What does a 65W i7 processor do that a 95W i7 can't?

[jana519]

Asking as an amateur PC builder. The 35W Skylake i7's can be passive cooled which has a big impact on chassis selection. But I'm not sure what would be the difference for the home user between 65W and 95W other than a small power savings. Are the 65W i7 mainly intended for enterprise customers or is there a reason to buy them as a home user?

 

[IEC]

(On average) lower power consumption, and therefore lower heat production.

However, they also have lower default and boost clocks. Which is just fine for the non-power user, and for OEMs looking for easier-to-cool CPUs.

If you're a power user/gamer, better go for the unlocked K model CPUs.

 

[XavierMace]

(On average) lower power consumption, and therefore lower heat production.

However, they also have lower default and boost clocks. Which is just fine for the non-power user, and for OEMs looking for easier-to-cool CPUs.

If you're a power user/gamer, better go for the unlocked K model CPUs.

If you're overclocking. You can be a gamer and not overclock. The price different between the standard and K model processors on some of the newer SKU's is fairly substantial.

 

[Burpo]

"What does a 65W i7 processor do that a 95W i7 can't?"

Saves on power..

 

[myocardia]

If you're overclocking. You can be a gamer and not overclock.

No, what he said applies, whether or not the person wants to overclock. The i7-6700, for instance, has both a lower base clock and a lower boost clock than the i7-6700k. The 6700 is a 65 watt CPU, while the 6700k is a 91 watt. As you can see, the 91 watt CPU has a base speed that is equal to the maximum boost clock of the non-k CPU. For someone who is needing maximum performance, that is a lot of performance to leave on the table, whether they are overclocking or not.

 

[IEC]

If you're overclocking. You can be a gamer and not overclock. The price different between the standard and K model processors on some of the newer SKU's is fairly substantial.

Yes, but amortize the cost over however many years you plan on keeping the processor, and it will make a lot more sense. You aren't going to get a significant leap in performance for another 5 years if the trend from Sandy Bridge continues...

An i7-6700K is 4C/8T and 4.0GHz/4.2GHz base/turbo + more cache. An i5-6500 is 4C/4T and 3.2GHz/3.6GHz base/turbo. At $350 RCP vs $202 RCP, you are paying $148 more (+73%) for >30% more performance. But amortized over 5 years that's <$30/yr difference for a pretty substantial difference in performance (and that's before overclocking).

 

[ElFenix]

anything can be passively cooled if you're brave enough...

(i maintain that any cooling solution that cannot move heat against a temperature gradient is passive and that computer people got it wrong 15 years ago - slapping a fan on a bunch of fins doesn't make the cooling active - you're just helping nature)

"What does a 65W i7 processor do that a 95W i7 can't?"

Saves on power..

for the same workload i doubt it. every time i've seen low TDP parts tested against regular TDP parts, the lower TDP parts take enough extra time to finish the task to erase any power savings.

 

[Dufus]

A processor wit a 65W TDP can be made to use more than 65W and a processor with a 95W TDP can be limited to use less than 65W. Possibly the 65W TDP processor runs hotter than the 95W TDP processor Watt for Watt.

 

[AtenRa]

for the same workload i doubt it. every time i've seen low TDP parts tested against regular TDP parts, the lower TDP parts take enough extra time to finish the task to erase any power savings.

not always true,

For the same CPU, increasing the performance by raising the clock frequency decrease the perf/watt due to higher voltage needed.
So it may finish the workload faster but it will consume more power in order to do so.

 

[AtenRa]

A processor wit a 65W TDP can be made to use more than 65W and a processor with a 95W TDP can be limited to use less than 65W. Possibly the 65W TDP processor runs hotter than the 95W TDP processor Watt for Watt.

Didnt get that, could you elaborate ??

 

[coercitiv]

not always true,

For the same CPU, increasing the performance by raising the clock frequency decrease the perf/watt due to higher voltage needed.
So it may finish the workload faster but it will consume more power in order to do so.

Indeed, not always true. However, a certain ratio between idle system power usage and load CPU power usage will lead to surprising results. Not sure which is more likely in consumer products though.

Didnt get that, could you elaborate ??

Nowadays we have a number of options when it comes to limiting CPU power usage, including configurable TDP. A combination of system settings and workloads can make a 65W TDP CPU to use more than that (because it's limited bellow it's true wattage potential to begin with).

Also, a 95W TDP CPU is likely to be a better bin of it's 65W counterpart, hence somewhat better characteristics when running at the same TDP. However, this is an artificial finding, likely dictated by market conditions.

 

[XavierMace]

No, what he said applies, whether or not the person wants to overclock. The i7-6700, for instance, has both a lower base clock and a lower boost clock than the i7-6700k. The 6700 is a 65 watt CPU, while the 6700k is a 91 watt. As you can see, the 91 watt CPU has a base speed that is equal to the maximum boost clock of the non-k CPU. For someone who is needing maximum performance, that is a lot of performance to leave on the table, whether they are overclocking or not.

I'm aware of the clock speed differences (600mhz), but the question is, is that actually a noticeable difference to the user? Enough of a difference to justify the $80 price difference as well as the TDP increase? Maximum performance is an entirely vague goal post. If you truly need MAXIMUM performance, you're looking at the wrong processor entirely, K or not. For 90% of users, I highly doubt they are going to notice that clock speed difference.

An i7-6700K is 4C/8T and 4.0GHz/4.2GHz base/turbo + more cache. An i5-6500 is 4C/4T and 3.2GHz/3.6GHz base/turbo. At $350 RCP vs $202 RCP, you are paying $148 more (+73%) for >30% more performance. But amortized over 5 years that's <$30/yr difference for a pretty substantial difference in performance (and that's before overclocking).

You're talking two different models of processors. Apple and oranges. I'm specifically talking the difference between the K and non-K models of the same processor. The price difference between the 6700 and the 6700K is $80.

I think we also all know a 30% clock speed increase doesn't necessarily mean a 30% increase in performance when it comes to everyday tasks.

 

[coercitiv]

Maximum performance is an entirely vague goal post. If you truly need MAXIMUM performance, you're looking at the wrong processor entirely, K or not. For 90% of users, I highly doubt they are going to notice that clock speed difference.

I'm specifically talking the difference between the K and non-K models of the same processor. The price difference between the 6700 and the 6700K is $80.

I think we also all know a 30% clock speed increase doesn't necessarily mean a 30% increase in performance when it comes to everyday tasks.

You already made up your mind about what users need their computer for, which is a very slippery step towards completely missing the point. The "need for performance" is always a scenario based requirement: you are guided by a variable mix of ST, MT performance, budget and prices.

For example, what if I told you that in my country the difference between 6700 and 6700K is only $40?
(illustrative purpose only, but also happens to be true in my case)

 

[Yuriman]

For those interested in performance per watt vs clockspeed:

http://forums.anandtech.com/showthread.php?t=2443083&highlight=

On 22nm, isolating the CPU, performance per watt continues to improve as clockspeed drops, all the way to 1600mhz (the minimum I was able to test). Taking whole-system power consumption into consideration, performance per watt peaks at 2800-3200mhz, and begins to decline sharply past 3600mhz.

However, if you assume bursty workloads, and tend to leave your computer on all day, you're actually going to have the best performance per watt at minimum clocks anyway, because the PC isn't going into a sleep state after finishing its work and thus fixed power costs will continue to exist. Averaging these in, you end up with lower performance per watt the higher your clocks are.

 

[sm625]

For those interested in performance per watt vs clockspeed:

http://forums.anandtech.com/showthread.php?t=2443083&highlight=

That is a really good thread. I was thinking about that thread while reading this one. It would be nice to have similar data for skylake.

 

[XavierMace]

You already made up your mind about what users need their computer for, which is a very slippery step towards completely missing the point. The "need for performance" is always a scenario based requirement: you are guided by a variable mix of ST, MT performance, budget and prices.

For example, what if I told you that in my country the difference between 6700 and 6700K is only $40?
(illustrative purpose only, but also happens to be true in my case)

No I didn't. I simply stated you can be a power user/gamer and go with the non-K model. All the OP told us about his usage is he's a home user and amateur builder. He's given ZERO indication that he's got performance concerns/needs, yet everybody seems to be arguing that he should be going with the K. Without more info, that's just silly.

 

[Dufus]

For example, what if I told you that in my country the difference between 6700 and 6700K is only $40?

Where I live it's less than USD30. As good as they are still not buying one though.

On 22nm, isolating the CPU, performance per watt continues to improve as clockspeed drops, all the way to 1600mhz (the minimum I was able to test).

JFYI even though the LFM is 1600MHz you'll probably find it goes done to 800MHz, at least my 3770k did anyway.

 

[coercitiv]

He's given ZERO indication that he's got performance concerns/needs, yet everybody seems to be arguing that he should be going with the K. Without more info, that's just silly.

All I see is people saying there is a performance difference between the two models. I see no clear purchase recommendation, moreover I see no purchase inquiry either. The OP only asked if there's more to it than just TDP difference.

PS: also in my opinion, when lacking specific info on user needs, i7 should be out of the question altogether.

 

[coercitiv]

Where I live it's less than USD30. As good as they are still not buying one though.

Better hang on to that BCLK BIOS then!

 

[Dufus]

I meant buying Skylake in general.

 

[ElFenix]

not always true,

For the same CPU, increasing the performance by raising the clock frequency decrease the perf/watt due to higher voltage needed.
So it may finish the workload faster but it will consume more power in order to do so.

6700 vs 6700k test spcr found .1 watt-hour difference in energy consumption between the two when running their benchmark suite. As a data center, you might care about that. As an average user, you don't. The 6700k idled at 10% higher consumption, which seems odd to me. Considering that factors into their math, that may be up to sample variation.

http://www.silentpcreview.com/article1481-page5.html

 

[Yuriman]

6700 vs 6700k test spcr found .1 watt-hour difference in energy consumption between the two when running their benchmark suite. As a data center, you might care about that. As an average user, you don't. The 6700k idled at 10% higher consumption, which seems odd to me. Considering that factors into their math, that may be up to sample variation.

http://www.silentpcreview.com/article1481-page5.html

Possible, but it may also be the case that the "K" chips are binned for higher leakage.

 

[coercitiv]

The 6700k idled at 10% higher consumption, which seems odd to me. Considering that factors into their math, that may be up to sample variation.

In the IGP only test the idle power difference was 1W.

 

[ElFenix]

In the IGP only test the idle power difference was 1W.

Interesting that in one place it's a 4 watt difference and in another it's a 1 watt difference

 

[Yuriman]

It would be nice if they included the error in their measurements. I didn't in my tests, and in retrospect I probably should have. Idle power fluctuates, maybe 2-4w.