News Intel Core i5-10400 Benchmarked: Comet Lake CPU Edges Out i5-9400F

[Admin]

Intel's 10th Gen Comet Lake Core i5-10400 CPU was benchmarked by a forum user.

Intel Core i5-10400 Benchmarked: Comet Lake CPU Edges Out i5-9400F : Read more

 

[NightHawkRMX]

I mean, what did you expect. It has double the threads and a higher boost clock.

TDP makes no sense. If the CPUs TDP is measured at the base clock, and both of these CPUs have 2.9ghz base, you would think the one with 12 threads would have a higher base than the 6 thread one, not a lower TDP.

Just think of this new CPU as an 8700 with slightly nerfed clocks.

 

[InvalidError]

TDP makes no sense. If the CPUs TDP is measured at the base clock, and both of these CPUs have 2.9ghz base, you would think the one with 12 threads would have a higher base than the 6 thread one, not a lower TDP.

CPUs are usually far below their TDP while operating at base clock, Intel can afford turning SMT on without bothering to change the TDP. BTW, the i7-8700 is a 65W part too despite being one or two plusses behind. Would make sense that a slightly updated part on a more mature process that runs 300MHz lower base+boost would use at least somewhat less power.

 

[bit_user]

TDP makes no sense. If the CPUs TDP is measured at the base clock, and both of these CPUs have 2.9ghz base, you would think the one with 12 threads would have a higher base than the 6 thread one, not a lower TDP.

Just think of this new CPU as an 8700 with slightly nerfed clocks.

If it's made on a 10-core or even 8-core die, then they can pick the best 6 cores and disable the rest. So, even if the cores are the same (and they're probably not), simply having more cores on die makes a difference.

 

[bit_user]

CPUs are usually far below their TDP while operating at base clock, Intel can afford turning SMT on without bothering to change the TDP.

According to https://www.anandtech.com/show/1458...ng-an-interview-with-intel-fellow-guy-therien :

TDP is an average power dissipation at a maximum junction temperature operating condition limit, specified in Intel’s engineering datasheet for that processor, for which the processor is validated during manufacturing and when executing an associated Intel-specified high complexity workload at that frequency. What that means is when we quote a base frequency, we think about a worst case environment and a real world high-complexity workload that a user would put on the platform – when the part is run at a certain temperature, we promise that every part you will get will achieve that base frequency within the TDP power.

This is quoted (10 months ago) straight from the guy at Intel who actually defined what TDP means.

Based on that, there shouldn't have been enough headroom in the Coffee Lake-R spec to simply switch on hyperthreading without an impact on TDP and/or base frequency. So, either they changed their definition of TDP, or something else (e.g. cherry-picking cores) enabled them to run more efficiently at the same base frequency.

 

[mdd1963]

The new '8700'... for socket 1200....

Will be interesting to see it compared to MAD's new 3300X, which was VERY darn strong in gaming results...

 

[InvalidError]

Based on that, there shouldn't have been enough headroom in the Coffee Lake-R spec to simply switch on hyperthreading without an impact on TDP and/or base frequency.

The part that you highlighted effectively says that only the worst dies will get to TDP at base clock under the worst case conditions that are still within specs. Very few people are unfortunate enough to get the worst possible die that can still pass Intel's QA and very few people will also run those chips on the worst possible everything else that still meets specs. For any die that isn't a worst case and isn't operating under worst-case conditions, base clock should be achievable at significantly less than TDP.

 

[bit_user]

The part that you highlighted effectively says that only the worst dies will get to TDP at base clock under the worst case conditions that are still within specs. Very few people are unfortunate enough to get the worst possible die that can still pass Intel's QA and very few people will also run those chips on the worst possible everything else that still meets specs. For any die that isn't a worst case and isn't operating under worst-case conditions, base clock should be achievable at significantly less than TDP.

From the sound of it, their QA is to ensure that they can reach base clock, on a specified workload, at the specified temperature. So, it sounds like you're saying they simply relaxed their QA standards?

Call me skeptical. Because, there would be enough of those chips floating around for someone to get one who has a twitter and/or youtube account, and they could raise a decent fuss.

Remember the Der8auer guy that did a survey of Ryzen 3000 dies and found a worryingly large number that didn't hit the rated clocks? I doubt Intel wants to risk that kind of negative publicity.

 

[bit_user]

Will be interesting to see it compared to MAD's AMD's new 3300X,

Fixed that, for you.

 

[InvalidError]

From the sound of it, their QA is to ensure that they can reach base clock, on a specified workload, at the specified temperature. So, it sounds like you're saying they simply relaxed their QA standards?

Not seeing the relationship there. TDP is a ceiling, anything less is within spec. There is nothing for QA to worry about when a chip with a 77W ceiling only needs 40W to meet operational and performance specs, 40W is less than 77W and therefore within 77W TDP spec, QA is happy.

 

[bit_user]

Not seeing the relationship there. TDP is a ceiling, anything less is within spec. There is nothing for QA to worry about when a chip with a 77W ceiling only needs 40W to meet operational and performance specs, 40W is less than 77W and therefore within 77W TDP spec, QA is happy.

But they set the base clocks at what clock speed the lowest-quality CPUs can run their specified workload on all cores, without exceeding TDP. So, either you're saying they were holding back with Coffee Lake-R (and why would that be?) or that they relaxed their QA standards to let chips through that don't meet spec.

I prefer the explanation that the 6-core CPUs are made with 10-core dies, and they simply get extra margin by cherry-picking the best cores.

Perhaps there are some other efficiency improvements, as well. The i9-9900KS review seemed to show improved power-efficiency at the same clocks as the i9-9900K, but maybe that was just a really good review sample.

 

[InvalidError]

So, either you're saying they were holding back with Coffee Lake-R (and why would that be?) or that they relaxed their QA standards to let chips through that don't meet spec.

If you "relax" QA to the point where parts no longer meet or outperform specs, you have a problem.

And yes, having more cores to pick from does provide additional opportunities to avoid cores that are at the worse end of any characteristic.

 

[TerryLaze]

According to https://www.anandtech.com/show/1458...ng-an-interview-with-intel-fellow-guy-therien :

​

This is quoted (10 months ago) straight from the guy at Intel who actually defined what TDP means.

Based on that, there shouldn't have been enough headroom in the Coffee Lake-R spec to simply switch on hyperthreading without an impact on TDP and/or base frequency. So, either they changed their definition of TDP, or something else (e.g. cherry-picking cores) enabled them to run more efficiently at the same base frequency.

Look at groups of CPUs,intel uses the same TDP on groups of CPUs that can run at specs with coolers of that TDP.
They don't calculate the exact TDP for each CPU independently (or at least that's not what they post) .
And HTT as well as the iGPU is already a part of that TDP number.
Specs for intel is the CPU running at or below 100 °C and at least base clocks for everything.

 

[TerryLaze]

If you "relax" QA to the point where parts no longer meet or outperform specs, you have a problem.

And yes, having more cores to pick from does provide additional opportunities to avoid cores that are at the worse end of any characteristic.

Yup,they don't disable "worse" cores they just boost the better ones and send software to those better boosted cores.
https://ark.intel.com/content/www/u...0900k-processor-20m-cache-up-to-5-30-ghz.html

Intel® Turbo Boost Max Technology 3.0 Frequency ‡
Intel® Turbo Boost Max Technology 3.0 identifies the best performing core(s) on a processor and provides increased performance on those cores through increasing frequency as needed by taking advantage of power and thermal headroom. Intel® Turbo Boost Max Technology 3.0 frequency is the clock frequency of the CPU when running in this mode.

 

[bit_user]

If you "relax" QA to the point where parts no longer meet or outperform specs, you have a problem.

That was my point. So, I'm not really sure what you're saying.

 

[bit_user]

They don't calculate the exact TDP for each CPU independently (or at least that's not what they post) .

Don't lose the plot, here. The real issue is the base clocks and the presence or absence of of HT.

Specs for intel is the CPU running at or below 100 °C and at least base clocks for everything.

That didn't change any time recently. T_Junction is 100 C for all the CPUs we're discussing, so it doesn't explain how they could enable HT on the i5-10400 @ same base clocks, unless either they were holding back on the i7-9400 or there's some other factor (which is my position).

they don't disable "worse" cores they just boost the better ones and send software to those better boosted cores.

Base clocks is the minimum guaranteed all-core speed. Your point is irrelevant for base clocks, which is what we're talking about.

 

[TerryLaze]

so it doesn't explain how they could enable HT on the i5-10400 @ same base clocks, unless either they were holding back on the i7-9400 or there's some other factor (which is my position).

That's the whole point of the first part of my post though, the TDP of the i7-9400 is "they don't calculate the exact TDP for each CPU independently" TDP is for a group of CPUs that often spans a huge number of them from low to high core count.

EDIT:
https://en.wikipedia.org/wiki/Coffe..._Coffee_Lake_processors_(Coffee_Lake_Refresh)
There are 65W TDP CPUs with 4 cores with 6 cores with 8 cores and with 8c/16t,they are all 65w TDP.

 

[bit_user]

TDP is for a group of CPUs that often spans a huge number of them from low to high core count.

Again, you're missing the point, which is about the base clocks & HT. The definition of the base clocks is a function of TDP, as I cite in post #5 - not my definition, Intel's. Base clocks & HT is really what we're really talking about, here.

There are only 3 explanations for how the i5-10400 gets HT at the same baseclocks as the i7-9400F:

1. They held back on the i7-9400F.
2. They relaxed QA standards on the i5-10400.
3. There is some other factor that enables the i5-10400's cores (the ones not disabled, anyway) to run more efficiently.

 

[TerryLaze]

Again, you're missing the point, which is about the base clocks & HT. The definition of the base clocks is a function of TDP, as I cite in post #5 - not my definition, Intel's. Base clocks & HT is really what we're really talking about, here.

Look at my edit

 

[bit_user]

Look at my edit

Still missing the point.

Think about car engines. I can build a 4-cylinder, a 6-cylinder, and an 8-cylinder that all achieve the same MPG per some standard. If I start with the MPG and then work back to how high I should limit the peak flow of their respective fuel pumps, then the differences between the respective engines will be reflected in their output numbers. Not a perfect analogy, but I think it illustrates my point - you can start with one parameter, and then derive the others from it.

So, yeah, Intel reuses the same thermal solution across different CPU models, and that's what ultimately determines TDP. We get it. But their definition of TDP in turn drives their advertised base clocks that they can guarantee with that CPU and thermal solution.

 

[InvalidError]

That was my point. So, I'm not really sure what you're saying.

By writing that Coffee Lake must have had no headroom, you sounded like all CPUs must reach TDP at baseline. My point was that TDP is a ceiling for base clock, a value that parts must NOT reach. Unless you are the unluckiest person on Earth, all parts will achieve baseline at significantly less than TDP even under non-ideal conditions, which leaves Intel with wiggle room to add more product bins without changing TDP or loosening QA if it wants to.

As Terry pointed out, if TDP was an exact spec, you wouldn't see everything from i3 through i7 sharing the same 65W bracket. It is little more than a convenient cooling solution design/selection guide so Intel can tell people and SIs to ah heck off if they run into issues while attempting to get away with less cooling than that.

 

[bit_user]

By writing that Coffee Lake must have had no headroom, you sounded like all CPUs must reach TDP at baseline. My point was that TDP is a ceiling for base clock, a value that parts must NOT reach.

The lowest-quality parts must not exceed.

But, the test probably accounts for some amount of degradation over time, so they probably have an even more stringent limit for QC.

Unless you are the unluckiest person on Earth, all parts will achieve baseline at significantly less than TDP even under non-ideal conditions,

It depends on how they bin them and what their yield is. Due to this, they could actually end up with a lot of parts that nearly squeak by the spec.

However, if their yields are extremely good and a part is the lowest bin, then you're right that very few parts will be right against the cutoff point.

which leaves Intel with wiggle room to add more product bins without changing TDP or loosening QA if it wants to.

As Terry pointed out, if TDP was an exact spec, you wouldn't see everything from i3 through i7 sharing the same 65W bracket.

TDP is defined by the thermal solution. That's why you have numerous SKUs sharing the same TDP - because they share the same thermal solution.