Comparing Performance and Power Use For Vista vs. Windows 7 WIth Clarksfield Chi

crazipper writes "Back when Intel launched its Core i5/i7 'Lynnfield' CPUs, Tom's Hardware ran some tests in Windows 7 versus Vista to gauge the benefits of the core parking and ideal core optimizations, said to cut power consumption in the new OS. It turned out that Win7 shifted the Nehalem-based CPUs in and out of Turbo Boost mode faster, resulting in higher power draw under load, while idle power was a slight bit lower. The mobile version of the architecture was claimed (at the time) to show a greater improvement in moving to Win7. Today there's a follow-up with the flagship Clarksfield processor that shows the same aggressive P-state promotion policies giving Win7 a significant performance advantage with Core i7 Mobile. However, power consumption is higher as well."

Comments

But what about ECC?

[Timothy Brownawell]

Given the recent google study and the Folding@Home NVIDIA study, why would you want to run an i5/i7 system (which don't permit ECC)?

On what desktop system do you use ECC?

[Sycraft-fu]

I haven't seen a desktop in a long time that had ECC RAM, or even support for it. In the Core 2 era of chips desktop use normal unbuffered DDR2 or DDR3 DIMMs. For ECC stuff on workstations/servers you use FBDIMMs which are way more expensive.

Same shit with the i7. If you want i7 class hardware with ECC it is called the Xeon 5500. Running on a 5520 chipset, it supports ECC RAM, and lots of it (144GB is the most I've seen thus far).

That's all workstation class stuff. Desktop stuff is not ECC because it is cheaper.

Re:On what desktop system do you use ECC?

[Timothy Brownawell]

I haven't seen a desktop in a long time that had ECC RAM, or even support for it.

Any moderately recent AMD CPU will support ECC, and it's not hard to find a mainboard that does as well (for example I believe any ASUS mainboard for AMD will support ECC, I know the one I checked a couple days ago does (cheapest ASUS AM3 mainboard on Newegg then, probably still is, only like $5 more than the cheapest other AM3 board)).

In the Core 2 era of chips desktop use normal unbuffered DDR2 or DDR3 DIMMs.

Buffered/unbuffered is separate from ECC/non-ECC. For example I know the AMD desktop chips support unbuffered ECC memory.

Desktop stuff is not ECC because it is cheaper.

Maybe 10% cheaper. And of course it's easy to make things cheaper if they don't have to work correctly.

Re:On what desktop system do you use ECC?

[DigiShaman]

ECC support is more expensive on Intel platforms because Intel segments the market intentionally!!! If you want ECC support running an Intel rig, you will need both a Xeon CPU and a Workstation class board (even thought the new FSB desktop chipset supports ECC, it's not enabled)

FYI, I'm actually building a new desktop machine this week. It's an AMD Phenom II paired up with Asus Crosshair III board. I purchased matched ECC DDR3 memory (non-buffered and non-registered) direct from Crucial.com. Simple really. I just pulled down the make/model board and placed the order in two shakes... I could be wrong, but this might be the cheapest desktop class system that supports ECC DDR3.

Re:On what desktop system do you use ECC?

[Runaway1956]

"Intel segments the market intentionally!"

Don't forget virtualization. With AMD, you don't have to pay a premium if you plan to run virtual machines.

Good grief

[Anonymous Brave Guy]

I've built several high-end PCs from scratch and spec'd several more at component level, during a period of well over a decade and most recently just a couple of years ago, and I still have absolutely no idea what any of the fine summary meant.

Does anyone actually label/number components in any sort of logical way at all any more? Codename this, year that, version.subversion.minorversion.veryminorversion the other (revision C17, of course; the C16s and B17s didn't have the double overclocked doobreeflips in the L7 cache).

It's a wonder anyone can build a PC that runs at any speed at all any more. Sheesh.

Re:Good grief

[crazipper]

Quick summary:
Lynnfield = the internal name for the new Core i5/Core i7 CPUs for LGA 1156.
Core parking/ideal core = two optimizations from Microsoft in Win7 that are supposed to save power by consolidating background tasks onto as few CPU cores as possible, and then putting the idle cores to sleep.
Clarksfield = Core i7 Mobile; basically, the Lynnfield stuff with a different interface, more aggressive Turbo Boost, etc.
Nehalem = Another Intel internal name referring to the whole family of 45nm CPUs based on this architecture. Members include Bloomfield, Lynnfield, Clarksfield.

Isn't this what we want?

[Korin43]

Isn't this what we want? I mean, it's higher power under load because it switches to "fast mode" faster. Isn't that good? Yes it uses more power, but if the goal was to use as little power as possible, we'd just lock the processor in "slow mode".

Re:Isn't this what we want?

[SpelledBackwards]

Exactly. People seem to forget that power drain and energy consumption are not the same thing - power consumption is in energy consumed per some amount of time. For a completely unrelated example: If you run a 30 W load over 1 second, it will use 30 joules of energy (because a 1 W power draw means it consumes 1 J per second). But if you run a 500 W load over 1/100 sec, you'll only use 5 J of energy. Batteries store energy, not power, so what is likely to be more important for mobile platforms is which one used the least amount of energy over the time span of the test, not comparing peak power or power in short bursts of activity. That is, if you're concerned about battery life. Peak power might play a bigger role in talking about current load and CPU/battery temperature issues.

Re:MacBook Pro

[nxtw]

Isn't that what Apple introduced earlier this year on the MacBook Pros? The ability to switch off the high power GPU when it's not needed and fall back to a lower quality integrated GPU? I realize that Apple used an nVidia solution instead of an Intel, but that still seems a little disingenuous.

The Apple GPU switching implementation appears to require the user to restart his or her session (that is, log off and log on again.) Intel's implementation seems to support switching GPUs without logging off or restarting. The Intel solution also has to handle two different display drivers.

Some older laptops supported switching between integrated and discrete graphics as well, but I think they required a reboot to switch.

How about using XP sp3 for comparison???

[voss]

I cant be the only one who might think xp sp3 might actually win

Re:How about using XP sp3 for comparison???

[selven]

Windows 98 might win.

Powering the chipset and backlight

[tepples]

while under battery power the CPU will do everything it can to conserve power under the same software load conditions.

In many notebooks, the CPU does not dominate battery consumption; the northbridge, southbridge, and LCD backlight draw a significant fraction of the power. So when CPU usage hits 90%, clocking it up to full power is warranted because it gets the work done faster, meaning that the chipset and LCD don't run as long while the user is waiting for the CPU to finish.

Situations like this are why I run Intel chips

[dave562]

I primarily use Microsoft software (I know, get out the pitchforks) and over the years I have occasionally run AMD chips after being overcome by various AMD biased friends of mine. I've never been able to put my finger on it, but Windows simply doesn't run as well on AMD chips as it does on Intel chips. I always end up switching back to Intel. This article is just an example of why. Intel and Microsoft are in bed with each other, and Microsoft will always be putting out the code to take full advantage of the Intel chips. It wouldn't surprise me if Intel gives Microsoft the heads up on new features far in advance. It wouldn't surprise me if Microsoft works with Intel and encourages them to develop certain features in their processors that will help the Microsoft code base execute faster.

Re:Situations like this are why I run Intel chips

[dave562]

What is flame bait about my post? Intel and Microsoft work closely together to optimize the user experience. Must be AMD fans with mod points today.

Re:Situations like this are why I run Intel chips

[Anonymous Coward]

Riiight.

Certainly Intel and Microsoft work closely together, they have many reasons to. But I've used many AMD and Intel systems, and honestly they're pretty interchangeable in terms of user experience.

Claims that Windows only runs right with Intel is at best, inaccurate. Are you forgetting things like the adoption of the AMD64 architecture as The Way Forward for Microsoft in terms of 64bit support, over Intel's offerings..

Re:Windows Update

[Canberra Bob]

I am no Linux fanboi by any stretch of the imagination however I have to agree with the parent. In my personal experience, regardless of hardware configuration, even a brand spanking new build will slow to a dead crawl (for all intents and purposes unuseable) when performing updates.

Re:MacBook Pro

[PopeRatzo]

So the worst thing you can say about Win7 is that it performs better but uses slightly more power in some rigs?

Desperation sets in...