Intel's 14th Gen 'Raptor Lake Refresh' CPUs Nail a Total of 50 World Records (tomshardware.com) 40
Velcroman1 writes: Overclocking master Allen 'Splave' Golibersuch surfaced on Tom's Hardware to detail his work with liquid nitrogen to set a slew of new world records with Intel's Raptor Lake Refresh" CPUs. They include 15 world records with the Core i7-14700K and eight records with the Core i5-14600K, along with four records with the Core i9-14900K, spanning benchmarks from Cinebench to wPrime and H265.
"My top speeds were 7,730.11 MHz on all cores on the 14900K, 7,859.05 MHz on the 14600K and 7,600 MHz on the 14700K," writes Splave. "All of these achieved in Cinebench R23 while using Liquid Nitrogen cooling." "At the end of a week of playing around, I broke the 8-core Cinebench record at a crazy 7.73 GHz on all cores," concludes Splave. "Overall, these CPUs potentially OC better than their predecessors and cost the same. It was a rather refreshing refresh, I would say."
"My top speeds were 7,730.11 MHz on all cores on the 14900K, 7,859.05 MHz on the 14600K and 7,600 MHz on the 14700K," writes Splave. "All of these achieved in Cinebench R23 while using Liquid Nitrogen cooling." "At the end of a week of playing around, I broke the 8-core Cinebench record at a crazy 7.73 GHz on all cores," concludes Splave. "Overall, these CPUs potentially OC better than their predecessors and cost the same. It was a rather refreshing refresh, I would say."
Funny (Score:5, Informative)
In normal usage, these are almost identical to previous generation. Increases are tested to be less than differences between skylakes back in the day.
Guess you need liquid nitrogen level cooling to get any meaningful improvement out of them. That's the secret sauce.
Re:Funny (Score:5, Insightful)
Indeed, the performance gains are only seen when you have extreme cooling and are able to delivery 300+ watts to the CPU alone. Few consumers will be able to see any benefit over cheaper 13th gen parts.
Perhaps more relevantly, for most people getting an AMD part is going to offer better value for money and performance.
Re: (Score:1)
But with AMD, you're still stuck with retarded motherboards that keep on retraining memory, and take forever to boot, as well as random blue screens with high end memory.
It's become a norm. AMD gets great hardware and fails on software. Intel gets great software and fails on hardware.
Re: (Score:3)
You've got bigger problems if you boot so many times per week that that's a real problem.
Re: (Score:2)
You got very unrealistic expectations of a personal PC if you don't. Almost no one runs servers on home hardware.
It's been very informative watching people among top techs on youtube do the same thing I did back in post Athlon era. "Yeah, performance is excellent, but I'm going back to intel, because this great AMD system is fiddly AF, and most of the time I come home and want it to just start and work as intended".
Stop the fanboy apologia and press AMD on making better software. I the performance and stabi
Re: (Score:2)
It comes down to workload. Even without overclocking, Intel has been the single-threaded champ for quite some time, even though they occasionally get leap-frogged by AMD on single-threaded specs. If you have a single-threaded process that has to get done ASAP and power consumption is not important, then this Intel chip is the current top dog. Beyond gaming, there are very many real-world examples of this -- mostly having to so with existing CPU-bound algorithms that are impossible or impractical to re-wr
Re: (Score:3)
Thing is that even in gaming, the 3d cache CPUs eat the "single threaded performance king" anyway. Often for a lot less money, since 7800x3d is actually the better one of the two with one full 8 core unit with 3d cache, whereas the "top tier" model has one core unit with 3d cache and one without it, and then it tries to bounce the load around based on application. Except that AMD's software is awful as usual, so it misses the mark quite often, so performance is "great when it works, and not so great when it
Re: (Score:2)
Do you know who's hated even more than those who make bad jokes?
Those who feel the need to explain the bad jokes.
Re: (Score:2)
Pretty sure Russians are liked way more, as they have a lot of fans in the third world. And they beat the shit out of people who explain the jokes there.
Re: (Score:2)
That makes sense because the Intel 14th gen processors use the same core architecture as the 13th gen ones. The Core iX-14000 series was more of a rebranding effort than anything else.
Re: (Score:2)
X-band radio license (Score:2, Interesting)
Re: (Score:3)
Re: (Score:2)
I'll be interested when the X-rays they shoot are coherent.
Re: X-band radio license (Score:1)
Re: (Score:2)
That's be some pretty impressive overclocking, as the lowest energy wavelength considered X-ray band is 3nm, which is approx 100 PHz. Or 100,000,000 GHz.
Imagine a beowolf cluster of these!! (Score:2)
Re: (Score:2)
My question is where in the solar system you would source an abundant supply of chilled nitrogen.
I mean you could build a data center on Pluto but the lag would be astronomical.
Re: (Score:2)
Titan should do fine and is much closer - and cheaper. Plus, the sneaky bastards won't be able to hide in the nitrogen mist.
Not meaningful world records... (Score:5, Insightful)
If you take a modern heavily parallel benchmark it is completely meaningless to say you got an "8 core record". The whole point is that we can scale much better to more cores than frequency. The only meaningful limited record would be for power - cinebench score/Watt.
In the meantime, in the real-world, the fastest servers you can easily get (e.g. large cloud providers) are EPYC Genoa, which are cheaper and much faster than Intel's Sapphire Rapids (you'll have to wait quite a while for Raptor Lake). In fact, I could recently write about my experience with Google's new Genoa VMs [dev.to]: they were in a private preview this summer and I found out I could break many of the OpenBenchmarking.org records using the 90 core (180 thread) single or double processor VMs. Yes, I know these are server CPUs, but Xeon/EPYC is what is more relevant to "world record" benchmarking.
Obviously I welcome competition - We were on Xeons for the Haswell/Broadwell/Skylake generations - AMD only caught up with Rome and we switched everything to Milan when it became widely available. We'll switch back to Intel when they catch up and surpass AMD, but it doesn't seem likely to be very soon (I wouldn't be surprised if ARM solutions caught up faster than Intel - Graviton3 is fast) and liquid nitrogen benchmarks claiming world records by limiting to a few high frequency cores seem like little more than PR. Despite it being cool otherwise in both the literal and metaphorical sense ;)
Re: (Score:2)
If you take a modern heavily parallel benchmark it is completely meaningless to say you got an "8 core record". ...
I could recently write about my experience with Google's new Genoa VMs: they were in a private preview this summer and I found out I could break many of the OpenBenchmarking.org records using the 90 core (180 thread) single or double processor VMs. Yes, I know these are server CPUs, but Xeon/EPYC is what is more relevant to "world record" benchmarking.
The people overclocking their PCs don't care and wouldn't touch Xeon/EPYC with a 10ft pole. Single thread performance on high core count processors is less than abysmal and that's what matters to them.
Re: (Score:2)
Nah, lots of people overclock just to get the most out of their CPU - I'd say the vast majority of non-gamers would care about the all-core max clock rather than single-thread performance, simply because how most software we use nowadays works. I
Power requirements go to plaid as clock frequency and particularly voltages are jacked up. On a high core count system you will get very little actual performance in return for wasting and dissipating huge amounts of energy, reduced reliability and further erosion of available per core bandwidth.
If your workloads are such performance is a function of core count then likely it's something that should be running on a GPU anyway with (tens) of thousands of "slow" cores at its disposal.
Again the exception is people who are into games that are not that scalable - which is, admittedly, still quite a few games.
Going beyond vendor over
Re: (Score:3)
It's not meaningful because you can't buy a liquid nitrogen cooling system. This is only a stunt no matter what tests are run. 8 core performance might be valid to somebody, but liquid nitrogen performance is relevant to nobody.
Re: (Score:2)
liquid nitrogen performance is relevant to nobody.
I wouldn't be so quick to say that. Maybe it would be relevant if you had a particularly cold day... on Titan?
The point? (Score:3)
This is all entertaining and interesting, but what's the point for real-world use?
I could use nitro in a Porsche to achieve break-neck speeds over 100m, damaging the engine, entertaining the bros and chicks, but what's the benefit or point beyond bragging rights? ...shrug...
Re: (Score:2)
The chicks is the point.
I don't think that applies in this situation though...
Re:The point? (Score:5, Interesting)
This isn't one of those times, though. In this case, it's just marketing nonsense from a company that knows it put out a turd and wants to convince people it's not a turd. The past two generations were genuinely good; 14th gen barely adds anything and in some benchmarks actually regresses. I don't even need to look at the records it's breaking to conclude that they're irrelevant; if they were breaking any relevant records to any significant degree, standard benchmarks would also reflect that.
Re: (Score:2)
The point is scoring a higher benchmark.
The real world just buys more cores.
Re: (Score:2)
This is all entertaining and interesting, but what's the point for real-world use?
I could use nitro in a Porsche to achieve break-neck speeds over 100m, damaging the engine, entertaining the bros and chicks, but what's the benefit or point beyond bragging rights? ...shrug...
What's the point you ask? Perhaps ask YouTubers like WhistlinDiesel that question.
You'd have to catch him in between bank deposits and Ferrari purchases...not that we live in a world that rewards generously for that kind of behavior or anything...
I Have An Idea (Score:2)
Let's try this with an M3 Ultra. . .
More power Scotty. I can't make any promises Capt (Score:1)
time to move off planet? (Score:3)
Ho Hum (Score:2)
And my 6th gen i7 just keeps humming along.
Compare these results with the CPU size (Score:1)
7,859.05 MHz means just 3.8 cm per cycle at the speed of light. It's less than the physical size of the CPU (37.5*45 mm). Amazing.