Raptor Lake Leaks + Intel 4 developments

[Deleted User 456458]

With 400w power draw it could be a 720mm aio and it would still throttle, the block couldn't get the heat out no matter the size of the radiator

That's why DLVR was very appealing. You'd have lower load voltages which would lead to less power draw so you effectively give yourself more room. That doesn't seem to be case here. Instead it's 12900ks with 8 more ecores worth of current.

 

[uscool]

and someone said it was architecture improvements, its just refresh isnt it so not suprising , at the same locked clock it gets the same scores ? so pretty much just got that from increasing the clock which needs more power and multi core due to adding 8 more e cores which needs more power

wouldnt say its disappointing or makes me think wow , pretty much what to expect from refresh its decent but looking forward seeing the kind of power usage and thermals on multi core workloads

 

[humbug]

So, I think the big news is that the single threaded performance of the 13900K and 12900K is basically identical at the same clockspeeds:

I find it quite incredible that several people were insisting that 'Raptor Cove' cores were a thing, and that there would be an IPC increase. Despite no indication of this on Intel's core roadmap.
they can feed him alsots of
So, what you will actually get is an increase in cache, and the potential to clock the CPU higher, with high end cooling.

As usual it was MLID who said there would be an IPC increase.

One day he will learn he's being used as a marketing arm by Intel, they can feed him all sorts of "don't buy from our competitors, we have all this fantastical stuff coming" to put out without any consequences when it turns out to be non sense.

 

[Joxeon]

5.7% clock bump for 10% improvement. The word stellar doesn't come to mind.

10% on the same node is still decent, let's not forget AMD have only touted around 15% with a 10% bump in clocks on a superior process node.

 

[Journey]

10% on the same node is still decent, let's not forget AMD have only touted around 15% with a 10% bump in clocks on a superior process node.

It's decent, but not stellar as some people are making out, hence exactly what I said. And also yet I agree 15% is't amazing either, especially with the larger than normal clock speed uplift. Not sure what you actually added other than the word decent.

 

[Deleted User 456458]

10% on the same node is still decent, let's not forget AMD have only touted around 15% with a 10% bump in clocks on a superior process node.

Gains will be hit n miss depending of how cache limited a workflow is.

If I was building a RPL setup, I'd max it out as such:
8P cores, HT off, 1 ecore cluster
Wait for Hynix A Die to be out
Get Z790 2 Dimm board

Then you can have some fun and pull notably ahead of a 12900k but for plug n play crew, good luck.

 

[humbug]

its better than a decade of 3.5% increments.

 

[SmokSmog]

They fixed the Cinebench R20 numbers, decent single thread uplift. They must have improved the 10nm node because you can't just take 12900K then slap 8 E cores and expect it to clock to 5,8ghz.

RPL 862ST
ADL 4ghz 601ST

862/601*4ghz = 5,737

 

[g67575]

I'm afraid not, it's a no. Both the 12th gen and 13th gen P-Cores will perform the same if clocked at 5.0ghz or 5.5ghz. I think some people managed over 6ghz already on a 12900K, probably with liquid nitrogen.

The results are pretty much what you'd expect:

CPU-Z Benchmark for Intel Core i9-12900K (1T) - CPU-Z VALIDATOR

Best CPU performance - 64-bit - November 2024

valid.x86.fr

The voltage is pretty mental at 1.545v.

It looks like Intel have increased the voltage (around 1.34v) for the 13900K.

 

[humbug]

hhmmmm ^^^^

It does have more L2 and L3. A lot more.

32MB L2 vs 14MB
36MB L3 vs 30MB

That's a lot of L2, Zen 3 only had 512KB per core, Zen 4 is rumoured to up that to 1MB per core, so 8 vs 16 L2 total.

I can't remember who talked about this, it may have been Mike Clark who said you can't just increase the size of the cache and expect a performance increase, for one you have to improve the speed of fetch or it just takes longer wading through more data and you nullify any performance you might gain from capacity, and if you're not die shrinking the physical area increase can add its own latency, again same effect.

So i'd like to know how expanding the L3 by a factor of 3 has such a dramatic effect for them, there must be more to it that just gluing a slab of cache on top of the CPU.

I'd also be interested to see if Intel just made a bigger CPU.

 

[g67575]

You're right, the Cache (particularly L2) could give it a nice boost in games, vs the 12th gen.

The main thing to consider, is the amount of cache per core. There's a 60% increase in the L2 cache per core for the 13th gen.

AMD and Intel are both competing to squeeze out a few more frames in games with cache increases.

 

[humbug]

You're right, the Cache (particularly L2) could give it a nice boost in games, vs the 12th gen.

AMD and Intel are both competing to squeeze out a few more frames in games with cache increases.

Yeah, it could.

 

[Cyber-Mav]

Ram latency is so high that more cache helps boost speed by keeping more data closer to CPU cores

 

[humbug]

Ram latency is so high that more cache helps boost speed by keeping more data closer to CPU cores

That's a good point.

 

[SKILL]

hhmmmm ^^^^

It does have more L2 and L3. A lot more.

32MB L2 vs 14MB
36MB L3 vs 30MB

That's a lot of L2, Zen 3 only had 512KB per core, Zen 4 is rumoured to up that to 1MB per core, so 8 vs 16 L2 total.

I can't remember who talked about this, it may have been Mike Clark who said you can't just increase the size of the cache and expect a performance increase, for one you have to improve the speed of fetch or it just takes longer wading through more data and you nullify any performance you might gain from capacity, and if you're not die shrinking the physical area increase can add its own latency, again same effect.

So i'd like to know how expanding the L3 by a factor of 3 has such a dramatic effect for them, there must be more to it that just gluing a slab of cache on top of the CPU.

I'd also be interested to see if Intel just made a bigger CPU.

Being big/little it's a bit more nuances than just X vs Y cache, or at least the L2 is.

L3 is 6Mb bigger and shared between all cores, good but not going to make a big difference across varied workloads really.

L2 looks massive, 18Mb more! But when you dig a bit you find 12Mb of that is in the E-Cores, certainly from a gaming/single-thread perspective this is useless. So 'only' 6Mb across the P-cores, still a big increase per-core but not quite as impressive as it first appears.

 

[g67575]

Are L2 and L1 cache increases more likely to help with general CPU tasks then?

 

[Deleted User 456458]

Being big/little it's a bit more nuances than just X vs Y cache, or at least the L2 is.

L3 is 6Mb bigger and shared between all cores, good but not going to make a big difference across varied workloads really.

L2 looks massive, 18Mb more! But when you dig a bit you find 12Mb of that is in the E-Cores, certainly from a gaming/single-thread perspective this is useless. So 'only' 6Mb across the P-cores, still a big increase per-core but not quite as impressive as it first appears.

Correct. It’s real world impact will be interesting to see when frequency aligned to adl and mem tuned (to maximise the core throughput)

 

[Journey]

From a pure gaming point-of-view I do wonder if the 13th gen parts will be able to clearly surpass the X3D from AMD, its so close currently but just how much extra that L2 and L3 caches will help. Going to be interesting for those who only play games when the reviews land, so they can see a proper picture and side-by-side, same for AMD really they could have a really confused line up on their hands.

 

[Bencher]

From a pure gaming point-of-view I do wonder if the 13th gen parts will be able to clearly surpass the X3D from AMD, its so close currently but just how much extra that L2 and L3 caches will help. Going to be interesting for those who only play games when the reviews land, so they can see a proper picture and side-by-side, same for AMD really they could have a really confused line up on their hands.

Arent the current 12th gen already surpassing the 3d?

 

[uscool]

Arent the current 12th gen already surpassing the 3d?

Pretty sure he means 3d on zen4 ?

He meant pure gaming perspective
Is it though beating the current one isn't it pretty much close affair to say one massively ahead ? Also it does use less power whilst doing it

The i9 Killer? AMD Ryzen 5800X3D Tested

Check prices on Amazon belowAMD Ryzen 5800X3D: https://geni.us/9GXJrNSAMD Ryzen 5600X: https://geni.us/JXHkwGoAMD Ryzen 5900X: https://geni.us/Qt5VWIntel 129...

www.youtube.com