Intel Announces 8 Core I9-9900KS: Every Core at 5Ghz, All the Time
anandtech.com
anandtech.com
Then there is the "Intel" benchmarks as usual [1] on GPU. Trying to suggest the 2 CPU were both running at 25W TDP to give a "fair" comparison, without mentioning the Ice-Lake U CPU were running with 50% more memory bandwidth vs the AMD Ryzen. And we know Graphics Benchmarks do depend a lot on memory bandwidth. The memory used was somehow mentioned in Toms or Other Sites but not Anandtech. ( Although none of them had mentioned the bandwidth difference, it was up to the reader to work them out )
Anyway none of these Consumer CPU upgrade interest me anymore ( Although any improvement to iGPU would be great ) I am eagerly waiting for a 2S - 128 Core EPYC 2 on a Server or AWS to play around with it.
[1] https://www.anandtech.com/show/14405/intel-teases-ice-lake-i...
Edit: And the lesson here, never trust a single news source. Always have a few option opened and fact check yourself. ( If you have the time )
What about the Ryzen 3000 line up? The benchmark leaks make it seem like it is going to be a huge improvement and AMD isn't susceptible to Zombieload.
Looking forward to a less complex architecture even if it means cutting me off at the knees with execution speed (for a few years):
RISC-V
Because the future is looking increasingly in-order high-core-count with hard partitioning between security contexts.
I feel like the entire "PC enthusiast" review space has dropped the ball on hardware vulnerabilities. Reevaluating performance between microcode and OS patches is an afterthought, and when a new CPU hits the market the numbers are presented without the obvious disclaimer that these performance gains may evaporate within months.
Some even perpetuate the "only relevant to datacenter" myth despite the fact that security researchers have shown to be able to exploit these vulnerabilities with JavaScript in the browser.
Edit: I took a look and it seems they don't ask/record that info. I apologise for the offence this idea seems to have caused someone.
'PC enthusiast' is such a blanket term to start with, so applying a blanket statement to such a group is obviously doomed to failure from the very start.
From what i've seen web browser teams have taken the recent risks extremely seriously - I have sure had a worse track record of infection via downloading and installing software versus visiting sites with JS running
Not really. They didn't even properly apply band-aids.
Chrome and Firefox disabled number of features, that allow Javascript code to create high-precision timers. This makes exploiting slightly more difficult, but the gaping hole is still there — there is infinite number of ways to create a high-precision timer, just not as obvious as closed ones.
Chrome has enabled Site Isolation on desktop, but haven't done it on Android (presumably, because of associated increase in memory consumption).
All major browsers still allow Javascript to run in background, create CPU threads and consume unrestricted amount of CPU time. I don't believe, that any of them have mounted instruction-based defenses (lfence etc.), but I may be mistaken here.
Even though I'm personally in the fortunate position not to have any reasonable exposure to these vulnerabilities, I wouldn't be particularly worried even if this wasn't the case. It's been well over a year since Meltdown and Spectre came out and there still hasn't been a single case of anyone successfully using these vulnerabilities to productive ends in the wild that I know of. Obviously, cloud computing vendors need to pay attention and there are legitimate business concerns that are affected by this, but insofar as personal computing goes? If people persist in the ridiculous notion that constantly running completely arbitrary code in naive sandboxes is a great idea, I imagine there will eventually be issues, but so far the issue seems to be vastly overblown in the popular media.
Considering we are referring to attacks that can bypass your PC's security, "prudence" is a better word than hysteria.
Yes, if they are left alone, it is the "end of the world".
They can be used to make any modern OS and browser as full of holes as Windows 98.
That statement can be made about any vulnerability whatsoever. The merit of any mitigation can only be determined by a cost/benefit analysis that takes into account the potential impact of the vulnerability as well as the very real costs of mitigating it.
> Yes, if they are left alone, it is the "end of the world".
No offense, but this is exactly why the word "hysteria" seems far more appropriate than "prudence". Not a single one of these vulnerabilities has been used to cause any measurable damage anywhere that we know of, whereas the mitigations deployed have significant costs that everyone must pay. Despite this, emotional "the sky is falling" type pronouncements are far more common in the media - even the ostensibly technical press - than attempts to rationally weigh the costs and benefits of any particular approach to the problem.
That's like saying: "nobody was drowned that we know of, whereas there was a significant cost to building the dam that everyone paid". (And also not dissimilar to arguments about doing no major industry/lifestyle changes regarding climate change).
It's exactly because there were mitigations relatively quickly deployed that we didn't have a "hack em all" exploit doing the rounds in hundreds of millions of devices. The difficulty of exploiting also gave some leeway to deploying those mitigations.
That's putting it lightly. Exploiting Spectre to get private data is difficult. Turning that into a privilege escalation is exponentially harder, so any "hack em all" exploit on hundreds of millions of devices would have needed an entirely unrelated mechanism for spreading.
It is very dissimilar indeed - the sentence you quoted does not constitute an argument by itself. It is an observation regarding the present state of affairs (which you have not disputed), which to me indicates a need to take a breath and do a reasoned cost/benefit analysis as opposed to the hysterical "this must be fixed at any cost, externalities be damned" mindset that is fairly common in many circles.
If you really want a climate change analogy, though, consider this - however many mitigating workarounds you invent, as long as speculative execution exists there will always be side channel attacks, and eventually some of them will probably succeed to some extent. Perhaps, as you noted, some major industry/lifestyle changes are indeed in order - people could stop living in the delusion that a perfect sandbox is possible and realize that arbitrary code execution will always entail risks. Rather than turning every website into a potential security risk, perhaps it is our approach to software (rather than hardware) that needs re-evaluation.
I fully support a users right to bypass these mitigations, and you're correct that your typical desktop user, at least today, isn't a target. But it seems odd that websites dedicated to performance computing have a blindspot to how automatically installed updates will impact performance.
It's quite easy to sensationalize benchmark results even unintentionally. The average reader of PC hardware review sites is totally willing to latch on to a microbenchmark result that shows a 20% performance drop and claim that it's disastrous for performance, even if the actual added delay to real-world operations is a fraction of a millisecond and thus will almost never cause the result of your user input to be delayed by even a single frame. There's a certain degree of irresponsibility in publishing results that you know will be taken out of context by almost everyone who reads them. I've discontinued benchmarks in the past because it was frustrating seeing readers pretend like they show a meaningful difference between products when the reader's workload never comes close to the workload represented by that benchmark.
That would kind of defeat the point.
If they published a benchmark in the past and don't bother to correct the benchmark when it becomes out of sync with reality - that is just bad journalism.
Nobody is saying you should go and cherry-pick benchmarks after the mitigations hit, but you should definitely check the benchmarks you already published once.
These sites can and should expect an informed reader.
In any case: Leaving wrong information up uncontested helps neither "experts" nor laymen.
No, and you should know better.
> If they published a benchmark in the past and don't bother to correct the benchmark when it becomes out of sync with reality - that is just bad journalism.
Proper practice is to publish the full test conditions, including software, firmware and nowadays also microcode versions. The availability of newer versions does not make older results any less true.
At AnandTech, we make all reasonable attempts to keep a thorough database of older hardware tested on newer benchmark suites, but the time this requires means we cannot re-test everything multiple times per year. I have over 200 SSDs and counting in the collection, and that test suite is over 30 hours long. The collection of CPUs is much larger. GPU reviews typically have fewer back-catalog hardware entries because updating to new drivers a few times a year is often unavoidable. You can browse the results for current and previous test suites at https://www.anandtech.com/bench/
> These sites can and should expect an informed reader.
You don't read the comments as often as we do.
This behaviour angered me no end. I wasted significant time looking for workarounds, and deleting the microprocessor driver was the only way. I wonder what fixes I've now nixed, but there was seriously no need for Intel to kill my overclock.
On a couple of occasions Intel have pushed updates which have reset my fix. Dear $deity .. my next PC will be AMD for sure.
Why would the attackers let people know they've successfully exploited?
No need, if you really want to disable all mitigations, including future ones, use mitigations=off.
https://www.phoronix.com/scan.php?page=news_item&px=Spectre-...
If you want useful benchmarks that show the performance impact, go to phoronix.
"PC Enthusiast" websites care about gaming performance and single user desktop performance, and always have. This has been the same since I started following things when the fastest CPU available was a 300 MHz Pentium 2. Imagine how amazed we all were by the 1 GHz Slot A Athlon.
Because Phoronix cares is outside the argument.
https://www.tomshardware.com/news/intel-cpu-microcode-benchm...
Anand has been working in Apple for a few years now, I wonder what has he been up to.
Do you run everything with administrative privileges?
If you don't sign that NDA you can't buy the CPUs from intel to resell them. Even if you are able to buy the CPUs from them, there's no guarantee that you'll buy from the list price or get the discounts for big, prestige projects which require tenders.
It's a deep and ugly rabbit hole.
Microcode benchmarking is not the hill Intel wants to die on.
And if two competing processors are close enough in performance for these mitigations to change which one comes out on top of benchmark charts, then other factors like price, power consumption and IO capabilities are probably a much bigger deal at that point than minor CPU performance differences.
Most if not all of our benchmark suites have been updated to include at least the early Spectre/Meltdown/et al. mitigations, and new CPUs are being tested with the microcode they launch with.
https://www.tomshardware.com/news/intel-cpu-microcode-benchm...
In fact, I would bet that most professional software engineers could not correctly explain Spectre, Meltdown, and Zombieload without making at least a few mistakes.
And a minute of googling brings up multiple articles about Spectre/Meltdown including two comparisons. https://www.anandtech.com/show/13659/analyzing-core-i9-9900k...
https://www.anandtech.com/show/12566/analyzing-meltdown-spec...
The Spectre and Meltdown vulnerabilities made quite a splash earlier this year, forcing makers of hardware and software to release updates in order to tackle them.
https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9...
It's typical for news sites to report individual announcements, with little or no analysis; this is fair, as long as the post clearly specifies its nature (which, in this case, does).
Anandtech did something very interesting actually on the Intel subject, which I didn't see on other sites - it made an article about the performance of the i9-9900k locked at its nominal TDP (95W), which showed very significant losses.
They also had the Intel series 6 launch where they praised the 6600K and compared it to the 2500K to show “massive” improvements over the years. This while all the other websites noted “minimal speed boost for too big price”. Perhaps both true but the spin on it makes all the difference when showing the intention.
AT shows quite the Intel bias. And it’s not the Intel part that bothers me, it’s the bias part. They go out of their way to make Intel look better without outright lying, just selectively presenting the truth in a way that shines a much better light on Intel. This for me casts doubt on other articles.
I’m glad Andrei Frumusanu’s mobile reviews still have a home, being the best I have seen on the entire internet. But that’s the only segment on AT where I can be reasonably sure about impartiality.
I remember back when Sandy Bridge was first released, and I was extremely pleased by the performance improvements my new chip was able to provide. Did they really manage to mess everything up within such a limited timespan? Or was there just always a hidden incompetence that never showed itself until now?
It's eking. Eke, eked, eking.
But perhaps there's something I'm missing here. Is there a misconception or lack of information here on my end that needs to be clarified? I can only make my analysis largely as an outsider looking in when talking about semiconductors.
I don't think any of this represents a short-term problem for Intel, other than the general downturn in processor sales because fewer people will need to upgrade. But I think it represents a very serious long-term threat.
They have some really cool technical advances, like 3D xpoint. But I'm concerned that they do so badly on embedded and custom integration from a long-term perspective.
I don't know enough about this, but the GP's argument of "Intel hit the wall first because they were the first to reach that level of performance" makes logical sense to me.
Yes, Safari, but iPhones do compete with your average desktop processor (not the top end).
https://macdailynews.com/2018/09/23/apples-iphone-xs-is-fast...
And as you said, you're comparing the top-of-the-line iPhone to an "average" CPU.
They really don't. A12 added a couple of instructions for floating point conversions, but contrary to claims making rounds on Twitter at the time, they were not even generated by WebKit when the benchmarks were run.
Don't forget that Intel CPUs have things that A CPUs are missing like QuickSync, AVX2, massive PCIe interconnectivity.
Whether the A-series CPU could be modified into something competitive on that front is yet to be seen. Whether this actually matters considering the state of our compilers and software development is yet another question.
Also, Apple doesn't make its own A-series processors - it uses TSMC for that.
We've yet to see how competitive they'll be once Intel leapfrogs that 10nm node. That's assuming that they can, of course...
In terms of architecture and vulnerabilities, it's not prudent to bet Intel chips are more vulnerable to exploits than others - it's just that we know more about those vulnerabilities. If you want to find vulnerabilities with high impact in cloud and enterprise data centers, Intel Xeon CPUs will be your primary research target.
It's also naive to dismiss the possibility for Intel to have learnt a lot from some of the failures in 10nm that will prove useful in accelerating node development in the future.
I wouldn't expect any massive leads in any industry to last for long. This might just be regression to the mean.
They still seem to be producing the fastest processors available for desktop and server.
It doesnt matter if someone else is making even a 3nm chip if the chip still can't outperform the current offerings.
They aren't comparable between AMD and Intel. They absolutely are comparable between Intel and Intel.
The spoiler though is that different architectures have different scaling properties and limitations. IBM's Power architecture has already scaled past where Intel is, not because of the semiconductor process, but because the architecture is more streamlined. ARM is somewhere in the middle, it started off pretty streamlined but it has been adding warts (special instructions) to more directly compete with Intel and that creates impediments to scaling.
So much so that we're now we're down to two companies in the whole world who are successfully executing the smallest CPU nodes, unless Intel manages to make their "10nm" work, or pulls off their "7nm".
While we're hearing the very roughly equivalent TSMC "5nm" node is starting risk production (https://wccftech.com/tsmc-5nm-production-euv/ beta testing, you might say, someone outside of TSMC has to be the first, second, etc. to try to get real world dies that work on a new node). Intel isn't saying anything, but Semiaccurate has reported at least two fab lines that were slated to move to their 10nm are installing lots of EUV equipment consistent with using them for their 7nm node (and at least one fab moving back to 14nm).
Intel is struggling because of their struggles with 10nm. Apple and AMD are not because TSMC has pulled off 7nm. Architecture matters, but process node matters a lot too.
They compounded their problems by essentially stopping microarchitecture development on 14nm, which is why eg. their laptop processors still don't support LPDDR4, and they're still shipping basically the same CPU core they released in 2015. Coupling microarchitecture and fabrication development has at times been an advantage for Intel, but for the past few years it's been a huge mistake, and they've promised changes to their design processes so that they don't get stuck like this again in the future if fab advances aren't ready when new microarchitectures are.
TSMC naturally doesn't have this problem, because they're a pure play foundry. Their customers have to each make their own bets on when new fab processes will be truly ready, and how well they will perform in practice.
But it's a brittle model, if a company screws up a node and is too messed up to handle the failure gracefully, as Intel is doing with their "10nm", no doubt with pride as a factor. And it's not uncommon for institutions to permanently lose abilities, I'm not counting on Intel succeeding with their "7nm" node.
On the third hand, we're now down to 2-3 high end CPU fab companies, Samsung, TSMC, and maybe Intel. That also can be a brittle thing.
It might have been hubris at having been at the cutting edge of fab tech for so long. It could have just been the fruits of pushing the envelope - sometimes what you predict will happen when you put theory to application proves false.
It has warped their business heavily for 4+ years now, but in the same way AMD had to "get their act together" with their processor design after Bulldozer failed spectacularly in practice and took ~7 years to fix it companies at these scales cannot turn on a dime - Intel had their roadmap planned a decade in advance, and to have it so thoroughly trashed starting around ~2015-2016 will require until at least 2021 to correct in all likelihood.
0: https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-...
The GPU was disabled because it is those blocks which are the most problematic yield wise.
But isn't the whole die exposed and otherwise processed as a whole piece? Very much not deeply educated here and can't justify the investment to change that, my primary mental models for defects are either something that takes out a whole die, like one lithography step being misaligned, or spot damage like a piece of dust.
But there are clearly issues in between that are statistical inside a die, I recall Semiaccurate saying one of Nvidia or AMD did a GPU tape out to a TSMC process where they duplicated vias because that process' were iffy, and they compensated with a less dense design where either one or two working was OK. If Intel is suffering that sort of problem, then the GPU is a big part of the die that can be fused out while you still have something useful. If all your CPUs or all your L3 cache banks fail, a working GPU is pointless.
That article points out two particularly suspect things Intel is uniquely trying at this node: SAQP for the metal layers, which I've seen cited before, and which they generically officially blame, and cobalt in interconnects. And at least one other thing was mentioned as suspect, and four new things total.
One ray of hope is mentioned for Intel, in that they were the most aggressive in the industry with their 14nm and 10nm nodes, and in both cases paid the price in yields, while they're being conservative for their 7nm node, no doubt because EUV is a very big step for everyone. Semiaccurate also commented and/or theorized that a compelling reason Intel is continuing to work on their 10nm at one fab is that one or more things in it are also going to be used in their 7nm, so they might as well debug them now and there, and sell some chips while they're at it.
Now to do some catching up on SemiWiki, thanks!
Tell me when they have really fallen. Still very far from it.
In technology, downward swings of fate tend to come fast and hard. The camera world went from 100% film to 100% digital in the space of about five years, which extinguished Kodak. Or consider Palm/Nokia/ Blackberry, who went from collectively owning the entire mobile market to dead as doorknobs in even less time.
It’s easy to see how it happens to Intel too: AMD’s big-core-count chips start eating up server business, while ARM takes over PCs (at this point people consider it all but certain Apple is switching to ARM in the next few years, and Microsoft is building Windows on ARM as a hedge), and without another business for Intel to fall back on (they’ve shut down modems, mobile chips, and anything else that could’ve been a new source of revenue), that’s the end.
I’m not saying it’s certain, but I’m saying it’s totally possible and their current market share means nothing.
performance benchmarks mean even less when those security issues are band-aided by performance metric-hurting-workarounds.
Most consumer wouldn't care about this anyway and the risk was overblown.
Even when people started speculating (ha!) that speculative execution could be problematic it took years before they managed to exploit it.
Are they really? I don't have an extremely deep understanding of Intel's implementation of x86 ISA, but I do know enough to say that so far we've been able to effectively mitigate almost all of these attacks with existing instructions available on the Intel CPUs. That doesn't mean that they are still not open to other variants of these attacks - but at some point you have to assume diminishing returns. Spectre is still very difficult to exploit, for example.
Perhaps this has little to do with Intel and more to do with software authors cutting corners? LFENCE and SFENCE are reasonably well documented, after all...
https://www.theregister.co.uk/2007/06/28/core_2_duo_errata/
The scariest part is that many of the best security minds work for various intelligence agencies. They very likely have known about such things for a very long time.
Meltdown strikes me as an almost perfect vulnerability. It affects almost everyone. It is undetectable until exploited and once exploited, it immediately goes away until the next time. It's easy to keep secret. Most importantly, it's a one-way vulnerability. Keep your secure systems from running untrusted code and there's zero risk. Since this is standard protocol anyway for those systems, you don't have the risk of someone running across a code patch somewhere.
The only potential downside is that the juiciest targets also aren't running untrusted code (though most foreign affairs workers probably run untrusted code). The big point of interest here is information symmetry. In most cases, giving others secret information is bad. In this case, both the best and worse case situations work out well for the USA. If nobody else knows, they get free info. If everyone else does know, then everyone gets perfect information about everything. This favors the most powerful country. They can eliminate the unknowns (the only real danger). In contrast, knowing you are going to be crushed does nothing if you can't hide your own hand either. So, the best case is very good and the worst case is still acceptable.
These attacks work fine in the browser, as researchers continue to show. They allow complete bypass of any native app sandboxing layers. Surely you don't run everything on your box as root all the time.
By the way, does NoScript actually block in-SVG javascript?
> Second, the increasingly complicated mitigations that we designed and implemented carried significant complexity, which is technical debt and might actually increase the attack surface, and performance overheads. Third, testing and maintaining mitigations for microarchitectural leaks is even trickier than designing gadgets themselves, since it’s hard to be sure the mitigations continue working as designed. At least once, important mitigations were effectively undone by later compiler optimizations. Fourth, we found that effective mitigation of some variants of Spectre, particularly variant 4, to be simply infeasible in software, even after a heroic effort by our partners at Apple to combat the problem in their JIT compiler.
> Our research reached the conclusion that, in principle, untrusted code can read a process’s entire address space using Spectre and side channels. Software mitigations reduce the effectiveness of many potential gadgets, but are not efficient or comprehensive.
The “some variants” include MDS, which the author was aware of but which were not at the time of publication out of embargo.
> The only effective mitigation is to move sensitive data out of the process’s address space. Thankfully, Chrome already had an effort underway for many years to separate sites into different processes to reduce the attack surface due to conventional vulnerabilities. This investment paid off, and we productionized and deployed site isolation for as many platforms as possible by May 2018.
So with improved browsers it's still unclear why ordinary users need those performance-eating mitigations, when browser vendors managed to solve that problem themselves.
For Spectre, that’s enough; for Spectre-class Intel permission exploit vectors (aka, Meltdown, Fallout, ZombieLoad, RIDL, Store to Leak Forwarding and other MDS vulnerabilities) all of the same infeasability of browser mitigations apply but data also leaks across process boundaries, so process isolation does jack shit to protect you without lower level mitigations.
There’s nothing whatsoever browsers can do to prevent this. Process memory read isolation effectively doesn’t exist in the presence of unpatched Intel MDS vulnerabilities.
> So with improved browsers it's still unclear why ordinary users need those performance-eating mitigations, when browser vendors managed to solve that problem themselves.
The unclarity is only in your misunderstanding of the relationship of MDS vulnerabilities on Intel to Spectre vulnerabilities in general.
> This can happen when one has opened the other using window.open, or <a href="..." target="_blank">, or iframes. If a website contains user-specific data, there is a chance that another site could use these new vulnerabilities to read that user data.
Most browsers have pushed patches which eliminate known mechanisms of leveraging the exploit, but the pathway cannot be completely mitigated by browser patches, I believe.
[0] https://developers.google.com/web/updates/2018/02/meltdown-s...
And that's exactly what shereadsthenews's point is. They can't boost all cores, and they are not boosting any core beyond the all-core capacity if it's truly a CPU that runs at 5 GHz all the time.
So, for example, my old laptop CPU would clock itself up to 2.7GHz on all cores... well, okay, it was a dual core, so that's not saying much, but still. But it'd only maintain that boost for a few seconds- under sustained load it dropped down to 2.5. This wasn't because of thermals, but rather because 2.7GHz was a Turbo Boost frequency, and once the PPL timer runs out...
> As of 3/16/19, the top 38% of tested 9900Ks were able to hit 5.0GHz or greater.
> As of 3/16/19, the top 8% of tested 9900Ks were able to hit 5.1GHz or greater.
So, Intel'd cut their yield by more than a factor of four if they only let parts that could hit 5.1 into this bin. For a 2% single-core performance boost...
Might be wrong, though.
[1] https://www.intel.com/content/www/us/en/architecture-and-tec...
Desktop chips also generally don't have any AVX offset, which is almost always required for 5 GHz all core.
https://devblogs.microsoft.com/oldnewthing/20050412-47/?p=35...
This upper limit depends on process, layout, power design and power limits of the CPU.
Last but, not the least, not all CPUs are created equal on a wafer. I came from an era where we hunted plain blue AMD Athlon dies for higher overclocking potential, since they were from center of the wafer and they were more stable under high load/voltage/clock. I had a 2200MHz Athon (200 x 11) which was faster then AMD's own 2200MHz Athlons, since AMD wasnt offering a 200MHz bus version of their 2200MHz parts.
That's not too bad. I'd be much more worried about code silently doing the wrong thing .
Also, a BSOD or panic in the wrong time can cause massive data loss. That's beyond bad sometimes.
Edit: I mixed Prime95 with SuperPi. Thanks AaronFriel.
The BSOD, oops, or panic is a symptom of widespread errors.
It's because the code does the wrong thing, and that happens silently... until it hits some pointers or kernel structures and stops being silent.
If the processor fires an unrecoverable MCE event, you're frozen with a nice, explanatory panic.
Center of a silicon wafer is said to be have a higher quality (due to lithography, physical stresses and other processes which I don't know exact details of), and the result is a die with more homogeneous properties and color reflection. Since the die's tolerances were lower around the center of the wafer, the performance of the resulting chip was better.
AMD was also sub-binning most of these parts (they were sold as Athlon 1700 @ 1433MHz regardless of their performance level), so people were buying these unlocked sleepers and overclocking them to insane levels without voltage increases.
However, today the processes is so different and node sizes are so small that the dies' color are different and not perceivable anyway.
In the older days, this issue was more of an obscure, collective wisdom which resulted from trial and error days of overclocking wars.
CPUs are basically huge networks of transistors (on/off switches). They're sort of like tiny printed circuit boards; lots of individual 'parts' are connected by 'wires' on top of a silicon wafer.
The distances are miniscule, but the lengths of wires running between transistors still varies. So when a transistor switches between 'off' and 'on', the signal takes a different amount of time to reach to its destination depending on which transistors are being switched. The signal can also feed into multiple other transistors which it will reach at different times.
While signals are busy propagating through the circuit, the CPU's state will be unstable, including the 'output' value of its current instruction. The time that it takes for any given instruction to stabilize is tough to predict because it depends on a lot of things, including how far apart the transistors are and how many of them the signal needs to pass through.
The CPU's "tick rate" in Hertz relates to how quickly it "latches" its internal state. Between "ticks", the CPU waits for all of the signals to stabilize. If they haven't stabilized when the clock strikes, bad things can happen.
I'm not sure how the 'quality' of an individual chip can make it more amenable to overclocking, though; maybe they run into fewer issues from thermal stress? Maybe the tiny 'wires' between the transistors have slightly less resistance? I dunno, someone help me out?
A current 9900K might be some 20%-30% faster than a current Zen, but it will no longer be so when with the new lineup.
Meanwhile mitigations are eating up Intel's performance advantage..
(I agree, it's confusing.)
https://ianhowson.com/blog/cpu-clock-rates-are-meaningless-n...
What this announcement is basically saying is that Intel now has a 8 core chip where all 8 cores can run at 5GHz indefinitely "out of the box".
> Base frequency is when the Tau moving window time has expired. How most modern high end motherboards set it to an effective unlimited time.
https://www.anandtech.com/show/13544/why-intel-processors-dr...
> To simplify, there are three main numbers to be aware of. Intel calls these numbers PL1 (power level 1), PL2 (power level 2), and T (or tau).
> PL1 is the effective long-term expected steady state power consumption of a processor. [...] PL2 is the short-term maximum power draw for a processor. [...] Tau is a timing variable. It dictates how long a processor should stay in PL2 mode before hitting a PL1 mode.
> This is where it gets really stupid: the motherboard vendors got involved, because PL1, PL2 and Tau are configurable in firmware. [...] This lets them set PL2 to 4096W and Tau to something very large, such as 65535, or -1 (infinity, depending on the BIOS setup). This means the CPU will run in its turbo modes all day and all week, just as long as it doesn’t hit thermal limits.
> 1000 Hz
You must have meant to say 5 GHz and 1000 MHz...
vi? Probably. But not Emacs, although I guess you could run something like Linus's Micro-emacs. https://git.kernel.org/pub/scm/editors/uemacs/uemacs.git/
In all cases, anyway, a 10 MHz 68030 should be enough for full Emacs, it's commonly seen as the lowest hardware requirement for a useful workstation Unix.
With AVX 4.8-4.9 is still doable without hitting the top 30% of CPUs in the CPU lottery.
My 9900K does 5.1 without any AVX offset but this is a top 10-20% CPU if the figures form Silicon Lottery are to be believed.
So it’s not that surprising Intel can simply bin CPUs to do 5.0 at near stock voltages since many resellers have been doing just that.
Unless the KS would guarantee a 5.3-5.4 all core OC I don’t see it being anything more than a PR release anyhow.
That said I’m not even sure if the 9900KS doesn’t come with an AVX offset to begin with most higher end motherboards come with a 9900K 5.0 preset anyhow which sets the voltage to about 1.3-1.325 and an AVX offset of 3 it just yells at you that you need a good cooling solution and this is not guaranteed to work.
Since I lean more on the 'performance' side of things, that was the end of 'single thread performance scaling', or put another way, that was when the performance of a single core stopped doubling every 18 months or so. And everyone switched over to dealing with Amdahl's law instead.
Moore's Law is "the number of lowest cost transistors doubles at X interval", and 193nm UV immersion lithography limits have been hitting it hard lately (see https://en.wikipedia.org/wiki/Multiple_patterning). But chip manufacturing equipment makers haven't run out of tricks quite yet.
Under full AVX workloads using Intel stock cooler?
Highly doubt it.