News AMD Ryzen 9 9950X Engineering Sample gets a full suite of Blender benchmarks at various TDPs, showcasing major efficiency improvements

[TheHerald]

It's also much slower in that configuration.

A stock 7950X3D consumes 140W in MT ( source: https://tpucdn.com/review/amd-ryzen-9-7950x3d/images/power-multithread.png ), if both have the same points per watts and the 14900K is at 125W, it means the 7950X3D is faster to finish the task by ~140/125 = 1.12 = 12%.

To be clear at approximately ISO efficiency:

7950X3D @ 140W * 253.3 points = 35462
14900K @ 125W * 250.9 points = 31363

35462 / 31363 = 1.13069 = 13.1% faster for the 7950X3D.

With a bit of clever math ( and "MyCurveFit" Source: https://mycurvefit.com/ - reference of that particular curve:

https://imgur.com/6HcwGWe

View: https://imgur.com/6HcwGWe
) we can extrapolate the 14900K performance at 140W given the data points of 253W, 200W, 125W, 95W, 65W and 35 in the curve of TPU's results which lands around 31882 points.

So at ISO power:

7950X3D @ 140W = 35462
14900K @ 140W = 31882

11.2% faster for the 7950X3D

With that data we can also extrapolate for fun, the points per watts at 140W for the 14900K to 227.7 giving:

7950X3D = 253.3 Points per Watt
14900K = 227.7 Points per Watt

And at ISO performance ( around 35000 points for both processors, more or less a few tenths of a percent ) we can also arrive at:

7950X3D = 140W
14900K = 253W

80.7% higher power consumption for the 14900K.

In conclusion:

- At ISO efficiency the 7950X3D is faster by ~13%.

- At ISO power the 7950X3D scores ~11% higher.

-At ISO performance the 7950X3D consumes ~44% less power ( or the 14900K consumes 80.7% more, same thing ).

You can dance around reality all you want and try to distort the facts to suit your argument, it doesn't matter.

Whether you dispute those proven undisputable facts that have been confirmed over and over again by the whole professional reviewing community and beyond or not is irrelevant.

BTW, are you related to the guy running Userbenchmark? I'm getting similar vibes here. I just hope this demonstration will get you back on track with reality contrarily to that guy which seems pretty hopeless.

I rest my case, it is ironclad and I'm done here, there's nothing left to say.

Note: Sorry abufrejoval, I misquoted, my bad.

Your extrapolations are completely off...just saying. You think that from 125w to 140w it only scores 500 more points? LOL. Sure bud, whatever, it's pointless to argue wiith fanatics

 

[bit_user]

The myth lives on, doesn't it?

So, you're back to making mismatched comparisons, I see.

Restricted to 125w it has similar efficiency to the 7950x 3d and far better than the 7950x. Please, the data is there

Except that when you restrict it to 125W, it scores only 31293. The 7950X3D scores 35769, which is 14.3% faster. So, we're faced with the unpleasant dilemma that Raptor Lake offers, for multithreaded workloads: performance or efficiency, but not both.

Spoiler: Cinebench R23 MT

P.S. I don't even know why we're rehashing this issue yet again, when it will become moot in only 2 weeks.

 

[TerryLaze]

So, you're back to making mismatched comparisons, I see.


Except that when you restrict it to 125W, it scores only 31293. The 7950X3D scores 35769, which is 14.3% faster. So, we're faced with the unpleasant dilemma that Raptor Lake offers: performance or efficiency, but not both.

Not both at the same time but at least it does offer both, in stark contrast to ryzen that offers only efficiency with no option to do anything to get more performance out of it.

Also, as always, you have to restrict testing to a single app, or group of apps, to get a 10-15% difference.
For someone that always goes on and on about having as many data points as possible you sure always go for single results...
At 125W (average of 91w ) the 7950x3d is like 4% faster if you look at everything, including but not just, the server type apps.

 

[bit_user]

Also, as always, you have to restrict testing to a single app, or group of apps, to get a 10-15% difference.

That's not actually true. The problem with TechPowerUp is they don't have an aggregate of just the multithreaded apps, the way Toms and ComputerBase do.

​

For someone that always goes on and on about having as many data points as possible you sure always go for single results...

Oh, I do value multiple data points. However, TheHerald was presenting Cinebench R23 MT efficiency data without the matching performance data. I was just rectifying that.

At 125W (average of 91w ) the 7950x3d is like 4% faster if you look at everything, including but not just, the server type apps.

I disagree that things like rendering and compiling are "server-type apps". Lots of people do those things on their desktops.

 

[TheHerald]

So, you're back to making mismatched comparisons, I see.


Except that when you restrict it to 125W, it scores only 31293. The 7950X3D scores 35769, which is 14.3% faster. So, we're faced with the unpleasant dilemma that Raptor Lake offers, for multithreaded workloads: performance or efficiency, but not both.

Spoiler: Cinebench R23 MT

P.S. I don't even know why we're rehashing this issue yet again, when it will become moot in only 2 weeks.

"only".

Sure, it's only the 2nd fastest desktop chip in existence at 125w. Horrible inefficient product, lol. There is nothing mismatched about my comparison

 

[TheHerald]

Oh, I do value multiple data points. However, TheHerald was presenting Cinebench R23 MT efficiency data without the matching performance data. I was just rectifying that.​

The data I posted had performance included in them. Since the x3d is equally efficient at 140w it's damn obvious that it's faster. In fact at iso efficiency you can just divide the power draws of each cpu and find the performance delta. Which btw is small. Nothing like the nonsense that is constantly thrown around.

 

[TerryLaze]

That's not actually true. The problem with TechPowerUp is they don't have an aggregate of just the multithreaded apps, the way Toms and ComputerBase do.

​

Which part isn't actually true?
Because multithreaded is one group of apps.
And the sub group of multithreaded that computerbase uses is the apps that have the highest benefit on the ryzen platform. So cherry picked out of already cherry picked, and the performance difference at the same power (142w ) is 13% so everything I said is true and yet you are trying to make me out to be a liar by providing additional data that is proving me right...

I disagree that things like rendering and compiling are "server-type apps". Lots of people do those things on their desktops.

Lots of people eat on their desktops, doesn't mean that we have to judge them as dinner trays.

 

[bit_user]

at iso efficiency you can just divide the power draws of each cpu and find the performance delta.

That just gives you a point sample, though. The full perf/W curves are more useful.

 

[TheHerald]

That just gives you a point sample, though. The full perf/W curves are more useful.

Doesn't really matter, point is people act like amd has a huge efficiency difference when that's just not true. Especially in any other segment bar the 7950x one. This campaign of misinformation needs to stop, and you are not helping to make it happen.

As we speak the 13700kf is 20$ more expensive than the 7700x in both amazon de and amazon Com, and it's both a lot faster and a lot more efficient.

I really don't get where people get the idea that amd is more efficient. It's not. Not at current prices, not at msrp prices. It just isn't. In fact biggest drawback is their bad efficiency. Just because they are limited to very low wattages compared to Intels chips doesn't make them efficient, it just makes them slow.

 

[TheHerald]

-At ISO performance the 7950X3D consumes ~44% less power ( or the 14900K consumes 80.7% more, same thing ).

You can dance around reality all you want and try to distort the facts to suit your argument, it doesn't matter.

Whether you dispute those proven undisputable facts that have been confirmed over and over again by the whole professional reviewing community and beyond or not is irrelevant.

BTW, are you related to the guy running Userbenchmark? I'm getting similar vibes here. I just hope this demonstration will get you back on track with reality contrarily to that guy which seems pretty hopeless.

I rest my case, it is ironclad and I'm done here, there's nothing left to say.

The audacity to ask if I'm related to user benchmark when you are claiming that the 7950X3D has the same performance as the 14900k,lol. Even the data you yourself posted show otherwise. But as I've said, the myth must go on, reality isn't important

 

[bit_user]

Doesn't really matter, point is people act like amd has a huge efficiency difference when that's just not true. Especially in any other segment bar the 7950x one. This campaign of misinformation needs to stop, and you are not helping to make it happen.

From my perspective, the data is pretty clear:

​

The R9 7950X beat the i9-13900K from about 65 W and above, all the way until some crazy point to the right.

The R9 7900X beat the i7-13700K across the range. I think the i7-14700K reverses this, to some extent, thanks to those 4 additional E-cores.

Where the tides turn is that the R7 7700X loses to the i5-13600K after about 65 W.

However, I can try to do a Cinebench R23 (MT) comparison by combining the ComputerBase single-app data for the R9 7950X and the TechPowerUp i9-14900K data.

I really don't get where people get the idea that amd is more efficient.

I think you do. People have seen variations on this data enough times and in enough places. I know you focus a lot on idle power and single-threaded efficiency, but those don't tend to be highlighted in the same way as multithreaded, because people like to see what happens to these CPUs at the limit.

I'm not saying you don't have a couple points, but I think you shouldn't be acting so surprised.

 

[TheHerald]

From my perspective, the data is pretty clear:

​

​

The R9 7950X beat the i9-13900K from about 65 W and above, all the way until some crazy point to the right.

The R9 7900X beat the i7-13700K across the range. I think the i7-14700K reverses this, to some extent, thanks to those 4 additional E-cores.

Where the tides turn is that the R7 7700X loses to the i5-13600K after about 65 W.

However, I can try to do a Cinebench R23 (MT) comparison by combining the ComputerBase single-app data for the R9 7950X and the TechPowerUp i9-14900K data.


I think you do. People have seen variations on this data enough times and in enough places. I know you focus a lot on idle power and single-threaded efficiency, but those don't tend to be highlighted in the same way as multithreaded, because people like to see what happens to these CPUs at the limit.

I'm not saying you don't have a couple points, but I think you shouldn't be acting so surprised.

So going by current prices, Intel wins also in MT efficiency in all but one segment. So no, I really don't understand how people are concluding otherwise.

 

[Simon_78]

You made up a bunch of numbers which are not accurate (your picture also uses Blender power consumption not Cinebench though they're close to the same). Your overall point is still mostly accurate (the 7950X3D is better perf/W at multi, but @TheHerald also didn't say otherwise), but basically none of your math is an accurate reflection of reality.

Below are the actual numbers for Cinebench R23 Multi if you don't understand why I said what I did.

Spoiler

Then let's replot the results, see if it changes anything.

New predicted 140W results: 32324

VS. the 31882

A discrepancy of 1.4%...

Source:

https://imgur.com/TSV6Sg9

View: https://imgur.com/TSV6Sg9

Edit:

The Herald:

Your extrapolations are completely off...just saying. You think that from 125w to 140w it only scores 500 more points? LOL. Sure bud, whatever, it's pointless to argue wiith fanatics

I'm doing my best to keep the sarcasm out of it, but it's so hard at this point...

Surely a mere 1.4% error will validate your argument about the "myth" when the previous margins were more than 10% at the smallest margins of the ISO comparisons?

Common man, let it go. You're just publicly humiliating yourself at that point, it's sad to watch. There is no "myth", there is no "conspiracy". Please stop pushing disinformation, it's just not cool to those who seek correct information.

That's the only reason I'm here, it's not for you, it's for those who will do a google search and end up here looking for accurate information and stumble upon your disinformation.

To those people, I hope I've cleared it up for you, have a nice day!

 

[thestryker]

Then let's replot the results, see if it changes anything.

New predicted 140W results: 32324

VS. the 31882

A discrepancy of 1.4%...

Yeah... that wasn't the big problem part of your post...

With that data we can also extrapolate for fun, the points per watts at 140W for the 14900K to 227.7 giving:

7950X3D = 253.3 Points per Watt
14900K = 227.7 Points per Watt

And at ISO performance ( around 35000 points for both processors, more or less a few tenths of a percent ) we can also arrive at:

7950X3D = 140W
14900K = 253W

This is the big problem with your made up numbers. The 14900k at 200W scores around a thousand points more than the 7950X3D. Just stop making up numbers when it comes to power and performance scaling. It doesn't work unless the scaling is linear which it isn't at these power levels.

 

[Simon_78]

Other websites have plotted the curves as bit_user demonstrated.

Yet you still cherry pick, you claim "cults" and "myths" and anything but acknowledge reality.

I swear you guys are spawns of Userbenchmark at this point.

You guys don't need a fact check, you're beyond that.

You guys need therapy, and I ain't a psychotherapist.

Peace out.

 

[bit_user]

Other websites have plotted the curves as bit_user demonstrated.

I'd point out that mine were real data points (except for one or two, which were slightly interpolated). They apparently fit curves neatly enough that I could've ventured there, but I didn't want to give a false impression that we know more about what those intermediate datapoints look like than we do.

I tend to agree with @thestryker that there's enough nonlinearity not to trust extrapolations of any sort. I've jumped on TheHerald for doing that, in the past.

Peace out.

Thanks for stopping by! Your help is most definitely appreciated!
: )

 

[eX_Arkangel]

AMD is, unequivocally, using much less power for the same work.

You seem to like TPU, so I'll include their data too.

https://tpucdn.com/review/intel-core-i9-14900k/images/efficiency-multithread.png

https://tpucdn.com/review/intel-core-i9-14900k/images/efficiency-gaming.png

https://tpucdn.com/review/intel-core-i9-14900k/images/power-games-compare-vs-7950x3d.png

https://tpucdn.com/review/intel-core-i9-14900k/images/power-games-compare-vs-7950x.png

https://tpucdn.com/review/intel-core-i9-14900k/images/power-applications-compare-vs-7950x3d.png

https://tpucdn.com/review/intel-core-i9-14900k/images/power-applications-compare-vs-7950x.png

Quote from W1zzard ( reviewer at TPU ): Compared to AMD's offerings there's still a night-and-day difference. During rendering, the 14900K consumes 282 W, the 7950X needs 80 W less, while offering virtually the same performance. The 7950X3D even uses only 140 W, which is half of the 14900K, and it's only marginally slower. Similar situation in gaming: 14900K average gaming power of 144 W, 7950X: 89 W, 7950X3D: 56 W, 7800X3D: 49 W (!)—all while offering virtually the same FPS.

I'll also include Gamers Nexus results for good measure.

https://gamersnexus.net/u/styles/large_responsive_no_watermark_/public/inline-images/GN CPU Benchmark Blender 3.6.4 (GN Logo) Power Efficiency GamersNexus.png.webp

Quote from Steve: "The 14900K is towards the bottom-end of the chart. It’s more efficient than the 13900K, 14700K, 12900K, and 13700K -- and AMD’s 5800X, barely -- but it’s far below top performers like the 5950X and 7950X. It’s not the least efficient we’ve tested, but Intel is pushing its thermal envelope and power draw higher and higher."

Now if you still want to claim otherwise, it's your prerogative, but it's also deeply flawed.

you quoted the wrong user, your response was meant more likely for terryblaze who is known for his extreme mental gymnastics to be able to still believe Intel is more efficient than AMD at any range or point.