News Intel Launches Arrow Lake Core Ultra 200S — big gains in productivity and power efficiency, but not in gaming

[Admin]

Intel announced the first five models of its Arrow Lake desktop processors, otherwise known as the Core Ultra 200S series, with prices ranging from the $294 14-core Core Ultra 5 245KF to the flagship $589 24-core Core Ultra 9 285K.

Intel Launches Arrow Lake Core Ultra 200S — big gains in productivity and power efficiency, but not in gaming : Read more

 

[bit_user]

This correlates to up to 15C lower CPU package temperatures during gaming, which will then ease cooling requirements.

I wonder how particular they're being about this wording. While the package could indeed be 15C cooler, that doesn't necessarily mean the tile with the CPU cores is!

I'm eager to know if they're facing something of a hotspot/thermal bottlenecking problem like what became very apparent in Zen 4. Intel is now joining AMD with chiplets made on a smaller node, which means heat in a much more concentrated area (i.e. higher thermal density).

Also, the hot spot is off-center. That and being smaller should pose more challenges for heatpipe-based coolers (or, at least the vast majority which lack a vapor chamber in the base).

 

[YSCCC]

Interesting that it basically states it's on par for gaming with 9950X, and claims content creation benchmark vs a 7950X3D which is slower than 7950X in that regard...

Efficiency should be much better than RPL but in a month or so it will be interesting to see how the price and performance competes this gen

 

[dimar]

Will be waiting for 9950X3D benchmarks. 13900K serves me well so far, so it may not be worth upgrading until the next generation or the one after.

 

[TheHerald]

This is one freaking bold graph from Intel.

I find it hard to believe, but I also find it hard to believe that they'd make this up since it's very specific and easily falsified.

 

[JamesJones44]

With a 10c reduction in heat, I guess the question will be, how much headroom do these processors have for overclocking for those interested in the K series. If it's a lot, the base 10% slower might not be a big deal. If the headroom is low, it's going to be a hard sell for gamers.

 

[Notton]

And then there are rumors Intel won't add a refresh generation, so this socket is for a single gen only.

 

[TheHerald]

And then there are rumors Intel won't add a refresh generation, so this socket is for a single gen only.

Probably false, has never happened before.

 

[ingtar33]

Probably false, has never happened before.

no, the rumors for this are coming from some pretty reputable sources; so it's probably true.

 

[TheHerald]

no, the rumors for this are coming from some pretty reputable sources; so it's probably true.

There is no way it's a single gen socket.

 

[Gururu]

I am VERY confused if the comparisons are among integrated graphics or with an identical dedicated card. Does it mean that these chips will serve as bottlenecks for any graphics card we buy, even mid-range? I think the headline is deceiving, unless it means exactly that. Noone who is concerned with gaming is going to worry about integrated graphics.

 

[joeer77]

Who cares about less power consumption? I don't. I care about speed. Looks like the move from FinFETs to GAA not worth it so far for Intel. We need to see some real benchmarks in the next month.

 

[TheHerald]

Who cares about less power consumption? I don't. I care about speed

Those 2 are one and the same. If it's as fast as the 14900k at half the power then it will be faster at the same power.

 

[Roland Of Gilead]

Finally! A use for Foveros 3d!

 

[Neilbob]

Guess all this remains to be seen with the usual slew of third-party benchmarks. I particularly care about the (default) power consumption, so I hope this isn't just Intel blowing a bunch of hot air (pun intended).

I found the actual productivity benchmark slides a bit cringe-inducing: 12 of the 15 were explicitly centred around AI, which I still just don't see being as important in the world of consumers as they seem to be portraying.

 

[The Historical Fidelity]

There is no way it's a single gen socket.

I heard that the other cpu series for this socket was cancelled, so it could be true….but, to your point, I have no verifiable proof of it.

 

[bit_user]

This is one freaking bold graph from Intel.

I find it hard to believe, but I also find it hard to believe that they'd make this up since it's very specific and easily falsified.

Why is that hard to believe? This data shows the i9-14900K (pre-mitigation) is 23.8% faster on CB R23 at 253 W than 125 W:

Spoiler: CB R23 @ different power limits

So, essentially, what they're saying is that the new Core U9 285K will be about 23.8% faster at 125 W than the i9-14900K was (on a heavily MT floating-point workload). Given that they've basically jumped 2 major process nodes and accounting for all the IPC improvements they said they achieved in Lion Cove and Skymont, it doesn't seem hard for me to believe.

Also, I'm pretty sure it's a smaller improvement @ 125 W than what Raptor lake achieved vs. Alder Lake.

 

[bit_user]

There is no way it's a single gen socket.

Meteor Lake-S was supposed to be the first generation to use this socket. That got cancelled and instead we got Raptor Lake Refresh.

Probably false, has never happened before.

What happened before was the Broadwell desktop CPUs were pretty much cancelled. However, that was supposed to be the second generation in LGA1150. So, they released a "Haswell Refresh" that also used that socket, but it was basically the same CPU.

In this case, it would be like if Haswell were the one to get cancelled and then the Broadwell Desktop CPU ended up being the first LGA 1150. But, then Skylake was already planned to use LGA1151 and they just stuck with that plan, leaving LGA1150 as a single-gen socket.

 

[bit_user]

Those 2 are one and the same. If it's as fast as the 14900k at half the power then it will be faster at the same power.

The graph you quoted shows that. Just draw a line up instead of across.

 

[TheHerald]

Why is that hard to believe? This data shows the i9-14900K (pre-mitigation) is 23.8% faster on CB R23 at 253 W than 125 W:

​

​

Spoiler: CB R23 @ different power limits

​

So, essentially, what they're saying is that the new Core U9 285K will be about 23.8% faster at 125 W than the i9-14900K was (on a heavily MT floating-point workload). Given that they've basically jumped 2 major process nodes and accounting for all the IPC improvements they said they achieved in Lion Cove and Skymont, it doesn't seem hard for me to believe.

Also, I'm pretty sure it's a smaller improvement @ 125 W than what Raptor lake achieved vs. Alder Lake.

Yes, I calculated ~20% but you are correct, the 285k is supposedly closer to 25% faster than the 14900k. Indeed it's a smaller increase than ALD --> RPL but that one added 8 cores, now Intel is removing 8 threads. It's kinda crazy. And regardless, just because it happened with RPL doesn't make it any less insane. If those numbers are correct that would make the 285k ~70% more efficient than my 12900k with both at 125w.

 

[TheHerald]

Meteor Lake-S was supposed to be the first generation to use this socket. That got cancelled and instead we got Raptor Lake Refresh.


What happened before was the Broadwell desktop CPUs were pretty much cancelled. However, that was supposed to be the second generation in LGA1150. So, they released a "Haswell Refresh" that also used that socket, but it was basically the same CPU.

In this case, it would be like if Haswell were the one to get cancelled and then the Broadwell Desktop CPU ended up being the first LGA 1150. But, then Skylake was already planned to use LGA1151 and they just stuck with that plan, leaving LGA1150 as a single-gen socket.

Haswell was still 2 generations. You can argue it was a refresh but that's generally my point, that no matter what get's canceled and what doesn't it's highly unlikely we will only see one generation on the socket.


According to anandtech's review 4790k was ~12-14% faster than 4770k. That's a bigger jump than..uhm. oh well, zen 4 to zen 5

 

[thestryker]

I am VERY confused if the comparisons are among integrated graphics or with an identical dedicated card. Does it mean that these chips will serve as bottlenecks for any graphics card we buy, even mid-range? I think the headline is deceiving, unless it means exactly that. Noone who is concerned with gaming is going to worry about integrated graphics.

All of the gaming benchmarks are done using an RTX 4090. If you're in a GPU limited situation then it's unlikely to make a meaningful difference.

 

[bit_user]

According to anandtech's review 4790k was ~12-14% faster than 4770k.

No, that's not what I'm seeing. At stock speeds, it's more like 3% - 10%, depending on the benchmark.

https://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/3

 

[JarredWaltonGPU]

Haswell was still 2 generations. You can argue it was a refresh but that's generally my point, that no matter what get's canceled and what doesn't it's highly unlikely we will only see one generation on the socket.


According to anandtech's review 4790k was ~12-14% faster than 4770k. That's a bigger jump than..uhm. oh well, zen 4 to zen 5

I mean, if you want to call Devil's Canyon a separate product from Haswell, you could even argue it was three generations on that socket. Not that anyone really bought Broadwell 5775C!

Anyway, just because it hasn't happened before doesn't mean this isn't the time it does. I have definitely heard multiple rumors, some from pretty credible sources, that LGA1851 is going to be a single cycle. And honestly, I don't really care that much. I know others do, but if I upgrade to a new CPU? Yeah, I'm getting a new motherboard to go with it.

 

[TheHerald]

No, that's not what I'm seeing. At stock speeds, it's more like 3% - 10%, depending on the benchmark.

https://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/3

Was talking about the productivity - not the gaming benchmarks, eg. handbrake is +12.2%. Still, according to TPU the 9950x is 3.5% faster than the 7950x in the application aggregate.

I mean, if you want to call Devil's Canyon a separate product from Haswell, you could even argue it was three generations on that socket. Not that anyone really bought Broadwell 5775C!

Anyway, just because it hasn't happened before doesn't mean this isn't the time it does. I have definitely heard multiple rumors, some from pretty credible sources, that LGA1851 is going to be a single cycle. And honestly, I don't really care that much. I know others do, but if I upgrade to a new CPU? Yeah, I'm getting a new motherboard to go with it.

Im not particularly interested about architectures but actual performance. Yes it was a refresh with basically just higher clockspeeds (and better thermal transfer) but the performance delta was bigger than it is between zen 4 and 5. Does that mean am5 is one generation platform? Of course not.