*Hiroshige's Goto Weekly overseas news*
This summer, Prescott is hot. Next year to electric power consumption 100W over
--------------------------------------------------------------------------------
- After all hot Prescott
Intel, 2 types throws CPU of the next generation production process 90nm, within year. 2nd generation Pentium of the schedule which is announced to around the end of 3rd quarter M "Dothan (?h?^??)" with, "Prescott (the press cot)" of the Pentium 4 succession which is thrown during 4th quarter is. Intel, it throws Prescott within year first with 3.4GHz, next year 1st quarter pulls up to 3.6GHz. As for Prescott verification of the sample has been advanced already, Intel is advancing the preparation to sale favorably. It was the expectation which well, is being advanced favorably.
Falling to there, those where it boils, are the thermal problem of Prescott. The electric power consumption of Prescott, being higher than Intel itself expectation substantially was ascertained. Somehow, TDP of Prescott (Thermal Design Power: Thermal design electric power consumption), it exceeds the mark of 100W finally at the frequency of 3.6GHz, it seems. Because the expectation of Intel of beginning was 89W, it means that also 15% increased at a stroke.
Because of that, Intel has begun the specification modification of the Prescott corresponding motherboard. As a result, even with the present Intel 875/865 motherboard which is assumed that it can correspond to also Prescott, the case which it cannot correspond to Prescott coming out, you rival, is. As for being more painful, with the note PC which loads Mobile Prescott which appears in the next year head, considerable influence probably will put out, is.
In addition, in the long term, as for this case the fact that the brake is not applied to the increase of the electric power consumption of Intel CPU, is hinted. If it becomes so, PC becomes something which more and more the design and the original work are difficult to do. Furthermore, 90nm process of the same company, has suggested also the possibility electricity consumption being more than schedule.
- TDP of Prescott which it has an influence on the FMB guideline
Intel before usually, actual electric power consumption measures with the sample tip/chip, conveys expectation electric power consumption to the OEM vendor. This fine control being in order the corresponding motherboard preceding, to have designing, after that, electric power consumption and product plan of the actual sample tip/chip (it commercializes to no GHz) together. This is usual pattern, but as for those where this time it is different, electric power consumption with pre- silicon and post silicon to be large is to slip.
Actually, as for at this story from about 1 - 2 months ago in PC industry it seems that is known. According to that, as for Intel, that being maximum, you estimated TDP of Prescott of the ?Ê PGA478 package that it is 89W, you say. However, as for the result and this numerical value which advanced verification with actual silicon it was rewritten in 103W. When electric power consumption increases, you must increase either the quantity of the electric current which is supplied to CPU. Because of that, the supply electric current to CPU on the motherboard (IccMAX) it was increased to 91A from 78A. In other words, as for the motherboard which cannot supply that much electric current quantity, it comes to the point of not being able to correspond to Prescott.
- With Prefetch architecture in many fold
The extension of TDP of Intel CPU
As for PDF edition this
Intel, guideline "Flexible of the motherboard design which is offered vis-a-vis the OEM vendor Motherboard (FMB)" influences this problem, largely. Intel offers FMB of 2 types to every each CPU generation. For example, present 0.13 ?Ê m edition Pentium 4 (Northwood: The north wood) vis-a-vis, there is "Northwood FMB1" and "Northwood FMB2".
Those where it corresponds to Prescott when with "Prescott FMB1", it conforms to this, had come to the point of being able to support the both of Northwood and Prescott. Those which already have conformed to these specifications it is many expectation in the motherboard.
William M. of Intel Siu (the ?E?C???A?? M sou) the vice president and the general manager (Vice President and General Manager and Desktop Platforms Group) you explain as follows concerning Prescott FMB1.
"(Prescott) FMB1 is the design which in the customer from Northwood makes the movement to Prescott possible. When the CPU socket and movement of chip set accompany (the movement of CPU), it cannot move the customer easily rapidly. But, in case of Prescott as for big modification there is no necessity (in the motherboard). Therefore, the movement to Prescott thinks "that it is rapid,
With Prescott FMB1, EOL (last edition) to Northwood (3.2GHz) with, it corresponds to Prescott of ?Ê PGA478 edition. Intel has been about on to move with Prescott 3.6GHz compared to new package LGA (Land Grid Array) 775 ?w. In other words, with Prescott FMB1 it means to be the schedule which it can correspond to Prescott to 3.6GHz. By the way, it has become the schedule which generation CPU "Tejas (?e?n?X)", corresponds with Prescott of the LGA 775 generation one after another with Prescott FMB2.
However, because TDP of the actual silicon of Prescott rises too excessively, as for Intel when plan is changed, forcing. According to a certain information muscle, you say that presently Intel having decided and offering Prescott FMB1.5 of the correction edition specifications. When with this Prescott FMB1.5, TDP 103W and IccMAX approximately 15% is increased to 91A, forcing. The fact that another specifications, FMB1.5 were made, because modification is large, is the case that the necessity to divide the FMB specifications arose. Somehow, unless it is the motherboard of Prescott FMB1.5 conformity, Prescott seems a story that it cannot support.
Because as for the influence to FMB being large, it is difficult already to stabilize to supply the heavy-current quantity even under present conditions through the motherboard. This is entwined closely with cost. Present Voltage Regulator of Intel Modules (VRM) specification, stabilizing, can supply heavy-current has become "VRM 10.0", with VRM 10 multiple phase is supported. And, Intel, is said that at the beginning, with Prescott FMB1 3 phases, 4 phases were required with FMB2. But, really you call 3 phases that it was not excessively accepted in the cost aspect. It is said But that because TDP keeps rising, in the future quickly the necessity to increase the number of beam phases probably will arise.
- TDP of of 100W over the head is painful
Another problem, until you say, is the thermal design with the rise of TDP without. Usually, when TDP rises, furthermore it becomes troublesome to cool CPU. However, this time TDP of Prescott rising, heat sink and Ta (basket internal temperature) the specifications without modification call Intel that it can go, it seems. In other words, CPU the case that does not change cooling into having become hot and, the ?Ä is good it is. The seed of this magic is simple. As for Intel Tcase (the package temperature of CPU) it modifies.
In the thermal design of PC, it becomes the element whose temperature difference of the air of CPU and basket internal and the electric power consumption of CPU, and heat resistance value of the heat sink are important. Concretely, (Tcase - Ta) the ??tdp = heat sink (the + thermal interface material) heat resistance value, with it becomes.
Among these, heat resistance of the heat sink is restricted first. If heat resistance value is displayed in degree C/W and value is small, small extent it is superior, but cost becomes high. Technological progress of the heat sink pace is slow, presently 0.33 degrees C/W with PC are made the limit.
On the one hand, temperature of basket internal is shown "Ta (ambient)" with. If it is low, the thermal design of the around low extent CPU becomes easy, but that much, technology in order to reduce the temperature of basket internal becomes necessary. For example in the past as for Ta 40 - 45 degrees C (, Northwood FMB2 was 42 degrees C), but with present Prescott FMB1 38 degrees C it has become harsher. Also these specifications, will lower above this, but it is in a state where it is not.
With being the case that it is said, in the circumstance where two elements are fixed, one parameter changed. So when it does, if one parameter which naturally, remains is not changed, it does not consist. In other words, amount and Tcase where TDP rose were pulled up. As for Tcase of Prescott of the time before those where they are 69 degrees C, with 1.5 have reached 74 degrees C. Actually Tcase rise ratio is smaller than ratio of TDP rise, but that is thought that is, because FMB1 room was seen. That this measure influences it is thought that it is the yield rate and the like of Prescott. In other words, the development whose are inconvenient for Intel it is reason.
But, the desktop still is better one. As for problem, Intel desktop replacement (DTR) throws Prescott to also note PC. Because for DTR substantially is TDP of the same level as the desktop, it has become very troublesome circumstance.
To recently TDP of Mobile Prescott was 74W with 3.46GHz. But presently, as for TDP of Mobile Prescott you say that it has become 94W. With note PC 94W! Some DTR say, as for this it is considerable challenge.
Intel even with Mobile assures the solution by the fact that the desktop in the same way Tcase is pulled up. Tcase of former Mobile Prescott was 72 degrees C, but now this has reached 76 degrees C, it seems. But, because even then it is not enough, with DTR also Ta 5 degrees C seems that is lowered. In other words, as for the note PC designer, you do not think of the waste heat mechanism which 5 degrees C in temperature of basket internal can reduce from so far it is the case that it becomes necessary.
With being the case that it is said, "hot Prescott" has started bringing many distortion. Though, Intel measure seems that is doing. There is also information that release it does Prescott which can be supported even with Prescott FMB1. Although Prescott of the high clock is supported with FMB1.5 you held down to TDP of the range which can be supported with FMB1, it is a story that it puts out also Prescott where the single step clock is low. But, as for details it is not found yet.