"Intel Cuts Power Use on Server Chips"

[kukito]

Anybody with a WSJ subscription care to chime in? link

The Inq mentioned this yesterday

http://www.theinquirer.net/default.aspx?article=38138

These are Conroe and Kentsfield based Xeon parts. 50 watts? Even for the Kentsfields?

 

[MU_Engineer]

It's very possible to get a quad-core chip to run within a 50W TDP. The Core 2 Duo LV chips (1.33 GHz L7200 and 1.50 GHz L7400) have TDPs of 17W. Put two of those together than you have a 34W TDP quad-core. This is what Intel will do with those quad-core server CPUs: keep clock speeds and voltages low to keep within the thermal envelope.

 

[kukito]

Thx Engineer. So I don't have to wait for Penryn then to get the low power quad core I want. The X3220 is here, with the same speed of my E6600 and four cores. At a price that I'm almost willing to pay. I might even go for the cheaper X3210 and overclock it. The power drop of 38% to 58% is amazing.

 

[ltcommander_data]

Thx Engineer. So I don't have to wait for Penryn then to get the low power quad core I want. The X3220 is here, with the same speed of my E6600 and four cores. At a price that I'm almost willing to pay. I might even go for the cheaper X3210 and overclock it. The power drop of 38% to 58% is amazing.

See this is where The Inquirer's editing is flawed. Nothing in that chart are low-voltage chips. They are all standard voltage single processor Xeon chips, with a 65W TDP for the Conroe dual cores and a 105W TDP for the Kentsfield quad cores.

Here is a more accurate story:

http://www.theregister.com/2007/03/09/intel_lv_fourcore_xeon/

The low-voltage Xeons with the 50W TDP will begin with a "L". What is noteworthy though is that they are going to launch a 1.86GHz L5320 quad core with a 50W TDP since originally they were only planning on doing a L5310 1.6GHz one. This definitely points to a maturing 65nm process. The pricing is $519 and $455 respectively. Note these low-voltage models are only for dual processor systems.

 

[kukito]

See this is where The Inquirer's editing is flawed. Nothing in that chart are low-voltage chips. They are all standard voltage single processor Xeon chips, with a 65W TDP for the Conroe dual cores and a 105W TDP for the Kentsfield quad cores.

Flawed and sloppy. They're listing totally different parts than what they're writing about. :roll:

Thanks for the clarification, data. I'll continue to wait for Penryn.

 

[MU_Engineer]

Yup, they're clocked about where I'd figured they would be. The C2D LV chips were 1.333 and 1.500 GHz and had a TDP of 17W per die and the fastest regular Core 2 Duo T7600 has a TDP of ~35W, so I'd have figured something in-between, in the upper 1 GHz range.

What will be interesting to see is how AMD will combat that push against their server space. A pair of 22x0HEs will consume 67W per socket and return overall a similar power draw to the LV Clovertowns once you add in the motherboard. But there's nothing that can come close for eight cores- an 82x0HE setup will use over twice the power as the QC Xeon LVs. Barcelona should be interesting as the HE versions of are suppsed run at about 70W up to 2.1-2.2 GHz, where they should outperform the Xeon LVs in both raw performance and performance-per-watt, assuming that the Barcelona is at least as fast clock-for-clock as the Core chips are. (I'm not going to get into THAT discussion here, but I think that it's a safe bet to say that they should at least get IPC parity.) However, the Barcelonas will draw greater power overall. Maybe they will just stick to the performance-per-watt metric and forget about trying to match Intel watt-for-watt, otherwise there might be some lower-clocked Barcelonas that aren't on the roadmap getting released. It will be interesting to say the least.