Early Core i9-13900K Review Hints that it Holds up to the "20-40" Claim

[btarunr]

An early review of a retail Intel Core i9-13900K "Raptor Lake" 8P+16E processor shows it holding up to the rumored "20-40" claim, the idea that the processor can be up to 20% faster in gaming, and up to 40% faster in productivity, compared to the current i9-12900K. Much of the gaming performance increase is attributed to the higher IPC of the new "Raptor Cove" P-cores, and the much higher boost clocks they run at (up to 5.80 GHz); whereas the multi-threaded performance boost comes from not just the faster P-cores, but a doubling in the E-core count to 16, and improved E-core cache structures, besides higher clock speeds that they run on. For tests that scale across P-cores and E-cores, the i9-13900K behaves like a 24-core/32-thread processor, which is what it is. Among the tests included are CSGO, AIDA64, 7-Zip, WinRAR, Cinebench R15, R20, and R23; and their average, in comparison to the i9-12900K.



View at TechPowerUp Main Site | Source

 

[Valantar]

Looks promising if true. IMO they're going the right way with this - keeping the same number of P cores but making them faster, while increasing the number of E cores for MT performance. I just hope they expand E cores further down the lineup this time around.

Edit: and, of course, at some point they really need to bring P core boost power under control. Regardless of overall chip efficiency, taking into account uncore power etc., there's something quite wrong when your core consumes 3-4x the peak power of your main competitor's core for a relatively slight performance advantage.

 

[Crackong]

(unlimited)

 

[Vayra86]

Someone explain to me why I would ever want an 1x900k CPU anytime soon though. It was pointless in 12th gen, and its going to be pointless in 13th.

 

[Valantar]

Someone explain to me why I would ever want an 1x900k CPU anytime soon though. It was pointless in 12th gen, and its going to be pointless in 13th.

That's the thing: just like with smartphones, any real need to buy a high end CPU has long since evaporated for most people. As soon as we got 6c mid-range CPUs (thanks, AMD!), the only meaningful differentiator became boost clocks, and now even those are high across the board. Tbh I only see myself buying lower end CPUs in the future unless it's specifically for a production-oriented build that needs a lot of cores. The only reason my main PC has a 5800X is that the 5600X was entirely unavailable at the time - and these days as I'm not at home I'm using an i5-4670S. Which, while slow in a lot of cases, is still perfectly usable. But then again I don't play competitive online shooters, which do tend to need a bunch of threads these days.

 

[Sabotaged_Enigma]

What about power consumption? Tell me bravely!

 

[Vayra86]

What about power consumption? Tell me bravely!

Scientists & engineers are still hard at work trying to get the power station back online after the recent Prime95 run, they didn't get a proper measurement that time.
Source:

 

[Jimmy_]

What about power consumption? Tell me bravely!

shhh! dont talk about that - its still on fire

 

[Tomorrow]

What about power consumption? Tell me bravely!

It's in the headline: 20-40. So 20-40% higher power consumption for the same amount of performance increase.

/s

 

[Tomorrow]

RPL has the same out of the box power consumption as ADL.

Really? RPL has not launched but yet you somehow know for a fact that it's the same as ADL.

 

[mechtech]

I almost feel bad for all the ADD people who bought 12k series.

 

[Punkenjoy]

Looks promising if true. IMO they're going the right way with this - keeping the same number of P cores but making them faster, while increasing the number of E cores for MT performance. I just hope they expand E cores further down the lineup this time around.

Edit: and, of course, at some point they really need to bring P core boost power under control. Regardless of overall chip efficiency, taking into account uncore power etc., there's something quite wrong when your core consumes 3-4x the peak power of your main competitor's core for a relatively slight performance advantage.

The day they will finally have fixed their missmatched ISA so their big core do not have that much dark silicon due to AVX-512 being fused out, i would agree.

Most of the heavily threaded workload do not require high frequency/high IPC.

They a great for silicon area but in the end, intel will have to get out of monolithic sooner or later. With the same yields, they cannot outproduce AMD if with the same amount of wafer. Smaller chips on AMD just means way less waste in case of defect.

RPL has the same out of the box power consumption as ADL. This joke stopped being funny almost a year ago.

But in the enthusiast space, who run ADL using box power consumption? even less, when you think there are motherboard manufacturer that out of the box crank out the power consumption to the maximum.

In the end, most people don't run all core workload all day so it should not be a problem (same for Zen 4 and RPL).

 

[HalfAHertz]

...is attributed to the higher IPC of the new "Raptor Cove" P-cores.

You've stated in multiple news articles that this is a new core. There is no new core - it's the same golden cove based core with larger caches. It might actually end up having lower IPC in some workflows because of the higher latencies. Luckily gaming is not one of those workflows.

Most of the performance increase is purely from the frequency bump and the longer turbo times. (5.2Ghz vs 5.8Ghz)

879 / 778 -> 13% increase
5.8 / 5.2 -> 11% increase

Don't just repeat intel's propaganda leaks marketing material but actually inform your audience.

 

[ModEl4]

Zen4 and Raptor will be close in gaming with minimum (1-3%) 1080p Ultra (or High) differences for same tier models.
In multithreading the 2 top Zen4 models will be competitive with equivalent Raptor models but the 2 lower end Zen4 will not.
The problem is pricing, i said it in the past many times, Raptor Lake official 1K unit price will not exceed the following limits, so nowhere near the 20% increase that many AMD supporters are hopping:

13900K $649
13900KF $624
13700K $459
13700KF $434
13600K $319
13600KF $294

 

[Valantar]

There have been numerous leaks including those which included PL1/PL2 readings which are the same as ADL.

I'm not responsible for people living under the rock.

None of this is official though, and as with all leaks they should be treated with some skepticism. I doubt we'll see higher power limits - they're already crazy - but until launch or some really official-looking documents leak from a reputable source, pretty much anything is possible.

 

[Tomorrow]

There have been numerous leaks including those which included PL1/PL2 readings which are the same as ADL.

I'm not responsible for people living under the rock.

I said fact not reckless speculation. This can go both ways. It may show a CPU better than it is (default power) or worse than it actually ends up being (maximum limit).

 

[tussinman]

I almost feel bad for all the ADD people who bought 12k series.

There's nothing really to feel bad about. 13th gen will work on there motherboard and resell value is high so they can make most of there money back if they really did want to upgrade. Don't understand feeling bad or A.D.D ? (investing early in LGA 1700 wasn't a big deal)

 

[oxrufiioxo]

Zen4 and Raptor will be close in gaming with minimum (1-3%) 1080p Ultra (or High) differences for same tier models.
In multithreading the 2 top Zen4 models will be competitive with equivalent Raptor models but the 2 lower end Zen4 will not.
The problem is pricing, i said it in the past many times, Raptor Lake official 1K unit price will not exceed the following limits, so nowhere near the 20% increase that many AMD supporters are hopping:

13900K $649
13900KF $624
13700K $459
13700KF $434
13600K $319
13600KF $294

Why would AMD supporters want RPL to be expensive it's best for them if Intel price matches ADL or even lowers pricing it would force AMD to either release cheaper models or come out with X3D variants quicker.

Makes no sense to me unless they are crazy fanboys and in that case what they hope doesn't mean $h!+.

I almost feel bad for all the ADD people who bought 12k series.

Why? They've been rocking very good chips for almost a year and ST will be probably 10% at most better so not even worth the upgrade unless you need the extra MT of the 13900K. I see the wink but just like your 1700 being a very good cpu when the 2700X launched all the ADL chips will still be very good.

 

[freeagent]

Wow look at all the gains they found! How come they weren't doing all this finding while AMD was doing a bunch of their own?

 

[Jism]

Wow look at all the gains they found! How come they weren't doing all this finding while AMD was doing a bunch of their own?

They just looking for ways to extract the best possible performance for their line of chips.

Like we used to had "real" speeds which where pretty much locked. It's already pretty complicated such as Base clock (i.e 3.2Ghz) and a Boost clock "up to" $value

Very nice but the upper range of that boost clock wont be kept long if your cooling is'nt good enough. So on avg you buy a 3.2Ghz model and it usually sits above half of what it can do or is specified for.

GPU's do the same. Not all workloads are obviously the same. So as long as within current or temperature range you can push clocks harder.

If i had a engine of exactly 20HP and only had to power one wheel, it could go much faster, then when i would have to power 4 wheels including a differential distributing that power.

Something in that order really.

 

[oxrufiioxo]

Wow look at all the gains they found! How come they weren't doing all this finding while AMD was doing a bunch of their own?

They got stuck on 14nm, and honestly what they were still able to accomplish on a 5+ year old process node was actually impressive. That being said their P cores are still much larger than AMD Zen4 cores so they still have a ways to go if they ever want to do a more than 8 P core varient.

Keep in mind amd can also substantially increase cache size without making the cores bigger with the X3D varients.

 

[P4-630]

I almost feel bad for all the ADD people who bought 12k series.

Then you should feel bad for the people that bought AMD 5xxx as well

Everyone here knows that you can expect a new CPU gen every year.

I bought a i7 12700K, May 2022, I have no regrets, my previous CPU was a i7 6700K....
I can still do a Raptor Lake upgrade later on if I want to.

 

[Punkenjoy]

Then you should feel bad for the people that bought AMD 5xxx as well

Everyone here knows that you can expect a new CPU gen every year.

I bought a i7 12700K, May 2022, I have no regrets, my previous CPU was a i7 6700K....
I can still do a Raptor Lake upgrade later on if I want to.

Well that depend

Ryzen 5xxxx released in November 2020 and ADL in november 2021. People buying both right now shouldn't unless they need, but if they purchased it in 2021 and maybe even the first half of 2022, they got enough for their money.

 

[P4-630]

the first half of 2022, they got enough for their money.

I ordered my i7 12700K on May 1st 2022, glad I did since prices went up short after that.

 

[Berfs1]

At this point does more than 500 FPS in CSGO really matter?? I still do 240 FPS in CSGO (with frame lock) on my 9900K with HT disabled lol