News Intel 10th Gen Comet Lake Processor as Fast as AMD Ryzen 7 3800X? Early Benchmarks Point to Twice the Power Draw

[Admin]

Intel's upcoming i7-10700KF desktop CPU was allegedly benchmarked before release and seemed to be as speedy as the AMD Ryzen 7 3800X.

Intel 10th Gen Comet Lake Processor as Fast as AMD Ryzen 7 3800X? Early Benchmarks Point to Twice the Power Draw : Read more

 

[King_V]

Sorry, I can't help it....

 

[jgraham11]

Has anyone looked at Toms Hardware's little minute and a half video
on choosing the right CPU:
1. "As long as its current gen the brand difference is a wash"
a) Toms doesn't recognize the problem with all the Intel CPU bugs, many of which are not patched, some have been tried and failed.
Intel recommends you disable Hyper-Threading if you care about security.
b) Toms doesn't recognize the double the power for same work thing.
c) Toms doesn't recognize that AMD CPUs are 95% of what Intel
CPUs are for gaming while being more than double the performance
for everything else at almost every price point.

2. "Clock speed is more important than core count"
a) Coincidently that statement favours Intel CPUs
b) When a CPU does more per clock cycle that's not the case
c) AMD CPUs cores are better: better IPC, better L1 cache, etc.
d) Intel's ring bus latency is the only thing keeping AMD behind in
gaming, not clock speed.

The last 3 actually seem legitimate...

Tom's: why don't you just come right out with it: "We support Intel"

Just buy it!

 

[derekullo]

Has anyone looked at Toms Hardware's little minute and a half video
on choosing the right CPU:
1. "As long as its current gen the brand difference is a wash"
a) Toms doesn't recognize the problem with all the Intel CPU bugs, many of which are not patched, some have been tried and failed.
Intel recommends you disable Hyper-Threading if you care about security.
b) Toms doesn't recognize the double the power for same work thing.
c) Toms doesn't recognize that AMD CPUs are 95% of what Intel
CPUs are for gaming while being more than double the performance
for everything else at almost every price point.

2. "Clock speed is more important than core count"
a) Coincidently that statement favours Intel CPUs
b) When a CPU does more per clock cycle that's not the case
c) AMD CPUs cores are better: better IPC, better L1 cache, etc.
d) Intel's ring bus latency is the only thing keeping AMD behind in
gaming, not clock speed.

The last 3 actually seem legitimate...

Tom's: why don't you just come right out with it: "We support Intel"

Just buy it!

Obvious trolling but I'll play along.

1a. Tom's has numerous articles on Intel CPU bugs
https://www.tomshardware.com/news/i...ities-in-cascade-lake-chips-and-a-new-jcc-bug

Intel vs AMD Processor Security: Who Makes the Safest CPUs?

Who has the most secure processors, AMD or Intel?

www.tomshardware.com

AMD wins 5 out of 6 of the safety tests.

1b. This article and numerous other Tom's articles mention that Intel's 14nm is not as power efficient as AMD's 7nm.
"The Ryzen 7 3800X has a 105W TDP and complies with the rating, according to our testing from our Ryzen 7 3800X review. However, Intel's TDP criteria is a bit different. Intel might market the i7-10700KF with a 125W TDP even if it only adheres to that when running at base clock speeds. In reality, the chip's peak power consumption would probably be far greater when operating at the boost clock speed. "

1c. Tom's benchmarks show exactly what you are saying for most games with Farcry 5 and a few others being Intel dominated.

2a. For the most part games favor clock-speed but it all depends on how a game is coded. As above Farcry 5 does appear to favor clock-speed.
Having said that most modern games are coded for at least 4 threads so you need at least 4 threads at a bare minimum for decent performance.

2b. Architecture does matter, but at the end of the day real world benchmarks are how we interpret how successful an architecture is.

2c. Can't argue/explain opinion.

2d. Intel does need a solution for this.
"Intel is starting to reach the core-count barrier beyond which the ringbus has to be junked in favor of Mesh Interconnect tiles, or it will suffer the detrimental effects of ringbus latencies. "
https://www.techpowerup.com/review/intel-core-i9-9900k/3.html

I don't have a bias for Intel or AMD.
I choose the one that has the best performance per dollar for the games and other software I use.

 

[alextheblue]

a higher TDP allows for higher sustained turbo frequencies when you strictly adhere to Intel’s spec (i.e. with power limits enforced). That is not something that the vast majority of reviews (including Tom’s Hardware own reviews) on the 9900K/9900KS did however. So, the power consumption figures for the 9900K/9900KS that we have are with power limits removed (hence and the reported violation of the TDP for the 9900K/9900KS).

Sorry, but Intel's "spec" is completely worthless. They should print it on toilet paper. Why? Intel doesn't actually bother enforcing it at a chip level. They tell the board manufacturers "OK this is what the spec is, now do whatever you want". So you can take a STOCK board, slap in the latest Intel i7 Core+++ KSFTW 10777, and change NO SETTINGS... and it will run roughshod over these supposed Intel specs, again WITHOUT overclocking or touching ANYTHING. Meanwhile AMD chips behave as expected regardless of board... until the user intervenes and changes something.

Intel absolutely could override the board settings and stay within spec, at a chip level, until the user enables auto or manual OC. They could absolutely force board manufacturers to default to those settings. They do not... as a result, blaming reviewers who "don't adhere to Intel's spec" is completely ridiculous. They would literally have to change FROM stock settings TO Intel "Spec" by detuning. Their results would no longer reflect what end-users experience.

Anyway, not all loads hammer these processors the same. You conveniently ignored the fact that they were supposing a peak power of 250W. That requires a high-end board, cooling, and a taxing load with AVX like you'd get with Prime95. With that being said, they are likely to pick up 10-20W over the previous gen flagships even in more typical heavy workloads.

 

[mdd1963]

250W?

I'd be highly skeptical of that, short of it perhaps actually drawing that much at perhaps 5.2-5.3 GHz on all cores....then, it might very well be accurate.)

 

[usiname]

(4) Since the 10700K has (slightly) improved silicon compared to the 9900K, it should be expected to have better power efficiency (closer to a highly binned 9900K i.e. a 9900KS). It should also have better thermals as the die it is based on is that of the 10-core 10900K which is larger than that of the 9900K and therefore heat is transferred away through a larger area. And better thermals help with power consumption as lower temps means lower resistive transistor power losses.

9600k - 149mm2
9700k - 174mm2
try again

 

[Olle P]

... Intel's "spec" is completely worthless. ... Intel doesn't actually bother enforcing it at a chip level. They tell the board manufacturers "OK this is what the spec is, now do whatever you want". So you can take a STOCK board, slap in the latest Intel i7 Core+++ KSFTW 10777, ... and it will run roughshod over these supposed Intel specs, ...

I couldn't agree more, but with one caveat: It does depend on what motherboard is used.
Would be nice to see some benchmarks of the higher end Intel CPUs with B360 motherboards used.

 

[mitch074]

250W?

I'd be highly skeptical of that, short of it perhaps actually drawing that much at perhaps 5.2-5.3 GHz on all cores....then, it might very well be accurate.)

I wouldn't be surprised if it actually peaked at 250W with default motherboard settings, as all mobo makers do push Intel's Turbo boost far past Intel's specs. A 5.2-5.3 GHz all-core overclock would probably pull more than that, since a 9900K OC'd at 5.0 already pull 250W - and both chips use the same process.

 

[germanium]

I have the Core I9 9900K & it can draw as much as 250 watts but usually crashes at that point even with water cooling. I have disabled hyperthreading & am running all core boost 5.0GHz. At these settings it can still draw up to 180 watts but can run this speed all day long without down clocking due to heat or power draw. This is due to my settings in bios. Most programs do not max out power at all though i have seen windows antivirus pull lots of power due to using AVX code but with hyperthreading disabled can still run at full 5.0GHz. With hyyperthreading turned on I have to reduce all core speed to 4.9GHz for AVX code work on windows antivirus. These are not stock settings. TDP only applies to stock non turbo speeds on these chips. on the older 4 core non k chips TDP was indeed fore all cores max turbo but not these 8 core chips not even at stock max all core turbo frequency. Most programs without hyperthreading max out at less than 150 watts with my settings.

 

[joeblowsmynose]

250W?

I'd be highly skeptical of that, short of it perhaps actually drawing that much at perhaps 5.2-5.3 GHz on all cores....then, it might very well be accurate.)

Agreed, the 9900k can pull 250w under certain loads (Tom's review of 9900k) .. so when we take vasically the identical processor and we crank up the clocks higher the consumption should go down, for sure. [/sarcasm]

Intel actually has two TDPs for their processors; the one that they advertise is the TDP for base frequency(PL1), the TDP for boost frequency on the 9900k (or maybe it was the KS?) was I think 250w (PL2 - "Peak") TDP. I actually saw an Intel document with these numbers on it, and I thought ... "so Intel actually does have accurate TDP figures, at each PL state ... but they only advertise and share one of them ..." -- If they actually shared both, everyone would be impressed with the accuracy of their TDPs ... because when you see both, then the Pl1 TDP that they advertise actually makes sense.

Anandtech noted this and had a similar sentiment back in 2018: https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo

I think most people buying Intel CPUs these days though have no intention of actually running any meaningful full load on them. Gaming generally doesn't peg CPUs much at all. So maybe it doesn't matter.

If a person is doing any serious muliti-threaded heavy workloads, I think they would have moved to Ryzen by now, or will be soon ... or wait another year for Intel gets their node size down ...

 

[joeblowsmynose]

I have the Core I9 9900K & it can draw as much as 250 watts but usually crashes at that point even with water cooling. I have disabled hyperthreading & am running all core boost 5.0GHz. At these settings it can still draw up to 180 watts but can run this speed all day long without down clocking due to heat or power draw. This is due to my settings in bios. Most programs do not max out power at all though i have seen windows antivirus pull lots of power due to using AVX code but with hyperthreading disabled can still run at full 5.0GHz. With hyyperthreading turned on I have to reduce all core speed to 4.9GHz for AVX code work on windows antivirus. These are not stock settings. TDP only applies to stock non turbo speeds on these chips. on the older 4 core non k chips TDP was indeed fore all cores max turbo but not these 8 core chips not even at stock max all core turbo frequency.

Thank you for the candid anecdote. Not enough people sharing their own personal results honestly, these days, generally replaced by rhetoric or hearsay. I have a higher appreciation for people who talk about their personal experiences.

 

[mattkiss]

"The Ryzen 7 3800X delivers an efficiency of 2.74 GOPS/W while the i7-10700KF seemingly offers a disappointing 1.18 GOPS/W. The i7-10700KF's peak power consumption would work out to around 250W in this specific benchmark, which draws considerably more power than most standard applications. Although this would be on par with the Ryzen 7 3800X, the i7-10700KF would pull more than twice the power of its competitor."

What would be on par with the Ryzen 7 3800X?

 

[alextheblue]

I couldn't agree more, but with one caveat: It does depend on what motherboard is used.
Would be nice to see some benchmarks of the higher end Intel CPUs with B360 motherboards used.

Can those be configured with settings beyond "Intel spec" as well? Assuming you are looking at a model with decent power delivery. Something to think about. Either way, an "Intel spec" board with be neat to test with but... who would pair the two? I wouldn't expect anyone buying a high-end chip with a board that (regardless of chipset) doesn't at least have really solid PD. Something in the $150 range would sort of be a minimum IMO.

 

[mdd1963]

At least they did not use the 3700X and attempt to infer it only actually uses 65W...

 

[Olle P]

Can those be configured with settings beyond "Intel spec" as well? ...

I suppose the motherboard manufacturers have the technical ability to go beyond the specs, but shouldn't.
The power delivery should be good enough to provide for the demands within specs. (Including delivery of relatively high power for shorter periods.)

 

[alextheblue]

I suppose the motherboard manufacturers have the technical ability to go beyond the specs, but shouldn't.
The power delivery should be good enough to provide for the demands within specs. (Including delivery of relatively high power for shorter periods.)

If they can, I'd be shocked if some of them didn't. It would mean a performance advantage out of the box vs the competition. That's what made me wonder if Intel restricted it at a chipset level. Otherwise I would think it wouldn't be entirely dissimilar to how some B series AMD boards have superior PD vs others.

 

[Olle P]

... It would mean a performance advantage out of the box vs the competition. ...

A performance advantage and a price disadvantage on a market where price is the deciding factor.
Have you seen a broad review of B360 boards comparing computer performance? I haven't.