available later with PR 366, 333, 266, 233, and 166 MHz;
- Supported FSB frequency 83, 95, and 100MHz (besides 60, 66, and 75MHz);
- Clock multipliers: 2x, 2.5x, 3x, 3,5x;
- 2.8V CPU core voltage;
- 16KB L1 cache (8KB for instructions and 8KB for data);
- Two pipelined FPU blocks;
- MMX superscalar module;
- Three superpipelined integer blocks;
- 296 Pin BPGA (Socket 7) or 387 Ball BGA package;
- Additional ways of power saving.
Let's specify the features absent by other CPUs, i.e. advanced power management and power reduction. Rise mP6 was initially designed as a notebook processor, that's why everything concerning power consumption was the major point of this project. And the developers really did their best. As a result the processor demands very little power when the internal clock generator is in "off" status. The needed energy is hundred times (!!!) lower than by other x86 processors, available in the market. And in normal working status the CPU can simply turn off the clocks for those parts of the processor, which are not needed for the current operation. It allows to work even without a cooler! And in order to provide sufficient CPU cooling, one single radiator is more than enough. Rise mP6 processor stays cold!
Besides, the processor has a very unusual for a CPU chipset-shaped package - Ball BGA. And it's a perfect solution for notebooks: looking as an ordinary chipset it needs less room than a CPU. And if you are going to install it into your desktop PC? Well, it's not a problem. In this case Rise solders the chip onto a special small adapter card and that's all. This converter-card can be seen on the picture below.
However, such brilliant features, which seem just irreplaceable for a mobile PC, will not guarantee a really wide use of mP6 processor in them. The matter is that notebooks manufacturers are mostly large companies. And they don't feel like sacrificing their image for the sake of cheaper products. The Rise processor, which hasn't yet acquired great popularity is very unlikely to win the favour and acknowledgement of computer giants. Even a famous Cyrix MediaGX processor is used only in a rather unpopular Compaq model. That's why Rise mP6 should be regarded primarily as a low-cost CPU for desktop systems, such as IDT Winchip.
We'd like to draw your attention particularly to PR (Pentium Rating) in CPU marking. What we mean to say is that the true working frequency of the processor does not coincide with the one indicated on the CPU. Here is the table of corresponding PR and mP6 frequencies:
|Rise mP6 Pentium Rating||366||333||266||233||166|
|FSB frequency, MHz||100||95||100||95||83|
|CPU frequency, MHz||250||240||200||190||166|
It can be easily noticed that relatively insignificant gains in working frequency cause tangible changes of the PR. This value is not invented. It is calculated depending on the real performance of the processor. So, let's try to find out why the performance increases so rapidly. It probably depends on a relatively small L1 cache - 16KB. Just to make it clear: the last CPU with such a small L1 cache was Intel Pentium. And the modern AMD K6-2 has increased its size up to 64KB. This may seem a vexing omission but on the other hand, Rise mP6 has a small processor core, which is a certain advantage for notebooks. Though it is sometimes doubted to be that cool.
And before going over to benchmarks let's try to predict the future of Rise mP6 processor in the market. And its future seems to us pretty gloomy. Remember Cyrix. They had to close up their business here because of the violent competition between Intel and AMD, which led to significant price drop on CPUs. IDT and Rise belong to the low-cost system market, to extremely low-cost market, shall we say. But they seem to be unable to keep their feet even there. Now Rise doesn't feel like giving in, but everything may change any time because it is the giant companies that boss the show, and not the small ones, like Rise, for example.
Well, emotions and thoughts are enough for today. It's time for the benchmarks. First, take note that it is not only those several mainboards officially listed on Rise web site, which support Rise mP6 processor. Practically all the Socket 7 mainboards with a more or less up to date BIOS recognize and support this CPU. BIOS identifies it as a Rise mP6 processor, when the system is getting started.
Our testing system was configured as follows:
- Chaintech 6BTM and Chaintech 5AGM2 mainboards;
- ASUS V3400TNT graphics card (on Nvidia Riva TNT chipset);
- 3D-accelerator Creative 3D Blaster Voodoo2;
- Soundcard on Ensoniq ES1370 chip;
- IBM Titan DTTA 371010 harddisk;
- 128MB SEC PC-100 SDRAM;
- All 3D-benchmarks were run at 800x600x16 resolution.
We compared the performance of Rise mP6 266 we had with that of its main competitor, which managed to survive for the time being - IDT Winchip-2 266, and with AMD K6-2 266. Besides, for a better comparison we also post the results shown by Intel Celeron 266.
And the first benchmark is as usual Winstone 99:
Here Rise mP6 performed much worse than all the other processors. Isn't it strange? Remember that the PR rating is based on the results of this particular benchmark. Did Rise really mean to make a fool of everyone? Of course not! It just outwitted them in the following way. IDT calculates the PR of its Winchip-2 basing on the AMD K6-2 processor, while Rise takes Intel Pentium MMX for the same purpose. And the latter falls behind AMD K6-2 in all office applications. That is why Winstone results are not that surprising at all.
Now let's try to explain such poor performance of Rise mP6. To do this we'll take the results obtained in CPUmark 99:
Here Rise mP6 is also not quite up to the mark. What's the matter here? The final index of the CPUmark 99 results from the system memory and integers operating speeds. If we consider only the way Rise mP6 operates integers, we will get the following picture. The first one is Rise mP6, which manages three instructions per time step. Then goes AMD K6-2, which fulfills two operations simultaneously. And the last will be IDT Winchip-2 with its single integer pipeline. Then it appears quite evident that it is memory operating speed, which spoils the show. And that's absolutely true. A relatively small 16KB L1 cache of Rise mP6 compared to 64KB one of its rivals prevents Rise from fast data access. So, Rise's memory bus turns out its main brake.
The results of the FPUmark prove this conclusion:
Operating memory is not that important here that's why mP6 almost catches up with K6-2, though its working frequency is 1.33 times smaller. Well, we have to admit that Rise's superpipelined FPU contributes greatly to its beautiful performance.
And now a few words about MMX. Intel Media Benchmark showed the following results:
The picture doesn't differ that much from CPUmark 99. Though Rise can boast three MMX blocks compared to only one by IDT Winchip-2 and AMD K6-2, it still failed to surpass its competitors. It has most likely happened because of the small L1 cache. It's size doesn't let Rise mP6 activate its full computing capacity in MMX-instructions as well.
Take a look at Rise mP6 gaming performance. The first game we offer you is, as usual, Quake2:
Well, the results are - to say the least of it! - quite oppressive. We even disabled the 3DNow! block of both: IDT Winchip-2 and AMD K6-2, so that the column denoting Rise mP6 performance could be at least seen on the diagram and didn't merge with the horizontal axis. :) So, our conclusion sounds as follows: forget the games unless you get a normal-size cache.
However, we decided to get to the bitter end of our testing and started the Unreal game on Rise mP6:
And again our worst expectations came true. Judging by all mentioned above, we announce our verdict: Rise mP6 is absolutely unsuitable for games. Neither superpipelined FPU, nor superscalar architecture can save it from being so discredited. We would actually advise the Rise company to think about increasing L1 cache size inside the processor.
If we take a look at the company's nearest plans, the situation will become very clear. First, Rise is going to switch over to 0.18 micron technology, to decrease the power voltage to 2V, which is typical of all SuperSocket 7 systems. As a result the energy comsumption and heat dissipation will beome even lower than they are now. Sounds very promising, especially if someone finally decides to use Rise mP6 or its next mP6 II version in notebooks. However, these changes will hardly help improve the performance and hence Rise is very likely to lose totally the sub-$1000 PC market.
Second, the company is planning to launch mP6 II (a newer version of mP6) for Socket370 with SSE-instructions support. This move is really worth mentioning in a couple of press-releases, isn't it? However, SSE is not so widely spread now that't why adding SIMD-instructions is mostly a marketing measure. As for Socket370, Rise will never succeed here, because the youngest Celeron processors are definitely much faster and cheaper. And Rise mP6 will hardly manage to survive in these conditions. The only thing that gives us some hope is a 256KB L2 cache working at the bus frequency, which is to be integrated into mP6 II core. The performance may increase by somewhat 20% in this case, but Rise will still fail to catch up with the competitors, because they are also constantly making progress.
Summing up we would like to say that making press-releases is Rise's only coolest point for the time being. Especially, taking into account their terrific skill to find every reason for that. And Rise seems to continue pleasing us with its bunches of different marketing stuff unless it follows in the footsteps of Cyrix. Actually, in the present situation in terms of pricing competition between Intel and AMD it may turn out exactly this way.
But, back to the present and to Rise mP6 overclocking. As we have already said, the working CPU remains absolutely cold. It was so cold that it kept on working even when I stopped the cooler fastened on the radiator. However, that was all we managed to achieve. Even a slight frequency increase - and the system doesn't start properly.
The first and the most important thing is Rise's low performance despite all the innovations they introduced. It is the small L1 cache, which is to blame, because it brings to nought all the efforts of core developers. It means that mP6 is not worth even thinking of until Rise changes its cache size. The only possible way to apply this device is to install it into a cheap notebook. However, it may happen only if there is a brave notebook manufacturer, who will mark its products with "Rise inside" and disregard all rumors and mockery.
Unfortunately, we have an impression that the number of x86 processor manufacturers is likely to keep on getting smaller. Last year there were five manufacturers, this year - only four. And it's still going down despite some optimistic forecasts about four new processor manufacturers. Intel and AMD's pricing competition has a pernicious effect on smaller companies. And they could actually add a new strength to this technical stagnation.