Intel Core i9-13900 (non-K) Spotted with 5.60 GHz Max Boost, Geekbenched

[btarunr]

An Intel Core i9-13900 "Raptor Lake" (non-K) processor was spotted in the wild by Benchleaks. The non-K parts are expected to have 65 W Processor Base Power and aggressive power-management, compared to the unlocked i9-13900K, although the core configuration is identical: 8 P-cores, and 16 E-cores. Besides tighter power limits out of the box, and a locked multiplier, the i9-13900 also has lower clocks, with its maximum boost frequency for the P-cores set 5.60 GHz, compared to the 5.80 GHz of the i9-13900K. It's still a tad higher than the 5.40 GHz of the i7-13700K.

Tested in Geekbench 5.4.5, the i9-13900 scores 2130 points in the single-threaded test, and 20131 points in the multi-threaded one. Wccftech tabulated these scores in comparison to the current-gen flagship i9-12900K. The i9-13900 ends up 10 percent faster than the i9-12900K in the single-threaded test, and 17 percent faster in the multi-threaded. The single-threaded uplift is thanks to the higher IPC of the "Raptor Cove" P-core, and slightly higher boost clock; while the multi-threaded score is helped not just by the higher IPC, but also the addition of 8 more E-cores.



View at TechPowerUp Main Site | Source

 

[HenrySomeone]

13600k almost matching 5950x in multi thread? Noice!

 

[john_]

I guess we are finally at 6GHz without LN2. I mean, Zen 4 is at 5.7, Intel is at the same Turbo frequency if not a little higher. Finally 20 years latter we are at a frequency where Pentium 4 could have a meaning as a CPU.

 

[Crackong]

Time for this

Another day another intel 'Leak'

 

[Solid State Brain]

I wonder at what voltage they are boosting by default. The built-in (fused) voltage–frequency curve for the i9-12900KS could use over 1.5V for the maximum single-core frequency of 5.5 GHz, depending on silicon quality; Intel datasheets allow up to 1.72V maximum at least for brief periods.

 

[Dirt Chip]

From this, 13900 looks like an segnificant increase over 13700k in multi test.
Hope it will consume much less W to make it a relevant option for my new setup.

The k parts seems to push it all-in style.

I'm dont see how 7900x rivel the 13900 or the 13700k in multi test.

AMD need an ace up there sleeves or intel will see the top of this round.

 

[Mussels]

Is it just me that's tired of seeing 65W TDP CPU's that rely on 200W for their performance numbers?
They cant sustain it, so it's an either/or situation: It's fast, or it's 65W. It takes turns, but it can never be both.


Even seeing the 125W CPU's that use 250W+ just makes it seem those TDP values are pure fluff to mislead people.

 

[Daven]

Is there any IPC increase? The 12900ks at 5.5ghz scores around 2070 in Geekbench. This 5.6ghz part is 2130.

 

[john_]

From this, 13900 looks like an segnificant increase over 13700k in multi test.
Hope it will consume much less W to make it a relevant option for my new setup.

The k parts seems to push it all-in style.

I'm dont see how 7900x rivel the 13900 or the 13700k in multi test.

AMD need an ace up there sleeves or intel will see the top of this round.

Extra cores, even E cores, will give extra points in tests that can see all those threads. But there will be cases where more P cores make more sense than just more cores. In those cases 7900K could be a better option. But it's platform's price could be a disadvantage at a first glance compared to a DDR4 or a cheap DDR5 Intel option. On the other hand, Am5 is a platform that could be getting newer CPUs for the next 2-3 or more years. In that period someone with Intel might change their platform 1-2 times just to be able to use future Intel series. So the platform cost will be an advantage to Intel customers that don't keep the same platform anyway, for more than 12-18 months, while AM5 will be an advantage for those who keep their platform for as long as possible, not rushing to get the latest and the greatest from day 1.

Power consumption will probably be the same between 13700K and 13900, with the exception when pushing the chips. 13900 will probably be offering higher multi performance when all cores/threads are used, while 13700K will be offering better performance at single thread, or cases where only a lower number of threads are fully utilized. But the price difference will probably make the 13700K an excellent choice and the 13900 a bad choice.

AMD needs nothing really, just to stay competitive and keep the Ryzen brand name strong. As it is shown in a screenshot in another thread, Ryzen keeps selling strong, even with all the Alder Lake might out there.

Is it just me that's tired of seeing 65W TDP CPU's that rely on 200W for their performance numbers?
They cant sustain it, so it's an either/or situation: It's fast, or it's 65W. It takes turns, but it can never be both.


Even seeing the 125W CPU's that use 250W+ just makes it seem those TDP values are pure fluff to mislead people.

This will be the norm from now and on. Tech press was never really damnatory against that, neither people commending under threads like this one or in other sites. Most comments where "look at the performance numbers, look at the performance numbers, who cares about power consumption?". So, now AMD is also abandoning their focus on keeping power consumption in check and we see Zen 4 coming at 170W(230W) to match Intel. If AMD was choosing to stay under 150W at all times, they would probably having difficulty to pass 5.0GHz and tech press would be calling Zen 4 a disappointment.

 

[Pumper]

I'm dont see how 7900x rivel the 13900 or the 13700k in multi test.

AMD need an ace up there sleeves or intel will see the top of this round.

13900 will be competing with 7950x and AMD already said that Zen4 will be 40% faster than Zen3 in MT, so there's no reason for 7900x not to equal 13700k in performance while being more efficient, plus, buying Intel will leave you with a dead end platform, as Raptor will be the last CPUs for current platform, while AMD will keep AM5 for at least 3 gens.

 

[dgianstefani]

It's nice seeing the non-K models not be gimped so much in boost frequency.

There's still a decent baseline even at quoted TDP, the boost is just there if you have cooling headroom.

 

[Solid State Brain]

Is it just me that's tired of seeing 65W TDP CPU's that rely on 200W for their performance numbers?
They cant sustain it, so it's an either/or situation: It's fast, or it's 65W. It takes turns, but it can never be both.


Even seeing the 125W CPU's that use 250W+ just makes it seem those TDP values are pure fluff to mislead people.

Single-core or even 2-core performance should be at a power below the stated long-term/base TDP of 65W.

For MT performance, you should blame hardware reviewers and motherboard manufacturers. The latter especially most often use high or no power-current limits and tons of load voltage (leading to effectively overvolted operating conditions, i.e. voltages exceeding values in the CPU-fused voltage–frequency curve), making default settings far from being true Intel defaults. They are allowed to, since power limits are not a processor specification and any current/voltage is allowed if below the specified limit and temperatures do not exceed TjMax.

Hardware reviewers seem generally clueless about all of this.

If Intel-recommended PL1 (65W) and Tau time for locked processors (recently usually 28s) were actually respected, due to how the algorithm works the CPU would go from 200W to 65W (PL1) within 10 seconds, making PL2 influence on long benchmarks like Cinebench scores limited.

People who want to efficiently use their 65W CPU at 65W no matter what, should tune their motherboard settings accordingly.

 

[hs4]

Is there any IPC increase? The 12900ks at 5.5ghz scores around 2070 in Geekbench. This 5.6ghz part is 2130.

There is a question of what to call IPC. Since the number of decoders, registers or ALUs are the same, the logical "Instructions Per Clock" should remain the same. However, data transportation such as caches and buses have been improved, and the "statistical IPC" seems to have improved for certain types of operations because cycles wasted due to waiting were reduced e.g. 7zip decompress, some memory-related tasks in the geekbench. This is similar to the question of whether the FPS improvement by 3DVcache can be called an "IPC improvement." In the generation change from Zen2 to Zen3, the increase in effective instruction processing due to improved data transport was mixed into the "IPC improvement".

 

[dj-electric]

Well said @Solid State Brain. People over-dramatisize the meaning of TDP in modern CPUs and frankly this is getting old.
If you really want 65W or 95W or whatever, you can set it as a limit in BIOS in seconds. You will lose some performance, and that's all

Modern methods of retaining performance without challenging a mid range air cooler exist a plenty.

 

[Daven]

There is a question of what to call IPC. Since the number of decoders, registers or ALUs are the same, the logical "Instructions Per Clock" should remain the same. However, data transportation such as caches and buses have been improved, and the "statistical IPC" seems to have improved for certain types of operations due to reduced latency e.g. 7zip decompress, some memory-related tasks in the geekbench. This is similar to the question of whether the FPS improvement by 3DVcache can be called an "IPC improvement." In the generation change from Zen2 to Zen3, the increase in effective instruction processing due to improved data transport was mixed into the "IPC improvement".

So it looks like these improvements you mentioned are only increasing specific benchmark performance by just 1-2% going from Alder Lake to Raptor Lake. If this is the case there are only two takeaways under this generation transition:

• 5-6% higher clocks
• 8 more E-cores

 

[Dirt Chip]

13900 will be competing with 7950x and AMD already said that Zen4 will be 40% faster than Zen3 in MT, so there's no reason for 7900x not to equal 13700k in performance while being more efficient, plus, buying Intel will leave you with a dead end platform, as Raptor will be the last CPUs for current platform, while AMD will keep AM5 for at least 3 gens.

I'm very hope zen 4 will see 40% uplift, that will be wonderful!
I hope AMD can out a real solid option for my use profile (mostly adobe workloads and some casual gaming from time to time).

I'm upgrading every 10-12 years so the AM5 longevity is insignificant for me.

Just performance, price and powe consumption (because of the heat generate) in that order.

 

[nguyen]

Is it just me that's tired of seeing 65W TDP CPU's that rely on 200W for their performance numbers?
They cant sustain it, so it's an either/or situation: It's fast, or it's 65W. It takes turns, but it can never be both.


Even seeing the 125W CPU's that use 250W+ just makes it seem those TDP values are pure fluff to mislead people.

Do you play cinebench all day? because these CPU don't need 200W for their performance numbers running games

 

[thelawnet]

Is it just me that's tired of seeing 65W TDP CPU's that rely on 200W for their performance numbers?
They cant sustain it, so it's an either/or situation: It's fast, or it's 65W. It takes turns, but it can never be both.


Even seeing the 125W CPU's that use 250W+ just makes it seem those TDP values are pure fluff to mislead people.

Just you tbh.

You get hyperbolic clickbait reviewers (hi, Hardware Unboxed) who will take CPUs and run them for 24 hours under synthetic nightmare workloads, and then whine about things.

Intel's specs are clear

Intel® Core™ i9-12900 Processor (30M Cache, up to 5.10 GHz) - Product Specifications | Intel

Intel® Core™ i9-12900 Processor (30M Cache, up to 5.10 GHz) quick reference with specifications, features, and technologies.

www.intel.com

65W is the base power. If you want to to run 24 cores at turbo frequency it's going to use more than 65W. They give a number for that as well - 202W (for the 12900)

If you are playing games then your CPU will happily do 5.6 GHz at base power on one core. There aren't any games that are going to thrash 24 cores at full speed. But let's say you are rendering or something then yeah you are going to go up to 200W or whatever, and also considering that you spent $600 on a CPU then you're not going to really care about the board requirements PSU etc.

And if you want to maximise efficiency then you can just set power limits in the BIOS.

Whereas a GPU is going to sit a full power during a lot of gaming, the overwhelming majority of PC usage doesn't use the full power, and even if you throttle the CPU then it's rarely the end of the world. Like oh no, my CPU frequency is reduced by 10% when using all cores because I bought a cheap board. Like big deal, you saved some money and you lost a theoretical tiny amount of performance in some situation you might not ever encounter. (Like, if your CPU turbos for 10 seconds and then goes back down because it's done with what it was working on, then you haven't lost anything - this only affects people with long-running high power usage tasks)

 

[Jimmy_]

these are pretty good numbers in MT but waiting for zen4 MT results
65W TDP ~PL1 -- what is the PL2 for this non-k part? ~100W+ or how much?

 

[hs4]

So it looks like these improvements you mentioned are only increasing specific benchmark performance by just 1-2% going from Alder Lake to Raptor Lake. If this is the case there are only two takeaways under this generation transition:

• 5-6% higher clocks
• 8 more E-cores

"Improved efficiency" should be added to that. In Raptor Lake
- A clock uplift of about 5% appears to be achieved without any power penalty.
- By leaving PL2 unchanged, the basic operation to improve efficiency is performed: increasing the number of cores and lowering the clock.
This results in a performance gain of about 25-35% for the 13900K compared to the 12900K without increase in power consumption (240-250W).

p.s.
Most consumers refer to "IPC" as the specific benchmark number divided by the clock, so it should look like "improved IPC" by that definition, especially in geekbench.

 

[Solid State Brain]

65W is the base power. If you want to to run 24 cores at turbo frequency it's going to use more than 65W. They give a number for that as well - 202W (for the 12900)

Even those 202W, they are nothing more than a recommended value according to the actual processor datasheet (available here or here).

There's no specified minimum/maximum value for PL1/PL2, although the underlying suggestion is that they should preferably be adjusted according to thermal and electrical capabilities of the motherboard and cooling solution.

 

[Crackong]

It is 2022 not 2017
No one believes that '65W' number anymore, from both companies.

 

[Denver]

Do you play cinebench all day? because these CPU don't need 200W for their performance numbers running games

So you are saying that the only scenarios that the extra toy cores are used properly are irrelevant? Does the discourse in defense of intel change according to the positioning of the planets or something? It's funny.

 

[nguyen]

So you are saying that the only scenarios that the extra toy cores are used properly are irrelevant? Does the discourse in defense of intel change according to the positioning of the planets or something? It's funny.

People who don't know how to set a power limit on their CPU shouldn't be buying high end CPUs....

Here is how mobile ADL scale with power vs Ryzen 6000

So just because Intel CPU use more power for more perf, Intel must suck? , some people have really skewed perspective

 

[Rowanero]

I'm upgrading every 10-12 years so the AM5 longevity is insignificant for me.

Just performance, price and powe consumption (because of the heat generate) in that order.

Power consumption should be very important if you value price. In 5 years of use, 5 hours a day, it could yield a difference of around 80-100$ depending on various factors. It IS a world of difference that many people does not take into account. But you need to know real world power comsumption under your typical loads, and that's not easy to obtain.