Intel Core 2 Quad Power and Thermal Performance Vs Modern CPU

[ttran7701]

Modern CPU's will destroy Intel's Core 2 Quad in terms of performance, that's a given. How do they compare in terms of power usage and thermal performance? The Core 2 Quad uses an old build process but is clocked much lower than say a i7 8700K. The Core 2 Quad also have two fewer cores. Would the lower clock and fewer core count make it run cooler than the i7 despite the bigger build process? I'm just curious how things have or haven't changed in the terms of power and thermal.

 

[Eximo]

Well, the short answer is, it depends...

Now for the long answer, might want some popcorn.

So, number one, Intel targets specific TDP ranges. Their highest end consumer chips have been rated at 77-95W for many generations now. With recent peaks up to 91W. Measured power draw on the other hand, is typically higher.

As you mentioned process node shrinks have been the primary driver towards greater efficiency. At each shrink you can reduce the voltage, which you have to do to avoid capacitance effects in the now tinier transistors. The greater density means a greater number of chips per wafer, so more money in Intel's pocket as long as yields are good.

In time with the process node shrinks, Intel has been increasing the area dedicated to graphics. So chip size has not really changed linearly with node shrinks. Now we have the addition of processing cores further increasing the size of the chip.

Then there is the whole soldering vs paste. With 3rd gen hardware consumer chips switched to using thermal compound instead of solder. Power consumption was low enough that they could save the cost of doing that. Now we are back to solder on the new 9th gen series. However, this came with a huge increase in silicon thickness. Silicon isn't as effective as say, copper, at heat transfer. Solder is better than thermal paste, but it was more of a zero gain, or possibly worse, with the increase in chip thickness. 8th gen runs as cool or cooler than 9th gen with paste.

Intel did it, not to please consumers(though the marketing team is no doubt pleased to have a new feature to offer), but because they have been stuck on 14nm and have just been pushing the clock speeds higher and higher. They know this isn't viable long term, so they are pulling out all the tricks to make these things work. The silicon thickness increase is probably necessary for the soldering, as fragility was the reason they said they switched to thermal compound (Which they eventually ended up doing on HEDT as well, so they could have been being truthful)

If you look at the non-overclocked CPUs from say Haswell to Kabylake, they are hands down cooler and more power efficient than older chips. Coffelake locked processors fall in there as well, but their higher end 6-core parts not as much since they can do more work. Also not terrifically fair to the core 2 quads since their processing power would be that much greater and skew the numbers. Would be more fair to take say a first gen i7 with six cores (i7-960+) and compare that to an i7-8700k.

If you assume a constant power, a bit false, and make a simple chart using something like passmark scores, you can see the efficiency gains over time.

 

[ttran7701]

Thank you for the great reply. I would be curious what the Intel Core 2 would do in terms of temperature under stress test like Cinebench. Would it score 80 degrees vs 90 degrees with a i7?? Again, the i7 would be cranking through the benchmark and wiping the floor with the Core 2. Could it do it with the same or higher running temperature?

 

[Eximo]

That is a somewhat different scenario and it would depend on the cooling available. Like for like, at full blast the Core 2 would run hotter and accomplish less. Once they introduced turbo boost, you would be comparing CPUs at throttled performance to CPUs that only had one speed, so clock speeds might end up similar.

A more fair comparison would be between something like a Core2 Quad 8400 or Core2 Quad 9650 and a modern i5-7600 or i3-8300. That would be:
Core2 Quad 9650: 95W = 4177 multithread
i5-7600 65W = 8766 multithread

So about twice as powerful and three times as efficient.

I forgot to add that there are also architecture and instruction set improvements over time. More efficient ways of programming the chip, making purpose built circuits to perform certain tasks, etc. Optimizations in the scheduler and prediction can also make a big difference in computation performance.

 

[ttran7701]

Wow! So an old Core 2 could conceivably run as hot as an i9 9900K?!? Again, thanks for all your replies. Great info!!

 

[CompuTronix]

So an old Core 2 could conceivably run as hot as an i9 9900K?!?

Guys,

I'd like to weigh in on this, as I think the question is very interesting.

As the author of the Intel Temperature Guide http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html - I've devoted nearly 12 years and 6,500 hours focused on researching, testing, analyzing and comparing thermal performance among Intel processors; the results of which I've shared in my guide for the benefit of the community. Except for the 6 Core i7 8700K, here's a relevant list of Quad Core processors I've personally owned, thermally tested and used to build and overclock various rigs:

Q6600 G0
Q6700 G0
Q9650 E0
i7 920 D0
i5 2500K
i7 2700K
i5 3570K
i7 3770K
i7 4770K
i5 4690K
i7 4790K
i7 6700K
i7 7700K
i7 8700K

Although Eaximo has offered some excellent insights, there's a few additional factors and variables to be considered which will provide a different overall perspective.

Regarding the original question, keep in mind that back in the day of the Core 2 processors, today's cooling solutions weren't available. However, we can closely compare similar Quad Core processors by minimizing as many variables as possible, and by utilizing the same test conditions.

Group 1 - Core 2 Quad vs i5 (4 Cores / 4 Threads):

95 Watts TDP
Soldered IHS
65, 45 & 32 Nanometers
2018 Mid-Range Air Cooling
Maximum recommended Vcore
Typical Best Overclock
Prime95 v26.6 Small FFT's
22°C Ambient temperature
Case covers removed
All Fans @ 100% RPM

Q6600 GO @ 3.6 GHz / 1.500 Vcore
Q9650 O/C @ 4.2 GHz / 1.400 Vcore
i5 2500K O/C @ 4.6 GHz / 1.375 Vcore

Note: The Q9650 overclocked @ 4.2 GHz earns a Passmark CPU score on par with the i5 2500K at stock.

(1) Overclocks and Core voltages are typical and appropriate per microarchitectures.
(2) Overclocked measured Package Power is ~110 to 115 Watts.
(3) Resulting Core temperatures are 75 to 80°C.

Here's the maximum recommended Core voltages per Microarchitecture from 14 to 65 nanometers since 2006:


Group 2 - Quad Core i7's (4 Cores / 8 Threads):

77, 84, 88 & 91 Watts TDP
Delidded IHS (Liquid Metal)
22 & 14 Nanometers
2018 Mid-High End Air Cooling
Maximum recommended Vcore
Typical Best Overclock
Prime95 v26.6 Small FFT's
22°C Ambient temperature
Case covers removed
All Fans @ 100% RPM

i7 3770K O/C @ 4.5 GHz / 1.300 Vcore
i7 4770K O/C @ 4.6 GHz / 1.280 Vcore
i7 4790K O/C @ 4.7 GHz / 1.280 Vcore
i7 6700K O/C @ 4.8 GHz / 1.400 Vcore
i7 7700K O/C @ 5.0 GHz / 1.375 Vcore

(1) Overclocks and Core voltages are typical and appropriate per microarchitectures.
(2) Overclocked measured Package Power is ~115 to 125 Watts.
(3) Resulting Core temperatures are 75 to 80°C.

The principal differences between the two Groups of Quad Core processors which actually affect raw test results are; number of Threads, soldered vs delidded and a better class of air cooler due to hyperthreading.

Using the same test conditions, the i7 8700K (6 Cores / 12 Threads) delidded and overclocked @ 5.0 GHz / 1.35 Vcore on big air or a mid-range 240 / 280 mm AIO runs ~150 Watts.

The soldered 9x00 8 Core processors overclocked @ 5.0 GHz / 1.35 Vcore can exceed 200 Watts, so cooling requirements increase yet again; a high-end AIO or a custom loop is required. As informative as the reviews may be, it's too soon to draw concrete conclusions, as there's little users feedback just yet.

It's been my experience that when variables are minimized, thermal performance between Core 2 Quads and delidded i5's are relatively comparable. I think the last thermally consistent, predictable and well behaved processor family was 2nd Generation Sandy Bridge.

CT

 

[ttran7701]

Wow! Thanks for great post. Good read!

 

[Eximo]

I think it is somewhat clear that the temperature between CPUs would remain basically consistent, but the added factor of performance was what really made the question interesting. I'd not looked at the numbers in that perspective and was intrigued to see the actual level of improvement.

It is interesting to see that Ivy Bridge really was the peak of low power, then Intel started pushing clock speed to keep up their annual performance improvements.

Electronics typically don't like running over 100C and any system designed around that is going to keep temperatures in check.

I disagree that the coolers we have today are hugely different then what was available 10 years or so ago. I had some beefy coolers even then. The really big ones did appear with Intel's LGA1366 and AMDs AM2.

Though I wasn't taking overclocking into account, either, as that comes down to more golden sample effects.

A proper study of this would require some serious data collection and testing specifically for a few factors:
Die size
Cooling interface
Temperature at a fixed voltage
A consistent test across many generations of CPU
And if possible, a single cooler style.

Pretty sure you could get away with that last one from LGA775->LGA115x

Would be even more interesting to do with mobile processors, though they would be trickier since their boost ranges are so extreme. Would also be a mess to apply the same cooler to all of them.