News Intel's new 35W CPUs aren't much slower than their 65W counterparts — Core i5-14600T exhibits 6% lower performance than Core i5-14600 in Geekbench...

[Admin]

A Geekbench 6 result for Intel's newly released Core i5-14600T shows that it's hardly behind the Core i5-14600.

Intel's new 35W CPUs aren't much slower than their 65W counterparts — Core i5-14600T exhibits 6% lower performance than Core i5-14600 in Geekbench... : Read more

 

[nogames]

So you're saying, the 65W counterparts are not very fast 😉

 

[ezst036]

I like 35w chips, they're very easy to turn into fan-less operation.

Give it a cooler that's rated for like 180w or some seemingly absurdly rated number, remove the fans, and keep right on going. Your case choice can't be a total after-thought, but you can generally do it with ease.

Those fanless PSU like the seasonic titanium put out virtually no heat at all.

This is not a gaming rig.

 

[CelicaGT]

More proof the standard components are too far outside the efficiency curve. Perf/Watt really suffers on the high end and most gamers would probably be just fine with the lower TDP option especially at higher resolutions. (Or just set a reasonable frame limiter and stop bouncing off the power limit on either component.)

(Edit: spelling)

 

[healthy Pro-teen]

More proof the standard components are too far outside the efficiency curve. Perf/Watt really suffers on the high end and most gamers would probably be just fine with the lower TDP option especially at higher resolutions. (Or just set a reasonable frame limiter and stop bouncing off the power limit on either component.)

(Edit: spelling)

With some tinkering, 65 watt CPUs can be easily made into 35W CPUs. Before overclocking was worth it, now underclocking or UV is. If we are willing to give up on a few fps.

 

[usertests]

14600T turbo TDP = 92W.

 

[TerryLaze]

14600T turbo TDP = 92W.

Non-t 14600 turbo TDP=154 ,what's your point?!
It's still more than 50% more power than the non-t uses.
(At full load)

 

[TerryLaze]

More proof the standard components are too far outside the efficiency curve.

No they are not, at least not if you use them as intended, if you do overclock them to infinity then yes.
Stock/default settings 14900k = 139 points 7950x = 158 points, that's a """huge""" difference of ~13% !
And that is only if you are doing any rendering or other full load job on the system.
That might not be on the peak of the efficiency curve but it is definitely not "too far" off.
https://www.techpowerup.com/review/...ke-tested-at-power-limits-down-to-35-w/8.html

 

[usertests]

Non-t 14600 turbo TDP=154 ,what's your point?!
It's still more than 50% more power than the non-t uses.
(At full load)

The efficiency curve. If it was actually 35W vs. 65W where the power really counts, there might be a bigger difference. By the time you get to 92-154 W there are diminishing returns.

 

[JTWrenn]

This is an unfortunate consequence of the who has the fastest CPU race. Efficiency should be our top line question nowadays but people gotta go full v12 turbo idiot on it all.

 

[TerryLaze]

The efficiency curve. If it was actually 35W vs. 65W where the power really counts, there might be a bigger difference. By the time you get to 92-154 W there are diminishing returns.

That's why base power is base power and max turbo power is max turbo power.
Most of the time you will be around base power even with moderate to high workloads. You only hit max turbo power if you run all of the cores at all of the clocks at all of the time running the heaviest of workloads.

 

[TerryLaze]

This is an unfortunate consequence of the who has the fastest CPU race. Efficiency should be our top line question nowadays but people gotta go full v12 turbo idiot on it all.

Wait! The fact that intel is making t models that use far less power is an unfortunate consequence of the who has the fastest CPU race?

 

[CelicaGT]

No they are not, at least not if you use them as intended, if you do overclock them to infinity then yes.
Stock/default settings 14900k = 139 points 7950x = 158 points, that's a """huge""" difference of ~13% !
And that is only if you are doing any rendering or other full load job on the system.
That might not be on the peak of the efficiency curve but it is definitely not "too far" off.
https://www.techpowerup.com/review/...ke-tested-at-power-limits-down-to-35-w/8.html

Kinda what I was getting at in a roundabout way with setting a frame limiter when gaming. All core loads are a different beast, and one I'm not terribly familiar with. The chart you posted is both enlightening and clear. That said I'd like to see something like this comparing the standard, "T", and "K" variants of the same die at stock settings. I think it would be quite interesting.

 

[Amdlova]

I have the T family here 😉 I only give for the 13500T 105W and for the 13100T 65w max 😉
don't heat... Can use passive heat sink 😀

 

[thestryker]

More proof the standard components are too far outside the efficiency curve. Perf/Watt really suffers on the high end and most gamers would probably be just fine with the lower TDP option especially at higher resolutions. (Or just set a reasonable frame limiter and stop bouncing off the power limit on either component.)

(Edit: spelling)

This is what makes the 7800X3D (and really all of the Zen 4 X3D parts) so good efficiency wise. They couldn't blow out the efficiency curve with the additional cache attached which keeps the CPUs at about peak efficiency.

It was really interesting seeing Buildzoid do some power testing on Zen 4 and RPL (may have been ADL) because Zen 4 is insanely efficient up to around 150W or so (I think it was 7950X) and then the power cost for performance starts flattening out to a point where it's basically not worth it at about 200W (stock 7950X is temp limited so will go to ~235W if you have the cooling).

Intel on the other hand flattens to a certain point on the curve and then the power/perf scaling is linear. I forget where his cooler wasn't good enough anymore, but it was into the 300s.

I don't think either company has any interest in giving an inch though. Their power controls have gotten a lot better so I think they rely on that to cover the lighter workloads. I'd rather they went back to some sort of temporary boost situation (on Intel's side some SKUs do still have temporary boost). Maybe in the case of Intel adding a PL3 so as an example you'd have 13900K/14900K with 125W PL1, 200W PL2 and 253W PL3 which was a timed boost.

 

[General_Cool]

I suspect this will be incredibly popular in countries where its 50 cents a kilowatt for electricity...rare Intel W. There is a great case to be made here for folks to save money and get basically unnoticeably worse performance, and I think the vast majority of people in the i5 space are not interested in insane overclocks more than they are purchasing and running costs...

Low TDP chips are the future of computing with the world's transition to climate friendly power consumption and efficiency policies and insane inflation. There could be a point in time where some countries limit the amount of electricity to each household for one reason or another (not gonna go into the politics of why this is insane!), and low consumption PCs with excellent performance could allow consumers to use their PCs for longer, reducing e-waste. Overall this is awesome. I hope we see AMD and Intel competing for the highest TDP per watt soon instead of just fighting over who's chip can double as a 300 watt space heater...

Speaking of the red team, no 7000 series GE chips in sight...not a good look AMD.

 

[vehekos]

It is depressing that it takes 6 P-cores, 8 E-cores, and 20 threads to multiply a single core performance by less than 6x.

 

[bit_user]

https://www.techpowerup.com/review/...ke-tested-at-power-limits-down-to-35-w/8.html

That's a i9, though. The i5-14600(T) has half as many E-cores and 3/4ths as many P-cores. They won't look nearly as good on efficiency at 35 W and 65 W. Not to mention that normal operation of those CPUs would allow them to turbo-boost (unlike the power-limited entries in that chart), which really hurts efficiency.

The standard formula for high-efficiency is to have a lot of cores and clock them low. With a smaller number of cores, efficiency will suffer.

 

[bit_user]

It is depressing that it takes 6 P-cores, 8 E-cores, and 20 threads to multiply a single core performance by less than 6x.

I wouldn't read too much into Geekbench scores. If you want to see how well these CPUs can scale, look at a program that's properly designed to use multiple threads, like Cinebench or Blender.

 

[bit_user]

You only hit max turbo power if you run all of the cores at all of the clocks at all of the time running the heaviest of workloads.

That's BS. A single P-core can easily use 35 W. Scale that up by 8 and you're already at 280 W. That's before factoring in hyperthreading or touching the E-cores.

​

Therefore, even a moderately multithreaded app should be able to max the i9's 253 W PL2.

 

[JTWrenn]

Wait! The fact that intel is making t models that use far less power is an unfortunate consequence of the who has the fastest CPU race?

Oig, no, the fact that they make a cpu that uses double the power to get 6% more performance is.

 

[dalek1234]

So you're saying, the 65W counterparts are not very fast 😉

Exactly.

Typical THG. They spinned a negative into a positive. They title should have said: "By increasing the power by ~86%, you get only 5% increase in performance." Intel must be desperate

 

[TerryLaze]

That's BS. A single P-core can easily use 35 W. Scale that up by 8 and you're already at 280 W. That's before factoring in hyperthreading or touching the E-cores.

BS is that you think that I meant that 35W is the max turbo power....do you even read?!
This is my whole post,

That's why base power is base power and max turbo power is max turbo power.
Most of the time you will be around base power even with moderate to high workloads. You only hit max turbo power if you run all of the cores at all of the clocks at all of the time running the heaviest of workloads.

You explained it perfectly, you need all cores to run at full speed, exactly the same thing that I said, to get at 280-300 ++ watts, that's max turbo power and not base power.

 

[TerryLaze]

That's a i9, though. The i5-14600(T) has half as many E-cores and 3/4ths as many P-cores. They won't look nearly as good on efficiency at 35 W and 65 W. Not to mention that normal operation of those CPUs would allow them to turbo-boost (unlike the power-limited entries in that chart), which really hurts efficiency.

The standard formula for high-efficiency is to have a lot of cores and clock them low. With a smaller number of cores, efficiency will suffer.

So "standard component" for you means only up to i5?

But also i5-14600k, stock or power limits removed = pretty much the same as the 7950x in efficiency, it does a little better than the 14900k.

 

[TerryLaze]

Intel on the other hand flattens to a certain point on the curve and then the power/perf scaling is linear. I forget where his cooler wasn't good enough anymore, but it was into the 300s.

I don't think either company has any interest in giving an inch though. Their power controls have gotten a lot better so I think they rely on that to cover the lighter workloads. I'd rather they went back to some sort of temporary boost situation (on Intel's side some SKUs do still have temporary boost). Maybe in the case of Intel adding a PL3 so as an example you'd have 13900K/14900K with 125W PL1, 200W PL2 and 253W PL3 which was a timed boost.

It is a timed turbo right now on intel (and on AMD the same way) , timed by how fast your cooler gets saturated and cuts off the CPU.
Just as your example from buildzoid, if you have a cooler that cools more than 300W 24/7 then it will boost indefinitely.
If you have a 200W cooler it will boost to 253W (or whatever you have set as max limit) only for a few seconds under heavy load and much longer on light loads, until it will have to go down to let the cooler cool off again.