When will we get a CPU that is 5.0 GHz BASE and not 4.9 GHz Max Boost?

Hulk

Supreme [H]ardness
Joined
Nov 4, 2005
Messages
6,216
Just like the title says, when will we be able to buy either a 4,8,16, etc. core CPU (either Intel or AMD) that has a base clock speed of 5.0 GHz?

It just seems like we've hit the wall as far as processor speed goes and the Big Two just keep adding more and more cores. Next year they will roll out a 32 core CPU with a max boost of 4.9 Ghz.

You would think by now the CPUs would be in the 8.0 GHz range.

Thoughts?
 
NetBurst

With this microarchitecture, Intel looked to attain clock speeds of 10 GHz,[7] but because of rising clock speeds, Intel faced increasing problems with keeping power dissipation within acceptable limits.

Different architectures, same problems. We simply cannot adequitely cool the processors enough to reach such high clocks. And now, with processes being so small, they have the additional problems of cross-talk and leaking voltage, among other things.

Long story short, it'll take a different process, with different materials, to even approach 6GHz. It probably won't happen soon, if ever. We may even change the type of math we use on CPUs before it happens. Who knows?
 
  • Like
Reactions: Hulk
like this
It just seems like no one is talking about how we've hit a wall with CPU speed. I was reading up on this and came across "organic computers" (wetware computers) which are supposed to have live neurons in them to hopefully break the 5.0 GHz threshold. That is some wild stuff but it seems like something from the future. What you've said makes sense why they just keep adding more and more cores instead of concentrating on higher base clock speeds.

Also, correct me if I am wrong but when these chips say 4.9 GHZ Max Boost it's not across all the cores, just a certain percentage. So a 16 CPU Core will have max boost only on like 4 cores and not 16?
 
Also, correct me if I am wrong but when these chips say 4.9 GHZ Max Boost it's not across all the cores, just a certain percentage. So a 16 CPU Core will have max boost only on like 4 cores and not 16?
Generally, yes. Usually if they support a certain speed on all cores, they'll use the separate term "All Core Boost," or something like that.

AMD, Intel, and other smaller competitors are also experimenting with different architecture designs, which may improve performance despite clock speed remaining the same or decreasing. For instance, stacked chips, "fabric" bus, additional cache, etc, as well as improving on ARM and other non-x86/64 designs.
 
  • Like
Reactions: Hulk
like this
It just seems like no one is talking about how we've hit a wall with CPU speed. I was reading up on this and came across "organic computers" (wetware computers) which are supposed to have live neurons in them to hopefully break the 5.0 GHz threshold. That is some wild stuff but it seems like something from the future. What you've said makes sense why they just keep adding more and more cores instead of concentrating on higher base clock speeds.

Also, correct me if I am wrong but when these chips say 4.9 GHZ Max Boost it's not across all the cores, just a certain percentage. So a 16 CPU Core will have max boost only on like 4 cores and not 16?

Try closer to max boost on 1 core under ideal thermal conditions on the higher end and only for a limited time.

Most motherboards let you override that though.
 
  • Like
Reactions: Hulk
like this
It just seems like no one is talking about how we've hit a wall with CPU speed.

Most of what's to be said, has been said. We've been hitting the wall since pentium 4 times. Advances in process and technique get us closer to the wall now, but it's the same issue. Higher clock frequency increases transistor power by itself, but also because higher frequency circuits have narrower timing windows, so you need transistors to switch faster which takes higher voltage and then they use more power. All that power turns to heat, and you have to get it out of the chip, which isn't easy, etc. Of course, if you push the voltage too high, you damage the transistors.

So, the way forward is more compute per cycle, not more cycles per second. Multicore is one way, but also superscalar operations, more than one adder, etc. Apple's M1 went significantly wider than others here; more units of almost everything, plus ARM has weaker memory ordering requirements so it's easier to have a larger window for out of order execution; it's harder to run a wider processor at high clock speeds, but Apple was never going to put good enough cooling to run at high clock speeds for long anyway. AMD and Intel would have a hard time pushing a processor design that tops out at 3.2 GHz, regardless of performance because of the Megahertz Myth, but Apple doesn't care about that.
 
But
Most of what's to be said, has been said. We've been hitting the wall since pentium 4 times. Advances in process and technique get us closer to the wall now, but it's the same issue. Higher clock frequency increases transistor power by itself, but also because higher frequency circuits have narrower timing windows, so you need transistors to switch faster which takes higher voltage and then they use more power. All that power turns to heat, and you have to get it out of the chip, which isn't easy, etc. Of course, if you push the voltage too high, you damage the transistors.

So, the way forward is more compute per cycle, not more cycles per second. Multicore is one way, but also superscalar operations, more than one adder, etc. Apple's M1 went significantly wider than others here; more units of almost everything, plus ARM has weaker memory ordering requirements so it's easier to have a larger window for out of order execution; it's harder to run a wider processor at high clock speeds, but Apple was never going to put good enough cooling to run at high clock speeds for long anyway. AMD and Intel would have a hard time pushing a processor design that tops out at 3.2 GHz, regardless of performance because of the Megahertz Myth, but Apple doesn't care about that.
Agree. 8.0 Ghz base is not for possible. Overclockers get 8.0 Ghz but they use liquid cooling extreme... Intel and Amd switched at multi core technology because Increasing the Ghz is more and more difficult
 
IBM's Power 6 (2007) and Power 8 (2014) architectures could hit 5GHz. Most architectures top out a bit lower because the tradeoffs needed to reach such high clock speeds end up resulting in somewhat lower overall performance either on an absolute or per watt scale.

https://en.wikipedia.org/wiki/POWER6
https://en.wikipedia.org/wiki/POWER8
Bulldozer too. It wasn't bad at those clocks either, but it sure was hot. Edit: think some got to 5.4, but that was late in production and marketed as an OCd chip, iirc.
 
Last edited:
Just like the title says, when will we be able to buy either a 4,8,16, etc. core CPU (either Intel or AMD) that has a base clock speed of 5.0 GHz?

It just seems like we've hit the wall as far as processor speed goes and the Big Two just keep adding more and more cores. Next year they will roll out a 32 core CPU with a max boost of 4.9 Ghz.

You would think by now the CPUs would be in the 8.0 GHz range.

Thoughts?
whe hell freezes over because ambient temperature will be adequtly cool enough
 
Bulldozer too. It wasn't bad at those clocks either, but it sure was hot. Edit: think some got to 5.4, but that was late in production and marketed as an OCd chip, iirc.

The FX9590 was a heat monster that made even the highest clocked Netburst CPUs look cool as ice.

I've only ever encountered two of them in the wild, and both were so unstable at "stock" 4.7 GHz speeds that they had to be dialed back to FX8370 clocks. When they were at 4.7, they'd overload even 360mm rads.
 
Base vs boost is by and large a marketing and legal matter. For example, the 12900K is for all intents and purposes a 5GHz, 8-core processor, but sustaining 5GHz requires good cooling and good VRMs. If Intel marketed it as a 5GHz part, there would be butthurt and class action suits from customers who don't see the advertised clock speeds due to cheap VRMs or insufficient cooling (remember the Ryzen 3000 fiasco?). By marketing it as an "up to 5GHz" part, Intel can achieve strong real-world benchmark scores, especially with reviewers who are normally equipped with top of the line boards, without exposing itself to the liability and backlash of marketing a part that will not perform as advertised without the correct set of ecosystem components.

A lot of it does boil down to exotic cooling as well: a lot of CPUs can do 5GHz all core at -100C but it isn't really practical to put a cascade in your daily driver...
 
IBM zEC12 had 5.5ghz base clock back in 2012. I think the chip companies realized they could get more performance with less power by not just pumping up clocks. I'm sure they could make higher hz processors, but they wouldn't perform as well so there's no point. I know even 20 years ago we could get ring oscillators running at 10-12ghz, so if you made a simple enough cpu you could get it to in the 20-30ghz range these days, but it would get smashed by a more complex cpu operating at 1ghz, so it's a foolish thing to chase.
 
not too unlike the Pentium 4. they couldn't crack 5ghz and ran hot, so they decided to stick two CPU dies on one package and the Pentium D was born and it ran hot but, with twice the threads lol.

EDIT: also amd came out with 64 bit around that time and then later the core duo was born.
 
And then Sandy Bridge came next and could overclock to (almost) 5.0ghz. And that number has been the benchmark for nearly a decade, despite cpus increasing in performance dramatically since then. There is a reason Intel ramped up the clock rate and there is a reason OP is asking about 5.0ghz. People remember simple numbers, it's effective marketing.
 
And then Sandy Bridge came next and could overclock to (almost) 5.0ghz. And that number has been the benchmark for nearly a decade, despite cpus increasing in performance dramatically since then. There is a reason Intel ramped up the clock rate and there is a reason OP is asking about 5.0ghz. People remember simple numbers, it's effective marketing.

And then Ivy Bridge came and only a select few chips could actually get to 5.0 ghz, which gave rise to Silicon Lottery. Most Ivy Bridge CPUs could only hit 4.6 ghz or so. Subsequent generations have all generally overclocked in the 4.5-4.8 ghz range before factory boost became commonplace.

Then again, OP was talking about base clock speeds hitting 5.0 ghz, which hasn't been done well on any CPU to date.
 
Back in the day MHz jumped by the month and we were all at a fever pitch. More MHz's!!!
There is still a bit of that deep in our reptilian brain. We like to go faster planes/cars/CPU's! It is in our competitive nature and it appears we have reached a limiting factor of silicon. Something will come along that allows higher sustained cycles eventually, or a old skool 1GHz CPU with 1000x the IPC?
 
And then Sandy Bridge came next and could overclock to (almost) 5.0ghz. And that number has been the benchmark for nearly a decade, despite cpus increasing in performance dramatically since then. There is a reason Intel ramped up the clock rate and there is a reason OP is asking about 5.0ghz. People remember simple numbers, it's effective marketing.
Westmere could just bout do 5ghz before that
 
The infamous "black ops" part (X5698) had a stock speed of 4.4 GHz with a 4.53 GHz boost, though it was only a 2C/4T part.
Vs a 4790k with a stock 4ghz and boast of 4.4ghz.

They x5687 could do 5 but was abit of a unicorn. The 6 core chips could oc to 4.4 on a good day as well.
 
And then Sandy Bridge came next and could overclock to (almost) 5.0ghz. And that number has been the benchmark for nearly a decade, despite cpus increasing in performance dramatically since then. There is a reason Intel ramped up the clock rate and there is a reason OP is asking about 5.0ghz. People remember simple numbers, it's effective marketing.
the more cores the harder it is OC as one of the bunch will be the weakest link but, then they come up with A synchronous clocks, i suspect the issues now is the transistor count. there are so many it's hard to hit higher speeds without a sub circuit like an ALU or FPU inside the core not able to meet the demand. The higher you pump those clocks the lower the yield and the higher the wattage. they can bin the lower quality units but, theres a limit till your celling al lcelerons and athlons and hardly making enough I series or X series to make some real profit.
 
Back
Top