Intel Will no Longer Disclose Multi-Core Turbo Boost Frequencies

#5480196 Posted on: 10/10/2017 01:01 PM

---

There is perfectly good reason why not to let anyone know any solid number for this...
Quality variance between chips is way too high. And as consequence, different chips have different leakage and require different cooling or power delivery.

In other words once fully loaded, one chip may not be able to clock that high on all cores due to stability issues, other due to cooling solution and next due to VRMs on MB hitting their limit.
In such situation, any hard number on paper is bad number for intel. Too high and people complain that chips are not getting there, too low and people see those chips as not so good. Not giving numbers and giving good chips to reviewers = they do not provide any false info, reviewers will.
Basically this kind of marketing strategy hints that customers are there for lottery, and may end up being disappointed.

#5480211 Posted on: 10/10/2017 01:33 PM

---

Probably to "avoid confusion" - since if they have 3 different clocks normal people will just be confused all around. And it looks worse, too.

#5480212 Posted on: 10/10/2017 01:40 PM

---

I guess without naming the all core boost frequency they could start selling CPUs with varying boost speeds on all cores depending on how stable the cores are. As long as one core is stable at the advertised frequency, they could cap the multi core boost if all cores aren't capable of whatever speed they used to target. Thus more units could pass validation this way and potentially be sold as a higher-end model at a higher price.

#5480218 Posted on: 10/10/2017 02:00 PM

---

Ultimately Intel likely suffers from bad yields while AMD can crank it up quite easily in contrast.


AMD sells CPUs with basically no OC headroom (1800X has a "boost" of 4GHz, and if you're lucky in the die lottery you can OC that to 4.1, but not further), and you think they can "crank it up"?

#5480219 Posted on: 10/10/2017 02:06 PM

---

With even more luck in CPU overclockability, I'd expect more returns. Poor retailers.

#5480240 Posted on: 10/10/2017 03:07 PM

---

It is quite simple.

By stating all core boost to be 4.3 ghz they make a legal commitment. With Turbo 2.0 stating 4.3 every of their 8700(K) has to hit 4.3 ghz stable on all cores. Every 8700(K) that only hits 3.8~4.2ghz all core boost becomes a defective, non-sellable unit - or garbage, so to speak.

Ultimately Intel likely suffers from bad yields while AMD can crank it up quite easily in contrast.

Intel needs to back to the drawing board and rather quickly at that.

That isn't how it works though, there is a binning process these CPU's go through before they get assigned if they are the 8700k or 8600k or whatever they might be, there is a reason they already have different models because they could not achieve these high clock speeds.

And even if what you said was true we have had cpus like this now for 6/7 years with all clock and single boost clock speeds and havent done this. This is just some random marketing thing that doesn't make sense... maybe it's just to have the big number at front or maybe it's to try and confuse the masses who are tech savvy... either way i think its poor practice of intel

#5480242 Posted on: 10/10/2017 03:20 PM

---

When u consider all the following together:

Fact: the node/architecture is hitting a power wall (the low quality TIM isn't helping either and 18/36 can fry your mobo if u oc (supposedly mobo partners fault))
Rumor: Intel provided binned ES to partners and reviewers (they might be an ES but with so little changes from previous iteration... I wouldn't expect diffs with final)
Fact: higher core count CPUs have lower all cores boost than AMD alternatives
Fact: CL availability is hyper low

I think the most logical reason is they want to "increase shields" and get rid of potentially bad advertisement for their almost "vapor-ware" high core count CPUs.

#5480245 Posted on: 10/10/2017 03:28 PM

---

More than likely that some one using watercooling are hitting the specified turbo frequencies while someone using cheap air cooling are not getting anywhere near as high. So in Intels defense better leave it off they are damn hot chips. Pretty much all reviews of the chip are using watercooling and thus providing better than expected performance. Heck all reviews i have seen have used watercooling have 2x or 3x sized Rads, pair this up with a 1x rad or aircooling i bet performance will suffer alot.

#5480263 Posted on: 10/10/2017 03:56 PM

---

I still don't understand why Intel even mentions base clock, or rather, why the all-core turbo clocks are higher than the base clock. The all-core turbo clocks makes for a better selling point than the base clocks. If thermal throttling is the issue, how about shipping a heatsink that isn't made out of a pack of soda cans?


Meh, GPUs are a little too different. Their operation is more complex, and as you pointed out, more variable due to silicon quality. Even power delivery for GPUs is far more variable. Some of them demand more power than the industry standard rating. Some need to draw some power from the motherboard, without accounting for any other PCIe devices. There tends to be more EMI near GPUs. Meanwhile, CPUs (and their VRMs) are relatively isolated, and get plenty of shielding from thick heatsinks. What I'm getting at is it's pretty much impossible for GPUs to guarantee their boosted frequencies. When it comes to CPUs, unless you're using the stock heatsink and/or a real crappy PSU, you are going to get the advertised turbo speeds just fine. If this was really such a problem, Intel would've made this decision from the very beginning. That, or they ought to have put more time and effort to refine their product - we all know they've got the money for it.

#5480281 Posted on: 10/10/2017 04:41 PM

---

I still don't understand why Intel even mentions base clock, or rather, why the all-core turbo clocks are higher than the base clock. The all-core turbo clocks makes for a better selling point than the base clocks. If thermal throttling is the issue, how about shipping a heatsink that isn't made out of a pack of soda cans?



The problem is consistency, well, the lack of it. To get the base clocks higher u need each core on each cpu to be able to get that high on fixed power and cooling parameters. Even if "most" previous generations 4-core chips could get high all core turbos, they all used different amounts of power and cooling, so would need to pack the highest performing cooling solution.
In other words, they just test the cores on a chip can get to a given turbo for 1ms without exploding and then throw the problem of finding the right power/cooling and clock limits to us......

#5480289 Posted on: 10/10/2017 04:55 PM

---

The problem is consistency, well, the lack of it. To get the base clocks higher u need each core on each cpu to be able to get that high on fixed power and cooling parameters. Even if "most" previous generations 4-core chips could get high all core turbos, they all used different amounts of power and cooling, so would need to pack the highest performing cooling solution.
In other words, they just test the cores on a chip can get to a given turbo for 1ms without exploding and then throw the problem of finding the right power/cooling and clock limits to us......
But like I was saying before, consistency isn't an issue unless you're using the stock heatsink and a very crappy PSU. My brother's PC has a quad core Xeon with HT and the stock heatsink. The heatsink is (to my annoyance) completely coated with dust, and yet the CPU can maintain turbo speeds on all cores just fine. You have to have really bad TDP issues in order for boost speeds to fluctuate enough for Intel to claim variability is an issue. If TDP consistency becomes that much of an issue, there's a good chance it is going to fail prematurely due to other quality problems. Granted, my brother's Xeon doesn't get reach the 4GHz range and Intel is finally going beyond 4 threads for mainstream products (more cores adds more variables). But as stated before, Intel has the time and money to refine their products to remain more stable. There is no good excuse for this.

The only reason Intel keeps pushing these high turbo speeds is because it's an effortless way for them to make their product seem better than previous generations and the competition. But the fact of the matter is, Intel may be pushing them too high, and as a result they can't promise results anymore. This is starting to look like the P4 "GHz wars" again.

#5480292 Posted on: 10/10/2017 05:08 PM

---

It's not that surprising, really. With higher core counts, the all-core turbo would have to be lower than the quad-cores. It's probably best for Intel not to advertise them.


Different workloads put different strains on the CPU. A CPU that's doing mostly integer work will product less heat and consume less power than when it's doing floating-point work. Different instruction sets will also produce different strain on the CPU, which is why the base and boost clocks on the Core i9 are so far apart (AVX-512 work is extremely stressful, so Intel can only guarantee a relatively low base clock). It also has little to do with throttling - the limiting factor is the TDP.

#5480294 Posted on: 10/10/2017 05:16 PM

---

This makes me wonder -- Is the single-turbo core a different one on each processor, depending on how it tests out? Or is it ALWAYS core #1?

Followup question: Is core #1 ALWAYS the same physical core, or can they label any of them core #1?

#5480296 Posted on: 10/10/2017 05:20 PM

---

But like I said before: why is this now an issue? I completely get what you're saying, but this isn't a new "problem". What you said has always been true, the only difference is Intel uses more cores at higher speeds. But if doing that compromises stability and consistency, then maybe they shouldn't do it.


Are you asking if the single tuboed core is always the same physical core? Because if so, the answer is no - watch Task Manager during an intensive single-threaded task and you'll see it swap cores once in a while. Windows doesn't seem to do core swapping as often as other OSes, but it does do it.

#5480302 Posted on: 10/10/2017 05:28 PM

---

Of course--it's marketing... What else? Intel wants to create the impression that its cpus run at a higher MHz speed than they actually do. Simple as that. Believe it or not, some people really do believe that the only difference among competing cpus is the MHz clock... As for why they are doing this only *now*, that is surely obvious, is it not?...