Intel Hits 50 Years and Its CPUs Hit 5.0 GHz (venturebeat.com) 161
Intel will turn 50 next month, so to celebrate that, its CPUs are hitting 5.0 GHz for the first time, it said. At Computex event in Taiwan this week, the chipmaker announced the limited edition 8th Gen Intel Core i7-8086K processor, the first-ever CPU from the company with a 5.0GHz turbo frequency. From a report: Intel, of course, is the world's biggest chip maker, and its fortunes are wedded to the success of the personal computer. "As we transition to the data-centric era, the PC remains a critical facet of Intel's business, and it's an area where we believe there are still so many opportunities ahead," Bryant said. "Today, at Computex in Taipei, I shared our vision for the future of the PC and introduced a wide range of new technologies that will help us and the broader ecosystem make this future a reality. One that transforms the PC from a simple computer into a platform that can power every person's greatest contribution."
Oracle already has a 5.0 GHz chip on the market (Score:5, Insightful)
"SPARC T8-2 Server Specifications
ARCHITECTURE
Processor
Thirty-two core, 5.0 GHz SPARC M8 processor
Up to 256 threads per processor"
Re:Oracle already has a 5.0 GHz chip on the market (Score:4, Informative)
Re:Oracle already has a 5.0 GHz chip on the market (Score:5, Insightful)
This is true, but the claim verbatim is 'first-ever 5.0 ghz', so it's correct to call them out on not hitting the front.
Re:Oracle already has a 5.0 GHz chip on the market (Score:5, Informative)
If that's verbatim, where is it?
The article says "first-ever CPU with a 5.0GHz turbo frequency" which was crafted to be correct on a technicality (it's not the base clock rate).
The summary says "its CPUs are hitting 5.0 GHz for the first time"
And Intel doesn't make SPARC chips, so that's also correct.
Re:Oracle already has a 5.0 GHz chip on the market (Score:4, Interesting)
WRONG!
AMD gave us the FX-9590 with 4.7 GHz base and 5.0 GHz turbo clocks FIVE years ago.
Re: (Score:2)
But they call the max speed a "boost" clock (even if they call the feature in general "turbo core"). It's literally all semantics.
Re:Oracle already has a 5.0 GHz chip on the market (Score:4, Informative)
https://hardware.slashdot.org/... [slashdot.org]
Re:Oracle already has a 5.0 GHz chip on the market (Score:5, Informative)
There isn't a direct quote on this point, so it may be the writer that said this wording, not the person being quoted. Meanwhile, the title is accurate, with "its CPUs hit 5.0 GHz", not it is the first to hit 5.0Ghz. As mentioned, the Sparc has had 5.0ghz, but AMD also has: https://www.anandtech.com/show/8316/amds-5-ghz-turbo-cpu-in-retail-the-fx9590-and-asrock-990fx-extreme9-review.
Re:Oracle already has a 5.0 GHz chip on the market (Score:5, Informative)
Intel: https://newsroom.intel.com/edi... [intel.com]
"the first Intel processor with a 5.0 GHz turbo frequency"
Intel actually qualifies their statement and all the reporters parroted it without the qualification. So basically, just another news day.
Venturebeat: https://venturebeat.com/2018/0... [venturebeat.com]
"the first-ever CPU with a 5.0GHz turbo frequency, said Intel’s Gregory Bryant"
CNET: https://www.cnet.com/news/inte... [cnet.com] ...
"the first-ever CPU with a 5.0GHz turbo frequency."
Re: (Score:3, Informative)
"Intel will turn 50 next month, so to celebrate that, its CPUs are hitting 5.0 GHz for the first time, it said. At Computex event in Taiwan this week, the chipmaker announced the limited edition 8th Gen Intel Core i7-8086K processor, the first-ever CPU from the company with a 5.0GHz turbo frequency. "
Noone is claiming "first ever", they are claiming " Intel's first ever".
Re: (Score:2)
No that's not the claim, where did you get that idea?
It's the first Intel processor with a specified 5GHz clock frequency (turbo) but of course not the first 5GHz processor (see IBM POWER 8 for that). So what is claimed that this is the first Intel processor at 5GHz clock frequency - which is correct.
Re: (Score:2)
Re: (Score:3)
Re: Oracle already has a 5.0 GHz chip on the marke (Score:2, Informative)
POWER6 was at 5 GHz in 2008
Re: (Score:2)
(facepalm)
Not even the summary says this is the first 5GHz CPU ever.
Re: (Score:2)
Skylake again (Score:3)
It's the same Skylake uArch which debuted three years ago and naturally this particular CPU is affected by both Meltdown and Spectre. It's still an accomplishment though since it's the first ever consumer CPU to run at a such an insane clockrate.
Oh, and it will be available in very limited quantities.
Re:Skylake again (Score:5, Funny)
and naturally this particular CPU is affected by both Meltdown and Spectre
Yeah, but it does it faster. So....there is that.
Re: (Score:2)
The more things change, the more they stay the same - Here's my post reacting to a similar 1.0GHz CPU announcement nearly 20 years ago:
https://hardware.slashdot.org/... [slashdot.org]
Re: (Score:2)
I'm sure they shrunk it down to 7nm to hit 5GHz. And the very limited quantities is probably because they already anticipate production problems.
Re: (Score:3)
Re: (Score:2)
Do they not create a die for an existing CPU when testing a new process? Seems like you wouldn't want to troubleshoot a new design and a new size at the same time. This might be just a die created for R&D.
Re: (Score:3)
This answers my questions. (Score:2)
This still has the NSA-required flaws, so they are unlikely to actually 'fix' any of the pending issues until people stop buying servers.
I'm done with intel for a while.
Re: (Score:2)
5GHz is only the turbo frequency though, meaning it can only do it on one or two cores and only for a limited amount of time without extreme cooling.
Given that it's crippled by Meltdown I think I'll take much cheaper Ryzen or Threadripper with more cores and especially more PCIe lanes.
Re:Skylake again (Score:5, Informative)
Re: (Score:3)
You will almost certainly be able to buy two or even three nicely outfitted Threadripper 2 workstations for the cost of that CPU alone. Intel sells 28-core Xeons now, but they aren't anywhere near 5 GHz, and they cost about $10k each.
Re: (Score:2)
Re: (Score:2)
To be honest, nothing Intel can currently touch this new chip either, because they won't be selling it for about six months.
If your workload is that parallel, there's a reasonable chance you can farm it out to multiple machines.
Re: (Score:2)
To be honest, nothing Intel can currently touch this new chip either,
How so? Since this is just a unlocked 28 core Xeon re-purposed for workstations. A chip you can buy today if you have $10K burning a hole in your pocket.
Re: (Score:2)
Intel's current 28-core Xeon runs at 2.1 GHz (3.8 GHz turbo), not 5 GHz. Intel said that this CPU would be released in Q4 of this year. Take off your fanboi hat for a moment and pay attention.
Re: (Score:2)
From your link:
"EDIT: According to this image we sourced from Engadget's compressed keynote video on YouTube, Intel apparently was running some sort of closed-loop cooling that required insulating material around the tubing. This could be a multi-stage phase cooler (sub-zero cooling), or possibly a more mundane water chiller, under the table."
Either way, it's probably not even close to ready for deployment in actual servers at 5 ghz across the board if they need a cooling system with insulation on it to get
Re: (Score:2)
Re: (Score:2)
Oh, and it will be available in very limited quantities.
I guess that's a good thing considering there's going to be very limited demand.
Re: (Score:2)
I mean, except for the AMD 9590 from June 2013...
https://www.amd.com/en/product... [amd.com]
Re: (Score:3)
Re: (Score:2)
That's still a substantial improvement over the fastest RAM available when the PIII 1.4 came out, 133 MHz SDRAM. If both the CPU and RAM clock cycle speeds had increased at the same rate, we'd have 16.8 GHz CPUs to go with the 1600 MHz clocked RAM.
To make it worse, the speed increases we did get also came with an increase in pipeline length, which makes the CPU substantially slower whenever branch prediction fails and the pipeline has to be refilled. So the 5 GHz of tomorrow isn't going to run all code 3
Re: (Score:2)
Pretty sure 3200MHz DDR4 runs at 400 MHz, not 1600. The "effective" clock speed of 3200 MHz comes from 4 bytes per access times two (since it acts on both the rising and the falling clock edge).
Re: (Score:3)
CPU and memory clock rates used to be identical, then CPU speeds grew way more rapidly than memory speeds, leading to the misnomer: memory multiplier. So the upside of what you note is that we're slowly moving away from ridiculously high multipliers. Yes, the end game would be processing integrated with memory.
Re: (Score:2)
It's still an accomplishment though since it's the first ever consumer CPU to run at a such an insane clockrate.
It does not sound insane to me. I mean, back in 2002, we had Pentium III running at 1.4 GHz, and not long after, the P4 3.06 GHz.
16 years later, and we haven't been able to double the PIII clock rate twice or the P4 clock rate once?
For perspective, in the same time span, we have gone from 133 MHz SDRAM and 533 MHz RDRAM to 3200 MHz DDR4. That's a far more impressive clock rate increase. :p
If this trend continues, we'll soon have to offload cycle dependent calculations from the CPU to the RAM controller because it's going to be faster...
While it's true that Intel hasn't doubled the clock rate yet of a P4, efficiencies gained in additional cache, predictive algorithms, out of order processing, faster memory, faster and larger PCI lanes, multi-core processors, etc. has improved performance by multitudes.
The one thing that I have noticed is that with the rise in mobile computing, it seems that building more powerful CPUs has taken a backseat to building more power efficient CPUs.
Re: (Score:2)
The one thing that I have noticed is that with the rise in mobile computing, it seems that building more powerful CPUs has taken a backseat to building more power efficient CPUs.
Not just mobile computing, but data centers where parallel processing is more important than raw speed, and the cost of a year of electricity multiplied by several thousand CPUs makes a difference.
But when you need linear computing power, the advances haven't been all that great over the last twenty years. That 3.06 GHz Pentium 4 from 2002 isn't all that much slower for that than today's CPUs.
Re: (Score:3)
In most consumer applications it isn't the speed of the memory or cpu that causes slowdown. It's the I/O. When I'm doing "real work" on my desktop my cpu hardly breaks 20% and I'm waiting on I/O to catch up. An this is with a Samsung 960 Pro NvME card.
With games its GPU. My cpu barely breaks 30% even with the most demanding game.
Faster CPU's are nice but they are not what determines overall computer power any more.
Single-thread performance doubling: 7.5 years! (Score:2)
PassMark - CPU Mark Single Thread Performance [cpubenchmark.net]
AMD Athlon XP 2800+ @ 2.25 GHz / rel. October 1, 2002 / Score: 627
Intel Core i7-8700K @ 3.70GHz (4.7GHz turbo) / rel. October 5, 2017 / Score: 2708 (highest-scoring processor for single thread performance as of June 5, 2018)
Single thread performance ratio: 4.01
**Single thread performance doubling time: 7.5 years**
Note: it's not clear what the best-performing processor was in the early 2000s, the performance doubling time may be even greater.
Moore's law ceased to
Re: (Score:2)
AMD Athlon XP 2800+ @ 2.25 GHz / rel. October 1, 2002 / Score: 627
P4 3.06 which came out at the same time scores 656 according to that table.
Yay! Progress! (Score:4, Insightful)
Re:Yay! Progress! (Score:4, Insightful)
Compatibility is a nice feature, and it's a testament to the design that they could remain compatible for so long.
Re: (Score:2)
While true, computer chips today would likely be significantly higher-performing if Motorola had came ahead in the early microprocessor days, with its 68000 series processors. A good fraction of modern x86 designs is dedicated to instruction translation. The Motorola 68000's instruction set was much more forward-looking, and requires less translation. If the research and money that had been dumped into x86 had instead been dumped into improving the 68000, computing might be pretty different today.
But tha
Re: (Score:2)
Turbo frequency (Score:5, Funny)
Will I need to hit the turbo button on the front of the PC to get this 'Turbo frequency'?
Re: (Score:2)
Re:Turbo frequency (Score:5, Informative)
I'm both old enough to remember those and young enough to remember how I thought it was a dumb idea. Thankfully processors change clock speed based on actual load now.
Turbo button was for legacy compatibility, A lot of old games used cycle based timing, so on a faster cpu the game ran too fast. Hit the turbo button to slow down cpu and voila, scaling based on load wouldn't work for that use.
Re: (Score:2)
Re: (Score:3)
Lots of games did a lot of similar things, right into this decade.
Go run the original C&C or Red Alert on a PC. Despite being available for Windows 95 (and thus much faster chips than anything a Turbo button was designed to cope with), you still have to play with "scroll speed" at the very bottom and "game speed" somewhere about half-way (top is way too fast, and the graduations are enormous between settings)
The timing was nowhere close to actually being based on wall-clock time, despite things like pr
Re: (Score:2)
A while ago, Origin (I think they owned it at the time), released the original C&C for free. So if you're getting nostalgic, you can get your fix for free.
With that said, the game did run much better on my old 700MHz Celeron system than it did my more modern system, though part of that may have been because that computer has Windows 98SE on it. As it was, I originally played the game on a Pentium 75, and while that was a long time ago, I don't remember any performance issues.
Re: (Score:2)
That's the EXACT re-release I'm talking about.
No, it still needs patching to even run on some systems, using the some 10+ year old third-party patches, and still has the speed problems.
Re: (Score:2)
This is true, but legacy compatibility also meant hardware support. Peripheral hardware would frequently not work at higher speeds either, including hardware supported in the BIOS.
Re: (Score:2)
Similarly, the old SCO Unix (back when SCO was merely stodgy and not evil) had a timing loop in the Adaptec 1540 and 1740 drivers, and would fail on boot on a 486/DX-66 (and probably others, but I'm just going from personal experience)
The fix was to take the system out of turbo mode, boot from the floppy, patch the driver image; reboot in turbo mode and install. Patch the driver image on the hard drive and relink the kernel; and then reboot in turbo mode.
The key thing is that the turbo switch was necessary
Finally! (Score:2)
In 2000, when I was still getting used to saying Gigahertz instead of Megahertz, and they'd bounded up from triple digits to quadruple digits in the span of just a couple of years, it seemed like 3, 4, and 5, GHz processors ought to be just around the corner. I can remember being mystified and disappointed as the path to 2 GHz became increasingly asymptotic. That had been the key metric for so long, watching computer manufacturers re-spin their marketing to talk about other features, or start plugging dual
Re: (Score:2)
Oh, they got up there quite quickly. But they had to make some bad architectural shifts to do it (the P4), which actually hurt overall performance. When they finally recovered from that, the march became a lot slower.
Re: (Score:2)
It was only a few years between the 1 GHz PIII and the single core 3.8 GHz P4, which held the x86 clock speed crown for close to decade. Of course, that wasn't the plan. Intel knew the P4 wasn't very efficient, but they thought they could ramp it up to 10 GHz and beyond so it wouldn't matter than the initial sub-2.0 GHz P4's weren't that impressive next to the final generation PIII's. Obviously that didn't work out for Intel.
I though we got off the mhz (Ghz) myth. (Score:3)
Is this the early 2000's where we were all drooling over the faster hertz speed of the clock.
We have been parallelizing the chips and software for over a decade now to reach meaningful speed improvements, while not really caring much about the clock speeds. This approach actually has been a good thing, it allowed great improvements in mobile chip design which cannot draw tones of power and doesn't need a radiator to keep it from melting itself.
Re: (Score:3)
Re:I though we got off the mhz (Ghz) myth. (Score:4, Insightful)
The higher the clock speed, the higher the wasted power to heat. And if the chips get to hot, they work faulty anyway, the smaller they get, the more faulty they get, too.
Re: (Score:2)
More cores also waste power to heat.
Re: (Score:2)
But far less than a higher frequency, and they can be disabled if not in use.
Double the frequency, square the power loss.
Double the cores, double the power usage. Keep in mind: in case of frequency increase it is simple loss, the power does not anything useful, in case of core doubling, the core is performing calculations.
Re: (Score:2)
Should also add that faster frequencies means faster switching, which means parasitic capacitances have to be charged faster which is generally achieved by higher voltage. Double the voltage, quadruple the power loss.
Re: (Score:2)
Wrong. Power dissipation is proportional to frequency if the voltage is unchanged.
Re: (Score:2)
You must be living in an interesting universe then:
https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: I though we got off the mhz (Ghz) myth. (Score:2)
Re: (Score:2)
Silicon is the limiter, other elements heat up less with current going through them and could potentially reach speeds up to 100ghz AFAIK.
Re: (Score:2)
From 2010:
https://www.google.co.uk/url?s... [google.co.uk]
Re: (Score:2)
Yea, but now I have to fire up my gas heating in winter all the time! You insensitive clod!
Re: (Score:2)
while not really caring much about the clock speeds
Speak for yourself. I'm still waiting for my 10GHz CPU. Decades of exponential clock speed improvements were a glorious thing.
I'm typing this comment on a machine from 2008, and the only thing it really needs is more RAM. Of course I don't use it for gaming, as that would be a graphics card issue.
It used to be desktop computers were considered outdated after 3 years.
Bah! (Score:2)
Re: (Score:2)
They missed an opportunity (Score:5, Funny)
I really wish they'd built a chip that ran at 4.77 GHz...
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
640 GB should be enough for anybody, then...
https://quoteinvestigator.com/... [quoteinvestigator.com]
Some specs since linked article sucks (Score:2)
Re: (Score:2)
average cycles per second per second (Score:3)
Obviously the growth in clock speed has been exponential (moores law) and goes in major steps process and design changes, but for fun, the linear average increase in clock speed since the launch of the intel 4004 in 1971 (740khz) to the present top line chips (~4.3 Ghz) is 3 Hz per second. Or 3 more cycles per second per second.
Re: (Score:3)
no, clock speed the one thing that flattened out years ago, about 2004
Re: (Score:2)
Intel top line CPU speeds have gone from 3.8 to 4.3 ghz in the last 10 years. That is still 2 hz per second increase over the last 2 years.
Re: (Score:2)
of course the important thing is how much work a CPU can do, a phrase that could give a notion of that graph's increase over the years would be more meaningful
Re: (Score:2)
OK, it's more sigmoidal and not exponential. This is actually true of all exponential growth in real life, they eventually hit another limit that stops the growth. Google bacterial growth curve for an example.
Thermal Loads (Score:3)
The amazing thing, then, is not simply that Intel have managed to ship a 5GHz part, but they have done so whilst essentially keeping the thermal profile of the chips more-or-less uniform for a good part of the last few years. In some ways this thermal efficiency is even more impressive than the outright clock speed; it talks to the materials science, packaging design and overall cooling effectiveness, that we've now come to expect from our current crop of processors.
Biggest Chip Maker (Score:2)
Samsung is now the world’s biggest chipmaker. [theverge.com] Happy birthday, Intel!
I get the point of the article.. (Score:2)
Re:5GHz too late... (Score:5, Funny)
Yep, they held them back at least 5 years so it would coincide with their anniversary. Idiots.
Re: (Score:3)
People have been running Intel processors at 5GHz for years. They may have needed aftermarket cooling solutions, but it's been something that's possible for quite a while. I guess the question is how high people will be able to overclock this 5GHz CPU. If it costs more than the current 8700k and doesn't actually provide any level of overclocking, then I don't really see it as big news.
Re: (Score:2)
8086K will allow overclocking, that's what the 'K' means.
Re: (Score:2)
What’s so special about this processor?
This processor is Intel’s first 5Ghz out of the box consumer desktop processor, featuring 6 cores and 12 threads, and is unlocked for overclocking*.
*Altering clock frequency or voltage may damage or reduce the useful life of the processor and other system components, and may reduce system stability and performance. Product warranties may not apply if the processor is operated beyond its specifications. Check with the manufacturers of system and components for additional details.
Re: (Score:2)
Yes, overclocking means operating out of spec.
News at 11.
Re: (Score:2)
They probably tried 10GHz on a Pentium IV, but melted down the facility in the process.
Re: (Score:3)
Re: (Score:2)
The engineers and the marketing department got locked in a room and battled it out to the death. I'll let you figure out who won.
Tricky question. The Marketing people won. At least in the US. Netburst was a huge engineering failure.
Luckily, some folks at Intel Israel had been working on making the PIII more energy efficient. When Intel's fortunes looked very bleak, somebody noticed this work and from there the Core line was born and AMD got set back a decade in terms of competitiveness.
P4 was the end of
Re: (Score:2)
Re: (Score:2)
The fallacy you were looking for is "post hoc ergo propter hoc" (after this, therefore because of this), not "correlation implies causation".
Re: (Score:2)
You had a 5GHz P4 in 1998? Makes me wonder why as a teen I spent all my money on a P3 600 Mhz in 2000...