Alder Lake-S sample has 5.3GHz boost clock scores 11300 points in Cinebench R20

not buying these scores.8 big + 8 little cores beating 5950x,by some margin.....

 

Cinebench and the like would probably be the best-case scenario for Alder Lake though, right?

Going to pull some numbers out of my backside just to illustrate my line of thinking here...

Let's say SMP adds ~30% to the performance of the big, conventional, cores under multi-core load. That would mean that in the top-end Alder Lake configuration (8+16) the little cores would need to achieve 65% (130/2) of the performance of the big cores to be competitive with a conventional, all-big, 16-core design with SMP.

That seems doable, or close to, and considering the die area used quite efficient.
Loads up to 8, or whatever number of large cores, should also be good.

Mixed loads between the end spectrums would probably be weaker but those workloads are likely rare in comparison, so Intel's design for Alder Lake makes a lot of sense.

All this assumes Windows can flawlessly manage hybrid processor designs and assign work appropriately. A big ask and I'd imagine the weakest point of this strategy.

 

I don't care about cinebench score, but care about gaming performance.
Intel better don't f..... up the latency like they did with Rocketlake.
Rocketlake has 20% higher cache latency than "Comedylake", so higher IPC goes down the drain with higher latency.

That's why tweaked cometlake with ~35ns memory is impossible to beat for tweaked Rocketlake in games.

 

11300 in Cinebench is only speculation by some random person on a Chinese forum. 9300 is the only actual (still only rumoured but at least not completely made up) CB R20 score so far.

 

I'd rather have lower clocks with more cores and threads.

Efficient, low power, highly optimized awesomeness - because clock speed is all intel can do, and that is not good enough.

 

rkl is the joke
cml is actually excellent in gaming efficiency and value
if you "care about gaming" you'd know to stay away from rocket and get a 10700f or 10900f

 

Clock Speed > Cores.
Always, always.

 

Tell that to the shareholders, they will believe anything

 

oc a 4/4 kaby lake to 7GHz and see how it does against 9900k in modern games.
clock speed matters,but if you're gonna peg those cores at 100% then your experience will be better with lower clocks and more cores.

 

I think he ment enough cores with high coreclocks
Enough cores for any game today is 8-12 cores.
Hell, we had 8 cores in 2014! With Intel 5960x

 

6 is fine.
it's not great,it's not futureproof and your cpu will have to really sweat in some games,but it's fine.

 

It is entirely plausible. We haven't seen what this new architecture can do. Though, like with previous Ryzen leaks, I will believe it when I see it.

I do share your sentiment. That's not a 20% boost over Tiger Lake IPC... 20% doesn't close the gap between the 11900K and this. That's something akin to a 50% increase. That's like "FX to Zen" increase...

 

Then go buy Epyc

You didn't mention price, so I guess that's your next buy

Now go play Cinebench 24/7

 

It will certainly be my next buy, games, VM's and Cinebench should fly...

Just hope there's good VRM water cooling and a $1000 motherboard to pick up at the same time! No expense spared, precious!

The worst thing I predict is just like my x399 platform, there'll be no upgrade path...

 

really ?
didn't amd bin every cpu in 5/7/9 series asking more money for every 100mhz of single core performance ? not to mention the xt series
some people really have selective memory.

wait,is cinebench actual software ?

 

Yeah, really.

Multi-threading is the key, just look at what intel did with Sandybridge, happy days.

 

look what 8086k does to threadripper 1000/2000 in gaming.
massacred.
40-60% difference at times.so I hope you tried to be wrong on purpose.
https://www.computerbase.de/2018-08...50x-test/2/#diagramm-project-cars-2-1920-1080
all amd's gaming performance increase came from improved frequency,latency and ipc
5600x is double the performance of 1920x although it's half its core count
it's the furthest thing from efficient.

all you have to do with core count is hit the sweet spot,which is 6-8 atm.Afterthat it's ALL core speed and latency.

 

Sure is, very addictive - you kick it off and multiple squares fly around the screen and you get a score at the end... I gotta admit, I'm not very good at it - my score is almost exactly the same every time...

 

so it's a benchmark ?
it's the same as buying a gpu for running 3dmark
for rendering any cpu will kneel to cuda acceleration,especially on ampere which is probably the most compute-oriented nvidia card ever.