First Ryzen Quad Cores Will not Pass 3.2 GHz?

Well, unless they have another production process in mind they will not be able to put anything above 4.0Ghz, and that completely without overclocking headroom. From published slides, R5 and R3 had low power use so no luck with that.
I do not think GloFo next production process is ready to roll and from voltage requirements there is no chance to reach far beyond 4Ghz.
Yes, people were able to reach over 5Ghz, on liquid nitrogen and at 1.9V - does anyone here would even consider running that 24/7 (even if we had enough LN to do that) ?
AMD already proved that ZEN is very good, their cpus are finally competent in gaming tasks and very good in semi a professional tasks.
We can expect fixes for gaming, and not from game developers but systemic ones because it is very clear that something is wrong with that just from differences in performance between Windows 7 and 10.
Yes, R5 will be easier to overclock, but with as steep a voltage increase as is required based on R7 samples, I don't have much hope for big gains, maybe 200 Mhz over R7....

 

I'm going to go with this one and say that since AMD already confirmed 4-core processors past 3.2Ghz, that this is bogus.

 

Probably basing it on using the same LPP process, right? I've been following this since the release. Exceptional samples for Ryzen see 4.2ghz at ~1.5v. I don't see that changing to any great degree with the 4 core parts due to the process.

 

Uh... are you forgetting that there are 4 fewer cores involved? You don't need a new fab node or more voltage when you have less of everything else. So yes, there should be plenty of overclocking headroom. Fewer cores is also why they consume less power...
Unless AMD puts a limiter on these CPUs, or, if the silicon is bad quality, nothing should prevent these CPUs from achieving at least the same freqs as the R7 series.

That's assuming R5's require the same voltages as an R7. Remember - fewer cores. Even if that is true, the amperage certainly is not the same, and it wouldn't surprise me if the R7s need more voltage to make up for the lack of current. From what I recall, 8-pin CPU power connectors are limited to 244W. There is some loss in power just by downsizing 12v to 1.4 (or whatever Ryzens are at). So, even if R5s are also 1.4v, they have effectively twice the amount of amperage available to their disposal. In an extreme scenario, that means the quad cores get to use 122W more than their 8-core counterparts.

 

You're also forgetting the power supplied by the 24 pin. It really doesn't work the way you're thinking. That the CPU draws less should not be a factor unless you're doing extreme overclocks or the board has poorly designed power delivery. And surely we are not using the extreme to characterize the norm.

 

Gents,

The tweet, if true, must be addressing quad-cores without SMT. AMD has already announced the Ryzen 5 1500X quad-core with SMT @ 3.5GHz / 3.7 GHz.

Google AMD Ryzen 1500X and hit the AMD homepage link on the Ryzen 7 news... Can't post a link yet as I've only written one post prior to this one... So daddy won't let me ;-)

 

Manufactured by GlobalFoundries.. maybe there's bit limitation on the design itself.. Can't push to higher frequency.

 

Again, CanardPC is one of the most respected and old French tech magazine/website, publishing since 2003.

 

The tweet states with and without

 

Didn't notice an actual translate in this thread so I used google:

Some things to note:

1. This is only the two lower R3's, their flagship quad core is actually part of the r5 lineup.

2. When canard said "do not exceed 3.2 ghz" we can assume they've tried an OC given that retail ryzen motherboards are now available where as pre launch they wouldn't of had that luxury. There's no getting around that, no "I'll just take it to 3.9 ghz". That is your limit.

3. No 4 core boost in other words no asus multicore enhancement or anything of the like, you'll only get one core that'll turbo.

It's obvious that AMD are doing this to keep the value of the r3's from being a little too good, at the end these have to only compete with the slower i3's and pentiums.

 

No, you still need more voltage, it does not matter how many cores you have, they still need the same voltage to remain stable. You only get less heat because only 4 core are producing it instead of 8. So you get easier cooling, easier power delivery and lower temps. While temps are a concern and the lower the better, they will not allow for much of increased overclocking potential - you may get 100-200Mhz out of that.
IF the silicon is of the same quality, they will work at the same frequency or slightly better (100-200) but if they are LOWER BINN chips left from R7 then you can forget about high clocks. Time will tell.
"That's assuming R5's require the same voltages as an R7. Remember - fewer cores." - no matter the number of cores, voltages requirements remain the same.
" it wouldn't surprise me if the R7s need more voltage to make up for the lack of current" - are you joking, is that some kind of new physics? They are in no way interchangeable. Voltage is all an all important king that decides how much heat is produced. Amperage matters only so far as that with more amperage it is harder to produce clean, stable current and stable voltage as a result.
From various test I have seen and temps produced by Ryzen, stability was not a function of overheating because chips would not clock over 4.1Ghz at 1.45V even when under liquid cooling and temps below 70C. Not even pushing more voltage would budge them.
So what I predict is that R5 will just clock up to 4.1 or 4.2Ghz at most but with lower temps - there is no silver bullet to that - IF they are not leftovers from R7. The same can be said for R3, R5 or R3 though, if they have a whole CCX disabled (hopefully not just 2 cores from each CCX) then they may have better performance in games. On the other hand since L3 is split between CCX they would have only 8MB of L3 cache.

 

I have overclocked dozens of chips. No, you don't always need more voltage, and yes, the amount of cores can matter. I'm not saying these statements are always true, but they often are. There are people who took 4+ core parts and reduced them down to 2 cores in order to achieve higher overclocks. And yes, you get less heat because there are 4 fewer cores; with 4 fewer cores, that means there are less components drawing power. When there are less components to power, the rest of the chip gets all that extra juice. Of course, this isn't that black and white. If your PSU and motherboard components can supply a stable voltage/current, the "extra juice" doesn't matter.

Agreed - this is an important factor.

No, they don't. In addition to what I said before, I'd say this is solid evidence to prove my point:
https://www.bit-tech.net/hardware/2017/01/03/intel-core-i7-7700k-i5-7600k-review/8
Notice how the i5 consistently uses less voltage than the i7, even during overclocks. Notice how the i5 manages to reach a higher overclock. The i5 even has the same amount of cores and yet there's still differences in voltage and overclocking capabilities.

You clearly have no clue how this stuff works... Y'know those CPU connectors on the motherboard? Take a close look - they involve yellow wires. Yellow is +12v. CPUs usually operate below 1.5v. When you downscale voltage, you either waste it through heat (which is a stupid idea considering there's a 3.3v source), or, you convert the voltage into more amps. Amps are so important for CPUs that there are 8-pin CPU connectors, involving 4x yellow wires. You can sure as hell bet they're not connecting those yellow wires in a series circuit; they're set up in parallel, in order to get more amps.

Agreed - overheating probably isn't the issue. I suspect it's more related to incomplete BIOSes, where even RAM seems to be suspiciously crippled. Stablity of RAM tends to have a correlation to the stability of a CPU, so temperature can't be the [only] problem.

Assuming what I said about the RAM is true, I would predict the same as you.

 

Won't the time gap between the R7 and the R5 parts, actually fix the binning issues? Something like three months after Polaris 10, XFX was selling cards that were not spinning their fans at 60C, and which could OC to 1.4GHz.

 

Maybe they arent soldering the IHS on for these lower end chips.
This explains partially why the low clock speeds.
Lower binning contributes.

 

No, the rest of components does not get extra juice, there never is "not enough juice" up to the point that VRMs can't reliably convert voltages due to overheating or other reasons, up till that point there is always enough, BUT voltage quality deteriorates with load put on VRMs and their temperature (hence additional cooling on VRMs). It is not the quantity of "juice" but a quality of it that matters.
People disabling cores do that mainly because in 4 core cpu, some cores are on higher quality part of silicon and some are on worse parts, and those are less likely to stay stable on higher clocks. I'm talking about extreme overclocking to beat records on LN for example. Good overclocking motherboards are designed to offer good quality power supply for X times more power then 2 core cpu would use. The deciding factor is silicon quality. Although a SMALL BIT may be caused by lesser quality of current flowing from VRMs that are supporting 4 instead of 2 cores, but it is a SMALL part as the change in quality is small - if that were not the case then you could not safely overclock at all if the jump from 60A to 120A would cause too much instability in voltages. BTW, think about how much current VRMs must supply to a chip working at 6Ghz at 1.8V on LN in contrast to 4.5Ghz at 1.35V on air (keeping in mind current requirements do not increase linearly with voltage but much faster) - that is the amount of power good motherboard is able to SAFELY provide for cpu (the better motherboards offer 12+4 phases of even 40A).
Power consumption of a CPU is proportional to the simple forumula f*V^2, with f the frequency and V the voltage or if you want to get technical and use static and dynamic power calculations - the static power consumption is equal to a defined constant k1 times voltage squared ( k1*V^2 ), dynamic is equal to another defined constant k2 (capacitance mostly) times the frequency times the voltage squared (k2*f*V^2) - together they constitute power use (and consequently current) and you should note the "V^2" meaning that voltage is most important to total power use.
Sorry for my rant, you did state "If your PSU and motherboard components can supply a stable voltage/current, the "extra juice" doesn't matter" but that one sentence kind of invalidates whole argument, because people do not disable cores for more power in typical overclocking scenario.
And yes, voltage along with adequate cooling is a deciding factor in how high you can overclock, up to the silicon limits (that are dependent on voltage quality, silicon quality and cooling - it is a circle of dependent variables)

First thing, do not equate 1 sample of 2 cpus (i5 and i7) with some RULE about how world works, they may have got a better binned i5 then they got i7.
Second, i5 has less "parts, that can go bad" during overclocking, meaning no HT and less cache, those are all transistors that can fail or require more voltage to remain stable, more transistors mean more heat and that heat caused them to abandon 5.1Ghz on new BIOS - and you know more heat forces more voltage to be applied to remain stable - it is a never ending circle.
Remember that they stated "(after BIOS change)... temperatures got too toasty above 5GHz, but the voltage required to hit lower frequencies appeared to be lower." so some of the differences can be caused by a BIOS change. This test I would call at best inconclusive to what you are trying to explain.

You know you are not writing to a small child? I have been building and overclocking computers (as a job, not for fun) since late 1980 and while that does not mean I'm an expert, I do know my physics lessons and while I do not have a degree in electric engineering I still can count my Volts, Amps and Watts.
Please keep such mode of writing for your kids (assuming you have them).

Also, keep in mind English is not my primary language so small errors may crop up in words use.

 

That may be true in some cases, but not all. Mind explaining why 8-pin CPU connectors exist? Having more power connectors does not give you more stability, it gives you more watts. Capacitors are responsible for giving a high-quality and stable power source; not the PSU, and not the VRMs.

Again, sometimes, but not always. For example, the 1700 is a result of what you said (where it would be an 1800X if it had better quality silicon) but for an X-series quad-core Ryzen 5, nothing implies the available running cores have faulty transistors. However, the non-X models likely do comply with what you state.
Hardcore overclockers can't bank on a magic chip that just happens to overclock well. Otherwise, achieving world records would be a real gamble and involve a lot of wasted time. Nowadays, CPUs are basically sold based on the quality of their silicon.

Yes, and? That doesn't invalidate that amps matter.

It doesn't invalidate anything. It's acknowledging that what I said doesn't apply to every case. Amperage delivery sometimes matters, where if you're reaching your limit, reducing the core count can help. Silicon quality sometimes matters, where sometimes a core really could be imperfect, and disabling it can help.

You're not wrong, but neither am I.

Did you not look at the chart? It also showed default voltages, and the i5's default voltage was still lower. Default voltages do not change; if you bought the same products, you'd get the same numbers.

Those facts you brought up support my point, not yours... You were saying how additional cores (which as you're aware, involves more transistors) do not requie a change in voltage. But as you've clearly just stated, having HT involves more transistors and therefore requires more voltage...

The only thing that becomes inconclusive is whether or not the i7 can overclock higher. But if more extreme cooling was involved, the i5 potentially could too, and at a lower voltage. The difference is the i7 will eventually run out of room for overclocking, because it will reach the wattage limit sooner.

You got what you did to me: condescendence. The fact that you seem to completely ignore the importance of amps requires me to spell things out. Again, I'm not saying it's the only factor, and I'm not saying it always applies.

Your English is perfectly fine.

 

Why 8-pin connectors exist? Because standard 24-pin connector is not able to transfer enough power (and not burn connectors), the same reason GPUs have additional power cables and not get all power from PCI Express. The second - to not type so much every time, and simplify.

World leading overclockers sometimes go through hundreds of cpus to get that one golden best overclocking chip. It's not a gamble, it require a lot of skill and some luck along with good motherboard, good PSU and good memory. Sure they go for the best potential overclocking chip and then test, test and test. The best ones have a pick between a lot of chips, either due to "connections" with suppliers of just friends. You may have something on 1700x vs 1800x - probably the same quality chips only with different stamp on them to create market separation.

They don't (much) - first frequency and voltage matters - power used and amperage are the function of changing voltage, frequency and intensity of calculations conducted.

We are both right here. If you have ****ty PSU and VRMs and, yes of course, capacitors, you won't be able to overclock because you won't have stable voltage (far sooner that you would ever run into amperage limit). But silicon quality ALWAYS matters, bad cpu won't overclock no matter PSU quality and what motherboard you put it into.

You're not wrong, but neither am I.

Why are you surprised with that, the i5-7600 default max frequency is 4.2Ghz and i7-7700 default is 4.5Ghz, it is normal that default voltage (which is max voltage by the way) for higher frequency is higher too. Even if you were to compare base frequency, the difference and reason for it is still there.

You disregarded previous statement. They may have gotten a better sample of i5 then they got i7. More cores in cpus of the same silicon quality will not require more voltage at the same frequency. By itself having more cache and HT will not require more voltage, but when overclocking it MAY require to have more voltage because every transistor is a link that can fail, while it can remain stable at some frequency it may loose stability at higher frequency and then need more voltage. Please keep my statements in context.

You forget that on initial bios they got the same 5.1Ghz. But that does not matter because we somehow can't get on the same page about reasons for why higher voltage is even required.

Ok, so we CAN call us even on that? Maybe my way of pointing a mistake (in my opinion) came by as condescending. Sorry for that.
Don't mistake my words as disregarding amps, they are important at the end (far end), it's just that in 99% of cases you will stumble on other factors first.
In the case of Ryzen (that we are supposed to be talking about), it is a 14nm LPP production process that from about 3.7Ghz require a steep voltage climb to increase frequency (it's why all core turbo for Ryzen is so low, well compared to 6900 it is not so low), then you get a heat problem once you get to about 4.1Ghz, you need very beefy cooling and from 4.2-4.3Ghz water cooling because of steep voltage increase and resulting heat. Even when they OCed with LN cooling and over 5Ghz on mad voltages of 1.9V (and hundreds of amps of current), the motherboard was still stable (granted that was ASUS Hero, but it was stable) and with all 8 cores active amps still were not a deciding factor, but silicon quality. You can understand that when after few increases in voltage you can't get any more frequency bumps out of CPU and the board is still going strong without power or thermal shutdown.

Thank you.

 

^ Hmm... Well, I tend to find that more believable. As for under 3.2 - I'd see that as more for mobile demographic/market. Has there been any talk of some sort of Ryzen APU's?

Update:

Found this - http://www.e te kn ix.com/amd-confirms-ryzen-apu-2h-2017/

 

Exactly... it can't transfer enough power, as in, amps. The voltage remains the same but the wattage increases - it's amps.

Got a source on that?

w=v*a (w=watts, v=volts, a=amps). Amps do not change based on the voltage. I'm not sure if that's what you meant, but it seemed that way. Frequency and voltage has a direct correlation of power consumption, but that power consumption cannot be achieved if the amps aren't there to supply it in the first place. It doesn't matter how much force you spray water onto a burning house if you only have a 1mm stream.

Agreed - silicon quality is imperative. But I think you're overestimating the liklihood of weak overclocks due to partial defects. Keep in mind, the average person isn't going to use liquid helium to get the job done. There is room for error. The 4.1GHz limit that Ryzen 7s seem to be stuck at is clearly not caused by silicon defects, because everyone seems to be stuck there, with the exception of those who have extreme cooling.

I'm not surprised. What surprises me is how you don't realize the contradiction in your point. You said the amount of cores has no impact on the voltage, meanwhile, these 2 CPUs have the same amount of cores and yet their voltages are different (more on that later). The frequency difference isn't different enough. I am sure you could drop the i7's voltage to 1.168v and it would still run stable (assuming your PSU and VRMs aren't garbage).

I have several PCs where making them efficient was my top priority. There were CPUs I managed to lower a whopping 0.225v (while retaining stock speeds) and it remained rock solid. Intel, AMD, and Nvidia tend to give a lot of extra voltage headroom in the event you get a mini brown-out in your system (which uh... is often caused by a lack of amperage).

You're still missing the point: nobody knows what the silicon quality is of an i5 or an i7 when they buy it from the store. Intel does not change the voltage of their individual components (so you're not going to find a 7600K with a different stock voltage from another 7600K).
Meanwhile, way back in the Bloomfeld days in 2008, people could get i7s to breach 4GHz on stock voltage.
In another perspective, the FX-8350 (a 4-4.2GHz part) has the same stock operating voltage range of 0.82 - 1.43V as the FX-4300 (a 3.8-4.0GHz part) despite the drastic transistor count difference and 200MHZ difference. Unlike Kaby Lake, these are very power-hungry parts, where every little change in voltage or frequency makes a big TDP difference. Those 2 CPUs actually disagree with both of us (though, I did say from the beginning that voltage can be affected by number of cores/transistors, but not always).
I'm not denying that silicon quality matters, but I can assure you it was not the deciding factor of the voltages applied. The companies give a healthy surplus of voltage to ensure reliability. That being said, a frequency difference of 300MHz isn't enough to require a voltage bump.

The only page I'm on is to prove that voltage isn't the only factor to consider, and, that silicon quality isn't likely the cause of these CPUs being OC limited.

Again, if amps aren't that important then why are 8-pin connectors used?

Understood. But I think this statement of yours got me to realize a better way to phrase my point:
As history shows (hence my varying examples), regardless of testing CPUs for silicon quality, CPUs with fewer transistors tend to achieve higher overclocks, while often (not always) using less voltage. Statistically speaking, fewer cores have a lower chance of something to go wrong. Perhaps that's a result of a lesser chance of encountering silicon defects. Perhaps it's because the current is more stable. There are a lot of factors. But ultimately my point remains - a CPU with fewer cores should be more stable, not less.


Anyway, none of this discussion really matters anymore because as pointed out by another Guru3D article, the Ryzen 5 quad cores are in fact much higher than this rumor claimed.

 

I'm not going to read the wall of text but this stood out so I'll comment.

V=IR
I=V/R

As voltage changes, current does proportionally as well.