Intel Teases 6 GHz Raptor Lake at Stock, 8 GHz Overclocking World Record (tomshardware.com) 55
Tom's Hardware reports: We're here in Israel for Intel's Technology Tour 2022, where the company is sharing new information about its latest products, much of it under embargo until a later date. However, the company did share a slide touting that Raptor Lake is capable of operating at 6GHz at stock settings and that it has set a world overclocking record at 8GHz - obviously with liquid nitrogen (here's our deep dive on the 13th-Gen Intel processors). Intel also shared impressive performance projections for single- and multi-thread performance.
Notably, the peak of 6 GHz is 300 MHz faster than the 5.7 GHz for AMD's Ryzen 7000 processors, but Intel hasn't announced which product will hit that peak speed. We also aren't sure if a 6GHz chip will arrive with the first wave of chips or be a special edition 'KS' model. Intel also claimed that Raptor Lake will have a 15% gain in single-threaded performance and a 41% gain in multi-threaded, as measured by SPECintrate_2017 and compared to Alder Lake, and an overall '40% performance scaling.'
Notably, the peak of 6 GHz is 300 MHz faster than the 5.7 GHz for AMD's Ryzen 7000 processors, but Intel hasn't announced which product will hit that peak speed. We also aren't sure if a 6GHz chip will arrive with the first wave of chips or be a special edition 'KS' model. Intel also claimed that Raptor Lake will have a 15% gain in single-threaded performance and a 41% gain in multi-threaded, as measured by SPECintrate_2017 and compared to Alder Lake, and an overall '40% performance scaling.'
Anyone can print a slide... (Score:3, Insightful)
Let me know when they start shipping silicon of this speed in large quantities. Most folks don't have liquid N2 setups to do this sort of overclocking, either, so that point's a bit moot. Intel's also still a lousy investment financially, which doesn't say much for its future technologically. Maybe the CHIPS act can bail them out, but I don't think it's going to provide enough money to do much. My.Magic Eight Ball says "Outlook not so good" which seems about right to me.
Re: (Score:2)
Intel's also still a lousy investment financially, which doesn't say much for its future technologically. Maybe the CHIPS act can bail them out, but I don't think it's going to provide enough money to do much.
CHIPS barely moves the needle on Intel's ready cash, they've got liquidable investments far exceeding the amount of our money Congress is handing them to do what they should be doing anyway. They could burn cash at their current rate for years without running out.
Re: (Score:3)
Re:Anyone can print a slide... (Score:4, Insightful)
My.Magic Eight Ball says "Outlook not so good" which seems about right to me.
Generating $0.8bn cash last quarter, paying out a $1.5bn dividend, and changing their end-year forecast from $65bn revenue to $68bn is "outlook not so good"?
I think your eightball is broken. The world is more complicated than a GHz number.
Re: (Score:2)
so, borrowing 700M to pay some dividends ? I'm not sure that's a good financial move...
Re: (Score:2)
A move that has been done in the past by basically every single company in the fortune 500. But you knew that right because you're *the* business expert.
Put your money where your mouth is, how big is your short position?
And yet, (Score:1)
Double standards (Score:5, Interesting)
Just days after writing about how hot the future Ryzens will be at speed ... with not a mention of heat from the Intel part.
A deja-vu even, the 12900K needed almost double the power to compete with the 5950X with hardly a mention of how hot the Intel ran then either.
Re:Double standards (Score:4, Insightful)
The competition is good for the industry. Let them keep leapfrogging each other.
Personally, I haven't felt the need to upgrade in years. I'll take my 64GB of RAM and 32 threads and won't care if it's bleeding edge. Plenty fast for anything I'd ever throw at it.
At work, we've been bouncing Intel in favor of AMD Epyc for a few years. It's been a much better deal in terms of cost and cost per watt....at least for our typical workloads. I'm sure there are companies out there that will pay a premium to eeek out that very last FLOP....but not me.
Best,
Re:Double standards (Score:4, Interesting)
Re: (Score:2)
So you loved the idea of easy upgrading but in reality that didn't matter? How does this argue for AMD?
Re: (Score:2)
Re:Double standards (Score:5, Insightful)
That's because most problems are RAM- or I/O-bound, not CPU-bound.
When the CPU has to go to RAM it's 100s of clock cycles. For a single-threaded application (i.e. most of them) and a cache hit rate of 99% that still means 99 clock cycles executing instructions then 100-300 clock cycles stalled fetching memory to fill the cache miss. I/O latency to something as fast as SSD is even longer.
Re: (Score:2)
Let them keep leapfrogging each other
What's it called when one of the frogs just keeps leaping over itself?
Re: (Score:3)
Re: (Score:2)
In addition, the clock speed a CPU is capable of is irrelevant for most people because the limiting factors are heat and power.
I.e. having a cooling system capable of removing enough heat, and having a power system capable of delivering enough juice.
I don't know why Intel bothers with stunts like this. Does anyone buying CPUs not check independent benchmarks on more realistic systems?
Re: (Score:2)
Re:Double standards (Score:4, Interesting)
Oh, there is a known limiting factor: The memory architecture. AMD has been better there for a long time now.
Re:Double standards (Score:5, Interesting)
>Oh, there is a known limiting factor:
Well there is the transistor aging. Good for at least 10 years worst case when *operated within the specified frequency, voltage, temperature envelope*.
We could make them clock faster if you were happy with a fraction of the devices bricking after 2 years.
The circuits I work on won't actually work down at super low temperatures. The thermal noise is not noisy enough. We design for and test down to -40C, but who knows what's happening below that? Nothing good I suspect.
Re: (Score:2)
Indeed. That is too technical for most people here though, so I left it out.
Re: (Score:2)
AMD has been less austere with its cache- which is fantastic, but memory architecture-wise, Intel is on top. Their latencies have been and continue to be better than AMD parts.
For a while, you could run faster memory on AMD parts, but then Intel leapfrogged with DDR5, and really, that's not about memory architecture anyway.
Re: (Score:2)
Nonsense. Do you remember how long it took Intel to integrate the memory controller in the CPU? That was years after AMD because they could not do it. Also, latency is by far not everything. Also notice how Intel is now planning multi-chip modules, again years after AMD and after they stated they would never do that. Intel is playing a game of catch-up and they are not winning.
Re: (Score:2)
Do you remember how long it took Intel to integrate the memory controller in the CPU?
Not relevant to the claim. Why the strawman?
Also, latency is by far not everything.
Eh, it really kind of is.
There's bandwidth and latency. Other than that, both consumer grade chips are quad channel. And Intel still outperforms in both metrics.
Also notice how Intel is now planning multi-chip modules, again years after AMD and after they stated they would never do that.
What the fuck does this have to do with the qualitative analysis between UC memory controllers?
Intel is playing a game of catch-up and they are not winning.
No, you're talking out your ass to to grind an ax.
Re: (Score:2)
Re: (Score:2)
I'm sure cache misses prevent linear improvement. For example, if you run at twice normal speed then you're going to hit main memory twice as frequently. For benchmarks that keep everything on processor caches that are tied to the CPU clock this will not matter. More serious programs will need to hit main memory and this will prevent linear scaling with a faster clock.
Re:Double standards (Score:5, Informative)
the clock speed a CPU is capable of is irrelevant for most people because the limiting factors are heat and power.
It's also getting limited by the speed of light, or for silicon, the speed of electronic signals through silicon that's about 2/3 the speed of light (with variation depending on what's going on in the silicon).
For understanding, 6 GHz frequency means 0.166 ns, so light travels about 4.79 centimeters, electric signals travel about 2.9 centimeters during that time. The electric signal bouncing all around the circuitry of the chip needs to flip all the transistor switches by traveling less than 3 centimeters.
The 8 GHz frequency means 0.125 ns, light travels 3.75 centimeters, electric signals travel about 2.25 centimeters.
For some of the winding circuitous paths around the chip and the large numbers of transistors that need to be hit, at these frequencies one clock tick's electric pulse can still be spreading around the chip when the next clock tick is triggered. One tick happens before the electric signal can travel the distance of the chip. So not just heat and power, but also the distance that must be traveled around the chip.
Re: (Score:2)
So not
Re: (Score:2)
For some of the winding circuitous paths around the chip and the large numbers of transistors that need to be hit, at these frequencies one clock tick's electric pulse can still be spreading around the chip when the next clock tick is triggered.
Which is perfectly OK.
As long as some other part of the chip cannot engage in superluminal communication, signal causality is preserved.
Re: (Score:2)
Winters DO get cold where I live. It will be nice to have a warm PC under my desk....
Grain of salt, more details required (Score:2)
Re: (Score:2)
Heard somewhere at Intel (Score:2)
Meaningless stunt (Score:2)
Buy AMD, get more, pay less. Also, Intel has a history to lie with these benchmarks and to demonstrated "products" that you later cannot buy or buy only after along wait and in very small numbers.
Re: (Score:2)
Re: (Score:2)
12600K base power: 125 watts. What an air fryer! My entire 8 core Ryzen system draws 16 watts at the wall under light load.
Re: (Score:2)
Re: (Score:2)
Sure it is.
Re: (Score:2)
Have you forgotten that you need more than the bare CPU? Apparently so.
Re: (Score:2)
Please focus on security (Score:1)
Re: (Score:2)
Take it with a grain of molten salt (Score:2)
Did we forget it's also about ops per cycle? (Score:2)
Did we forget it's also about operations per clock cycle, memory bandwidth, and I/O throughput?
The next CPU from Intel will be 1 THz! (Score:2)
And require a small nuclear power plant to power it. Anyone can write a number and put it on a slide. Show us the silicon (or diamond or whatever).
"and a 41% gain in multi-threaded" (Score:2)
One would surely hope so when it has 50% more cores!
spectre (Score:2)
Intel also claimed that Raptor Lake will have a 15% gain in single-threaded performance and a 41% gain in multi-threaded,
that'll help them overcome the performance losses caused by speculative mitigation in the linux kernel, then, eh?
Keep focusing on Ghz Intel (Score:2)
while your ARM competitors eat your lunch.
is this what we need? (Score:4, Interesting)
I mean, I like fast computers, don't get me wrong... but how many people actually have a good use for a 100W+ desktop CPU these days? Especially as GPUs are the dominant need for modern gaming, is a 6-8Ghz CPU actually all that valuable? No doubt next gen gaming engines are going to be even more GPU dependant and less CPU dependant.
Meanwhile (puts on flame suit) Apple is putting out competative chips at a fraction the power draw and qualcomm is eyeing that same market.
I just feel that making more performance/W is the better route for 99% of consumers. Obviously the gaming industry isn't jumping over each other to get on Apple silicon, but I think it is actually an interesting platform for them considering commonalities with the booming mobile gaming industry. If Qualcomm puts out something in the ballpark of Apple but 'open' so to speak, that could be a very interesting direction.
Intel seems to just be trying to pour on the brute force. I admit gen12 performace+efficiency cores contradicts that a bit, but really they are trying to hold onto the idea that you need a big fat CPU in an era where GPUs are dominant and modern software is more and more multi-threaded. AMD seems a little more keen on this than intel..
power requirements ? (Score:2)
no mention of the power consumption of these things...
At how many watts? (Score:2)
If you build a gaming PC based on Intel and any of the new gen GPUs, you will be very close to exceeding the power limit of a standard US wall outlet. You already cannot have 2 gaming PCs in a single room if the room is only one circuit. This is the wrong direction. I don't want to hear nonsense about clock speeds - I only care about MIPS/watt.
I'll have bacon with that, please. (Score:2)
I wonder how warm will they get. Would you be able to cook eggs on bacon on your CPU while it's in use?
World Record in Ghz... is 8.7Ghz and it's an AMD (Score:2)