News Intel Announces Core i9-9900KS With $513 RCP, Arrives October 30

[Darkbreeze]

If I had a system that could only boost on any given core for no more than 28 seconds, assuming desktop platform, I'd either return it and get a refund, sell it or throw it in the trash. If the CPU can't maintain the expected boost clock for the duration of whatever amount of time I need for it to do so, then it is not capable of that specification and shouldn't be marketed or presented that way, at all. Ever. Period.

If I need or want the full package at 4.8Ghz for two hours of a Prime run, or however long it is needed to maintain that speed while in game, then that's what it should do. Otherwise, it cannot, and should not, be considered capable "of" that speed. Hell, I can OC my CPU to handle 10 seconds of Prime at 5.4Ghz, that doesn't mean I can sell it as a 5.4Ghz 6700k. That whole argument seems, nonsensical to me.

 

[bit_user]

I believe the 28 second boost limit is only for mobile platforms.

Certainly not. Have you not been following the plot about some gaming boards that ship with unlimited tau, by default? That is no small point, here.

View: https://youtu.be/aVLuKqfyVyw?t=153

That makes perfect sense in that application where the limits of acceptable ergonomics are more stringent than with a desktop system.

Yet, I agree with this. Mobile was one of the better use cases I had in mind for time-limited boosting.

 

[bit_user]

If I had a system that could only boost on any given core for no more than 28 seconds, assuming desktop platform, I'd either return it and get a refund, sell it or throw it in the trash.

Single-core boost should be unlimited, I believe, as that should stay below the PL1 threshold.

If the CPU can't maintain the expected boost clock for the duration of whatever amount of time I need for it to do so, then it is not capable of that specification and shouldn't be marketed or presented that way, at all. Ever. Period.

I'm not sure you understand the concept of boost.

If I need or want the full package at 4.8Ghz for two hours of a Prime run, or however long it is needed to maintain that speed while in game, then that's what it should do. Otherwise, it cannot, and should not, be considered capable "of" that speed. Hell, I can OC my CPU to handle 10 seconds of Prime at 5.4Ghz, that doesn't mean I can sell it as a 5.4Ghz 6700k. That whole argument seems, nonsensical to me.

So, basically, you're arguing that nobody should be able to get a short burst of extra speed from their CPU, if the CPU cannot maintain that speed. Why not?

I'm frequently compiling code in small bursts, usually for shorter amounts of time than 28 seconds. I would appreciate it if my CPU would boost so that I could see the results of my change a couple seconds sooner, and then get back to more editing, etc. Over days, weeks, and months, those few seconds would really add up. Plus, anything to shorten the code/test/debug cycle helps preserve focus and maintain the flow of programming.

But, you're telling me that I can't have that capability, because the CPU needs to be able to sustain that speed indefinitely? In other words, if your use-case isn't optimized, then nobody else's should be, either? That seems rather childish.

 

[kinggremlin]

Certainly not. Have you not been following the plot about some gaming boards that ship with unlimited tau, by default? That is no small point, here.

View: https://youtu.be/aVLuKqfyVyw?t=153

https://www.tomshardware.com/reviews/intel-core-i9-9900ks-special-edition-review

"However, Intel's chips shift into Power Level 2 (PL2) during boost activity, which the company defines as 25% over TDP (158.7 Watts for the KS) for a boost duration (Tau) of 28 seconds. Motherboard vendors often ignore boost duration limits, as we see with our own tests above, so you can expect very aggressive boost activity with most motherboards. "

 

[Darkbreeze]

Single-core boost should be unlimited, I believe, as that should stay below the PL1 threshold.


I'm not sure you understand the concept of boost.


So, basically, you're arguing that nobody should be able to get a short burst of extra speed from their CPU, if the CPU cannot maintain that speed. Why not?

I'm frequently compiling code in small bursts, usually for shorter amounts of time than 28 seconds. I would appreciate it if my CPU would boost so that I could see the results of my change a couple seconds sooner, and then get back to more editing, etc. Over days, weeks, and months, those few seconds would really add up. Plus, anything to shorten the code/test/debug cycle helps preserve focus and maintain the flow of programming.

But, you're telling me that I can't have that capability, because the CPU needs to be able to sustain that speed indefinitely? In other words, if your use-case isn't optimized, then nobody else's should be, either? That seems rather childish.

I FULLY ASSURE you, despite your attempts to be snide, that I fully understand the concept of boost. I also understand the concept of snark and talking down to people as well, which is what you generally tend to do and don't think it hasn't gone unnoticed around here.

As well, it was clearly commented about, just a few posts up. So maybe, tone that down a bit, shall we?

I'm glad that short bursts of performance are helpful to you, but they are not helpful to everybody. I can 100% promise you that on all prior architectures, and given, I haven't worked with THIS architecture yet so I cannot REALLY comment on whether it does or does not function as discussed here, but on prior architectures, when I am doing things that are steady state, like running Prime small FFT, those all core boost clocks STAY where they belong so long as there is a steady state load present and demand for them to. If I'm gaming, or encoding or converting, or similar long term tasks, I sure as heck don't want to see that the CPU is not able to sustain the advertised all core boost. As for single core boost, sure, that's great. But honestly I can't recall EVER having seen any task that was demanding enough to require the CPU to boost, that didn't incorporate more than one core and thereby make the single core boost specifications largely moot and pointless, because rarely if ever is there ever going to be a demand for just one core when all processes are taken into consideration.

 

[bit_user]

https://www.tomshardware.com/reviews/intel-core-i9-9900ks-special-edition-review

"However, Intel's chips shift into Power Level 2 (PL2) during boost activity, which the company defines as 25% over TDP (158.7 Watts for the KS) for a boost duration (Tau) of 28 seconds. Motherboard vendors often ignore boost duration limits, as we see with our own tests above, so you can expect very aggressive boost activity with most motherboards. "

Correct. That's my point. Often, but not all, as you'd have seen, if you watched to the point where he plotted the power consumption of the two boards he tested, over time.

But, contrary to what you thought, we're talking about desktop products, and specifically the i9-9900K and i9-9900KS - not just mobile. If this were only a mobile thing, then why the heck did you think we were talk about it?

What bothers a lot of people is that Intel advertises these chips as having a 127 W TDP, but unsuspecting consumers can buy a board, plug it in, and have the thing sustain far higher dissipation then they expected their cooling solution to have to handle. All by default. It's not exactly false advertising, but Intel's claims of 127 W and all-core 5 GHz both warrant a "huge asterisk", as Steve said.

 

[bit_user]

I also understand the concept of snark and talking down to people as well, which is what you generally tend to do and don't think it hasn't gone unnoticed around here.

I'm sorry it comes across that way, as I don't (normally) intend to slight anyone. I'm really just here to share info. Where I see a potential misunderstanding, ignorance, or confusion, that's typically when I post. I know I'm not always right, and I try to express gratitude when someone corrects me.

What I don't have a lot of patience for is people who post very strongly-worded opinions on the basis of very little information. If one expects others to respect their opinion, then they have a duty to make it an informed one. Sometimes, we don't know what we don't know, and that's fair. But, it's not too cool to be both forceful and wrong. Especially without even making an effort to check one's facts.

As well, it was clearly commented about, just a few posts up. So maybe, tone that down a bit, shall we?

Where? I'd just like to know.

As for single core boost, sure, that's great. But honestly I can't recall EVER having seen any task that was demanding enough to require the CPU to boost, that didn't incorporate more than one core and thereby make the single core boost specifications largely moot and pointless, because rarely if ever is there ever going to be a demand for just one core when all processes are taken into consideration.

I'd encourage you to watch the above GamersNexus link - I took the trouble of linking the exact point where he explains it. For a deeper understanding of Coffee Lake's boost behavior, you might give this a read:

https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo

The point being that we're not talking only about a single-core boost. What we're talking about is power consumption limits. If your workload allows it, the CPU will give you a multi-core boost of as much as its current power limit allows. So, my incremental compilation example is entirely consistent with even an all-core workload! However, it's true that the fewer cores you're using, the more boost they will tend to get.

Like it or not, this is the era of boost. You're only going to see more boosting from CPUs, not less.

 

[Darkbreeze]

Where? I'd just like to know.

I forgot bit_user was on the prowl taking everything literally to the T just to make a counter point.

So, while not specifically saying that you're being condescending, it really is the implication. You can be right, without needing to point out every minuscule point or obscure tidbit. Not everybody is, or wants to be, an engineer, nor do they all even CARE to take things to that level. Anyhow, I'm done here. Appologies if I slighted you at all. I see where you on the boost issue and as far as there BEING more boost behaviors, I'm not against that at all. What I'm against is false claims. If a claim is made, it better be able to be backed up and realistically, and COMMONLY be accurate. What we don't need are politician style marketing tactics when it comes to well, our CPU can boost to 5Ghz on all cores, but only on Sundays in Indiana if the rain is falling slightly sidways and there's a low pressure area moving in type BS.

 

[bit_user]

Thank you for clarifying that.

So, while not specifically saying that you're being condescending, it really is the implication. You can be right, without needing to point out every minuscule point or obscure tidbit.

To his claim about me looking to make counter-points, I'm aware of the pitfall that a couple invigorated posters can end up just nit-picking each other to death. So, I try to avoid that (but don't always succeed), and when I feel like I've made my case, I back off and let them have the last word.

One thing I've learned is that it's all too tempting to try and force the counter party to accept your position, but that almost never happens. So, my goal is simply to post good-quality information and try to let that stand on its own merits.

Appologies if I slighted you at all.

No, we're good. I did certainly jump on your earlier post, so a sharp reaction should've been expected.

Thank you for helping support the community, here. I appreciate your light-touch moderation and guidance.

 

[Darkbreeze]

No, actual appologies, because I think I was bit harsher, ok, I was harsher, than I intended to be. You are right that you try to put accurate information out there. It just feels like sometimes you strive for being TOO accurate, almost to the point of offering excuses for what I see as design shortcomings or faults, but that's more so my "opinion" than it is anything else and as such is perhaps no more relevant than anybody else's. You contributions are definitely appreciated too. Just, maybe keep in mind that it might not always be seen the same way that you are intending it to be. It's ok to call a turd a turd, just, maybe not necessary to actually describe the aroma as well. Heh.

 

[Darkbreeze]

For those who just fire up the occasional game, I think there's no cause for concern.

I suspect the 1 year warranty is just so they're not on the hook if some turkey buys it for heavy compute jobs and runs it full-tilt, 24/7.

I think this is what actually got me going on this thread. I just can't, cannot, can not, understand paying that kind of money for a CPU and there being ANY kind of expectation that a person shouldn't be able to absolutely run that at the advertised all core boost speed of 5Ghz, full time, 24/7, and have it AT LEAST last three years. At least. I feel like any CPU you pay this kind of money for, either can do what you have marketed it on the surface as being able to do, or it's pure BS, of the stinky kind.

It would be a lot more understandable for a light duty CPU like an i3 to have a lighter warranty, since it might definitely get used in scenarios that are much more rigorous than what it was designed for, than to think that somebody's top of the line consumer flagship model has to be babied and cannot be used heavily in the manner in which we normally would use these types of processors. It makes no sense to me, at all. If the CPU isn't fit for the job, because it can't do what you say it can do, then don't sell it that way. Knock back the clocks or cores and sell it as what you are confident it CAN do, assuming it will be a workhorse.

 

[kinggremlin]

What bothers a lot of people is that Intel advertises these chips as having a 127 W TDP, but unsuspecting consumers can buy a board, plug it in, and have the thing sustain far higher dissipation then they expected their cooling solution to have to handle. All by default. It's not exactly false advertising, but Intel's claims of 127 W and all-core 5 GHz both warrant a "huge asterisk", as Steve said.

I actually learned how to overclock CPU's (my killer Pentium 133) from this site in the 90's back when it really was Tom's Hardware. I have been on enthusiast message boards for almost 25 years. Not once, have I ever seen someone say that they bought a cooler capable of X TDP which is the spec for their CPU and they're seeing throttling when running above base clocks, what's going on? TDP is a meaningless number. Most enthusiasts don't know what Intel's TDP means, and there are zero laymen that have any idea what it is. Here is a coolermaster spec page for a HSF:

https://www.coolermaster.com/catalog/coolers/cpu-air-coolers/masterair-ma620m/

All sorts of specs, no mention of cooling capacity or TDP.
Newegg product page, same cooler:

https://www.newegg.com/cooler-master-masterair-g100m-ma-g100l/p/13C-000X-003C1?Description=MASTERAIR MA620M&cm_re=MASTERAIR_MA620M--13C-000X-003C1--Product

Click on specifications, same thing.

Cooling capacity is never listed on the product pages of online vendors, though it would be very useful if they did. Depending on the manufacturer, if the information is even available, you really have to dig to find it. You're arguing for a non-existent crowd. If you spend $500+ on Intel's top of the line special edition CPU, then try to cheap out on the cooler, you deserve whatever you end up getting. That's my stance, and no amount of nitpicking by you about this post is going to change my mind.

 

[kinggremlin]

I think this is what actually got me going on this thread. I just can't, cannot, can not, understand paying that kind of money for a CPU and there being ANY kind of expectation that a person shouldn't be able to absolutely run that at the advertised all core boost speed of 5Ghz, full time, 24/7, and have it AT LEAST last three years. At least. I feel like any CPU you pay this kind of money for, either can do what you have marketed it on the surface as being able to do, or it's pure BS, of the stinky kind.

I still believe it has to do with the extreme level of binning Intel has to do for these retail samples. They're not interested in keeping a stash of these around for warranty replacements when they know the people who are buying these are likely going to beat them to death. Removing the power and time limit for boosting, does not flip the overclocking bit on a CPU because it is still considered to be running within spec even with those two limits removed. So you could have your 9900KS running at 5Ghz all core 24-7 and not void the warranty. Look at the 9990XE. Cost $2000 to win at auction straight from Intel, and you got 0 warranty for that. The more extreme the binning, the less Intel is interested in keeping spares for 3 years.

 

[Darkbreeze]

And as per MY original point, the less interested "I" am in even considering that product. Might as well not even market it, or offer it to consumers. Sell what you have to Linus, De8auer and the any other ranked competitive overclockers out there, say "this is what we accomplished" and then move on. Why even bother for that matter. You know they're not really making any money of these limited sales. Seems more harm than good.

 

[JQB45]

So we are all friends now. As for the Boost/Base clock its simple. Intel and AMD mention both so whats the problem? You will get 5GHz for a brief burst of time and then fall back to say 4.3GHz for the duration. I don't have a problem with that.

Has anyone noticed its much easier to control the CPU in Linux, or is that just me?

I do mostly web browsing now days as my love for coding has died out. 1.5GHz all cores x 6 is enough for me...

 

[kinggremlin]

And as per MY original point, the less interested "I" am in even considering that product. Might as well not even market it, or offer it to consumers. Sell what you have to Linus, De8auer and the any other ranked competitive overclockers out there, say "this is what we accomplished" and then move on. Why even bother for that matter. You know they're not really making any money of these limited sales. Seems more harm than good.

Perfectly valid viewpoint. I don't tend to care too much about warranty length as I rarely use anything for more than 2 or 3 years outside of the case and the monitor. On that schedule, I almost never have any hardware failures. Never had a CPU fail, only one video card. However, I wouldn't buy anything new with no warranty. $2000 for a 0 warranty CPU is pretty insane.

 

[Darkbreeze]

So we are all friends now. As for the Boost/Base clock its simple. Intel and AMD mention both so whats the problem? You will get 5GHz for a brief burst of time and then fall back to say 4.3GHz for the duration. I don't have a problem with that.

I don't have a problem with that either, if THAT is what they advertise to be true in advance. Finding that out when the marketing has been indicating that it's an all core 5Ghz chip, after I've purchased it, would not make me happy and I WOULD have a problem with that. The whole point isn't whether the behavior is correct/normal or not, it is simply, if you SAY a thing can do something and you sell it based on that premise, then it BETTER be ABLE to DO that.

 

[Darkbreeze]

Perfectly valid viewpoint. I don't tend to care too much about warranty length as I rarely use anything for more than 2 or 3 years outside of the case and the monitor. On that schedule, I almost never have any hardware failures. Never had a CPU fail, only one video card. However, I wouldn't buy anything new with no warranty. $2000 for a 0 warranty CPU is pretty insane.

You're replying to your own posts now? LOL. (I think your quote got borked up)

 

[kinggremlin]

You're replying to your own posts now? LOL. (I think your quote got borked up)

Whoops. Yea, not sure what happened to the quote there.

 

[bit_user]

I just can't, cannot, can not, understand paying that kind of money for a CPU and there being ANY kind of expectation that a person shouldn't be able to absolutely run that at the advertised all core boost speed of 5Ghz, full time, 24/7, and have it AT LEAST last three years. At least. I feel like any CPU you pay this kind of money for, either can do what you have marketed it on the surface as being able to do, or it's pure BS, of the stinky kind.

It's a fair opinion, but I look at it a bit like race car engines. I believe formula 1 engines last about a half season and I don't even know how long top fuel dragster engines last. But, the point is that by running it hotter and harder, you're shortening its life. So, if people are cool with that, then they can pay the money and get a more highly-tuned and less-restricted engine ...er, I men CPU. Whether it lasts 1, or 3, or 5 years then is probably down to how you drive it. Or... cool it and run it.

It would be a lot more understandable for a light duty CPU like an i3 to have a lighter warranty, since it might definitely get used in scenarios that are much more rigorous than what it was designed for,

The i3-9350K only turbos to 4.6 GHz. Maybe they bin them in such a way that they know nobody is going to be able to run one at high enough clocks to cause such accelerated wear. Or, maybe it has lower thermal throttling thresholds. Either way, it might be somehow protected from redlining in the same way as the i9-9900KS can.

If the CPU isn't fit for the job, because it can't do what you say it can do, then don't sell it that way. Knock back the clocks or cores and sell it as what you are confident it CAN do, assuming it will be a workhorse.

I definitely agree that Intel needs clearer communication around the whole "all-core 5.0 GHz" and 127 W TDP specs. So, I'm not trying to excuse them from that. However, if taken to task for selling a CPU that they don't guarantee for more than 1 year, they would probably say you should buy a Xeon if you value better reliability and longer service life - especially for heavy workloads. The KS is a race car engine - not for long-haul trucking.

 

[bit_user]

I actually learned how to overclock CPU's (my killer Pentium 133) from this site in the 90's back when it really was Tom's Hardware.

Yeah, good 'ol Thomas Pabst. I believe it was hosted on pair.com or some such, but I discovered it before he even had his own domain.

Cooling capacity is never listed on the product pages of online vendors, though it would be very useful if they did.

I wonder if the industry is moving away from that? My last cooler was rated for 130 W, which is the TDP of the CPU I bought it for.

 

[bit_user]

I don't tend to care too much about warranty length

It can sometimes serve as a useful proxy for expected product life, such as hard disks. I once read that car batteries are a really good example of a product that's designed to last just beyond the warranty period.

I rarely use anything for more than 2 or 3 years outside of the case and the monitor.

I tend to stretch out my hardware for maybe 5-7 years.

On that schedule, I almost never have any hardware failures.

For me, it's been one motherboard, some hard disks, a couple routers (which I keep even longer), a UPS, a DVD-RW drive, and the fan died on a Radeon 9700 Pro, at work.

I wouldn't buy anything new with no warranty.

Right. The famous "bathtub curve" of failure probabilities - you get those "infant mortality" cases, where something dies in its first couple days of service.