Background info: I have a comfortably quiet PC built with a C2D e8400 which I run on passive cooling. At 3.16 GHz and 1.06 v it maxes at 55 C give or take a degree of ambient temperature. I also run an AMD 5770 passively in this system. However C2D is starting to show its age and I'd like to replace it with one of the upcoming Sandy Bridge CPUs.
As a silent computing enthusiast, the most important number I consider when buying components is TDP. But now that all of Intel's products have integrated graphics (which by the way is a strategic move I will never ever understand) and 'turbo' frequencies, the real TDP of these products has become a matter of some mystery to me. The TDP of the e8400 is rated 65W, and I simply cannot consider going back to active CPU cooling, so this is a number I must either match or beat in whatever I replace it with.
Looking at a table on Wikipedia, the highest-performing parts at 65W TDP should be the Core i7-2600S (2.8/3.8 GHz turbo) and the Core i5-2500S (2.7/3.7). No doubt I'll be paying a premium for the 'S' designation, but that's fine with me. My questions are about integrated graphics and turbo.
First of all, do the integrated graphics in the Core iX processors featuring them have a kill-switch? I mean a setting, BIOS or otherwise, that completely, electrically, disables them so that not one transistor is adding extra heat to my system? If so, does this lower the 'effective' TDP of my CPU, or does it simply bring it down to the 65W stated? In other words, does Intel's quoted 65W refer to the total power when all X cores are maxed, AND the integrated graphics is also 100% maxed (i.e. running furmark)?
Second, how will my passive configuration affect the function of the turbo? I gather from reading reviews and product literature that the CPU automatically adjusts its operating frequency according to workload, remaining within its thermal/power envelope. Since I plan on running passively, will the turbo be unable to reach the specified maximum frequencies? Will I be able to undervolt, and will that improve the thermal envelope with respect to the turbo function? It doesn't look like these will be easily overclocked, and I don't want to be stuck at 2.7 GHz. Even if the architecture is 25% faster or whatever, I can easily sacrifice a few degrees C and run my C2D at 3.5 GHz or higher, REAL 24/7 GHz not 'turbo'. Will I really be getting a performance increase out of these new chips relative to my old faithful C2D? Obviously 4 cores will allow greater performance than 2 in those applications which can be designed to utilize them, but it is equally obvious that there are still many applications which are not so designed.
Q's about Intel, integrated graphics, TDP, and silence
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
silentplummet
- Posts: 33
- Joined: Wed Feb 28, 2007 4:55 pm
Re: Q's about Intel, integrated graphics, TDP, and silence
Looks like you're talking about the upcoming Sandy Bridge processors. As these parts have not been released, the following comments can only be seen as estimates.
The TDP on Intel processors represents a 'realistic' load number. This includes everything on the chip. On the Sandy Bridge chips, this will include the processing cores, the integrated memory controller, and the integrated graphics. The important thing is that this signifies a 'realistic' load. So, if you loaded your processor with Prime95, the graphics with Furmark, and the memory subsystem with something rediculous, it's possible that you may exceed the TDP on the processor. Intel TDPs seem to be quite liberal, however. I've seen some of the Core2Duo's which are rated for a TDP of 65w to consume no more than 45w in Prime95. But, as this is an unreleased processor, we don't quite know.
As for the turbo implementation, from what I've heard it will be more advanced than the one on current Lynnfield and Clarkfield chips. Whereas the current chips will turbo to make sure a TDP is not breached, the Sandy Bridge chips will turbo to make sure a certain temperature is not breached. This means that they may turbo more aggressively than normal until a certain temperature, then it will go down to the 'max' turbo frequency. Using passive cooling may limit this higher level of turbo-boost, but we won't really know until we get the actual chips to tweak.
If I were to guess, I would say that under full load, the i7-2600S would be hotter than your current E8400, but it would be a lot faster than your current processor. Under normal loads (such as email, a program maxing out one core, etc) I would imagine it would run cooler or at least as cool as your current processor. But again, we won't know until we get our hands on the chips.
The TDP on Intel processors represents a 'realistic' load number. This includes everything on the chip. On the Sandy Bridge chips, this will include the processing cores, the integrated memory controller, and the integrated graphics. The important thing is that this signifies a 'realistic' load. So, if you loaded your processor with Prime95, the graphics with Furmark, and the memory subsystem with something rediculous, it's possible that you may exceed the TDP on the processor. Intel TDPs seem to be quite liberal, however. I've seen some of the Core2Duo's which are rated for a TDP of 65w to consume no more than 45w in Prime95. But, as this is an unreleased processor, we don't quite know.
As for the turbo implementation, from what I've heard it will be more advanced than the one on current Lynnfield and Clarkfield chips. Whereas the current chips will turbo to make sure a TDP is not breached, the Sandy Bridge chips will turbo to make sure a certain temperature is not breached. This means that they may turbo more aggressively than normal until a certain temperature, then it will go down to the 'max' turbo frequency. Using passive cooling may limit this higher level of turbo-boost, but we won't really know until we get the actual chips to tweak.
If I were to guess, I would say that under full load, the i7-2600S would be hotter than your current E8400, but it would be a lot faster than your current processor. Under normal loads (such as email, a program maxing out one core, etc) I would imagine it would run cooler or at least as cool as your current processor. But again, we won't know until we get our hands on the chips.
Re: Q's about Intel, integrated graphics, TDP, and silence
I am also planning to build a sandy bridge passive system, I will go for one of the 35W CPUs, they should be safe enough. What kind of case are you using for your passive build, do you have any recommendations?silentplummet wrote:Background info: I have a comfortably quiet PC built with a C2D e8400 which I run on passive cooling. At 3.16 GHz and 1.06 v it maxes at 55 C give or take a degree of ambient temperature. I also run an AMD 5770 passively in this system. However C2D is starting to show its age and I'd like to replace it with one of the upcoming Sandy Bridge CPUs.
With regards to the turbo, I always assumed those CPU runs 2.5 GHz all the time and only when the user clicks turbo it goes to 3.5, you know like that turbo button on PCs 15 years ago. Gotta say it would annoy me a little if it would constaly switch to 3.5 whenever I open a youtube video or go to a flash website.
Re: Q's about Intel, integrated graphics, TDP, and silence
Not so much. Basically it's same number, same price.No doubt I'll be paying a premium for the 'S'.
What software are you using?Will I really be getting a performance increase out of these new chips relative to my old faithful C2D?
You said it yourself, some software does, like video encoding, other software doesn't.
It's a little more sophisticated than that. When only one core is loaded the CPU ups the speed from 3.3GHz to 3.7GHz in the i5-2500. If it's more than one the frequency is lower, something like 3.5GHz for two cores.I always assumed those CPU runs 2.5 GHz all the time and only when the user clicks turbo it goes to 3.5, you know like that turbo button on PCs 15 years ago.
Not if the CPU is adequately cooled. What cooler are you using?Second, how will my passive configuration affect the function of the turbo?
The C2D E8400 is probably closer to 45W TDP than 65W. A four core CPU, even if it's in 32 nm will use more power. 65W is the maximum power to cool completely passive.
Bottom line, it should work fine, but CPU might throttle when you run Prime95 or LinPack or other test software, but not using normal software.