i3 and i5: Turbo boost useless on i5??

Liam m · May 31, 2016

Hi

I have a little question regarding the i5 and i3 lineup from intel. Doesn't matter which gen.
I currently have the i5 6600 ( non K at 3.3ghz, turbo to 3.9 if one core is used, I think to 3.6 if 4 cores are used?). I also see the i3 6300 runs at 3.8ghz without turbo boost.
I was wondering: why do the i5's even have turbo boost and i3 not, if the i3 has a higher base clock?
In other words: what's the point of calling it turbo boost on i5 processor if the frequency is lower than an i3 cpu (when 4 cores active)...
Is this pure marketing to mislead people?
Could intel also sell their i3 6300 as "3.2ghz, up to 3.8ghz turbo boost frequency"? Or am I missing something here?

ps: I know turbo boost can only be used when power draw etc are not too high... And I also know the basic differences between i3-i5-i7 cpu's...

Thanks

rgd1101 · May 31, 2016

because like you said, i5 have 4 cores, i3 have 2 cores with HT

Liam m · May 31, 2016

rgd1101 :

Then why is it clocked at a lower speed?

Kurz · May 31, 2016

Liam m :

i5 Its clocked at a lower speed for efficiency and stability.
The i3 is clocked higher because its gimped by the fact it only has 2 cores, and a higher clock speed helps mitigate that a little.

guanyu210379 · May 31, 2016

and..clock speed is not everything
Higher clock speed does not always mean that the processor is better.

Liam m · May 31, 2016

Kurz :

Thanks for the answers..
My next question: why is the i7 6700k (4 cores +HT) then clocked at 4 ghz stock speed? Is the stock voltage than higher or how does that one run stable while it is clocked higher?

Groveling_Wyrm · May 31, 2016

The i7 67000k is part of the I7 series. They are the best processors that Intel has produced. The ones that don't test so well get labeled as I5's and I3's.

Liam m · May 31, 2016

guanyu210379 :

Yes of course the i5 is better, but will the single core performance of a same generation i processor not be the same for all i3, i5, i7's when you put them at same frequency?

Liam m · May 31, 2016

Groveling_Wyrm :

I knew this happened with gpu's, but was not aware from this being happening with cpu's...
But I assume only a few i5 were normally meant to be i7? Or are all i5/i3 failed i7 cpu's?

Groveling_Wyrm · May 31, 2016

Liam m :

It is done with both GPU and CPU's. They call it product binning. They get each one to a suitable frequency and voltage and go from there.

They all start as I7 then get downgraded. They get put into a machine where they are tested several or hundreds at a time to see where they fail.

teknobug · May 31, 2016

Liam m :

AMD did that with the Athlon and Phenom lines, if they failed at 4 or 6 cores they were released as 3 cores or even dual cores. Many of the Phenom II X3's were unlockable to 4 cores, and a Phenom II X4 B55 unlockable to 6 cores or such.

Rogue Leader · May 31, 2016

Groveling_Wyrm :

That is not exactly true. The i3 is based off the lower level Pentium processors with hyperthreading enabled, while the i7 and i5 are essentially the same with the i5 getting hyperthreading and some cache disabled.

AMD was the ones that were disabling poor testing cores to create the Athlon and Phenom X3 and X2 processors off of X4's that didn't test well.

Groveling_Wyrm · May 31, 2016

Rogue Leader :

I have found nothing that backs your statement up, regarding Intel's binning procedure on the I3 being different. This would require a separate chipset design, requiring removal of 2 cores, as well as a separate manufacturing setup using different wafers. It would be Inefficient for Intel to do that.

What we do know is that the Skylake H chipset is the same for all I3, I5, and I7's, the Pentiums, Celerons, and the server Xeon's, as well as having the same motherboard chipset. If they changed the chipset then they would have change the Land Grid Array(LGA) to suit as well, to minimize waste.

Logic says that they use the shmoo plot, and get a percentage of processors from each class of processor out of each wafer, instead of wasting wafer space. In most likelihood, the I3, and Pentiums are probably on the exterior.

And a few questions you have to ask yourself regarding your answer: If Intel downgrades their I7 to I5 (Which we know they do), then why would they reverse that and do an upgrade to get the I3 from a Pentium? What would Intel do with the I7 chips that only have 2 working cores, but yet have hyperthreading? What would Intel do with the I7 chips that only have 2 cores and no hyperthreading? Junk them? I don't think so.

If you have any information that proves me wrong, I would be glad to know where you found it.

Rogue Leader · May 31, 2016

Source: A very close friend of mine actually works at Intel, and I work for a company that makes the machines that creates those wafers, as well as the lappers and saws for cutting them.

The short version is they have 2 different die sets for their main market chipsets, dual core and quad core. Dual core is used for everything from Pentium and Celeron up to i3, and quad core is used for i5, i7 and some E3 Xeons. This is also why I said you are not exactly right but close, and in fact I didn't describe it in my first post exactly right. In very basic terms i3's are the top level dual cores and they get dropped to Celerons and Pentiums, and i7's are the top quad core which gets dropped to i5's. Its the same idea, just that they don't go disabling/cutting cores out to get to an i3.

Their quality control on build these pieces is EXTREMELY high, they just aren't running into dead cores like they used to. Thats thanks to advances that companies like mine has made.

EDIT: wanted to add to relate back to the original question, thats why the i3 gets itself up to 3.9 ghz because it has the best of the dual cores, while the i5-6400 runs at 2.7ghz because it was binned the weakest of the quad cores.

Groveling_Wyrm · May 31, 2016

Rogue Leader :

Fair enough on your claim, but no data to back it up.

The data I can find comes from a former engineer at Intel, as seen in this LINK. And everything else that I that I have found while searching through the internet, to answer to this question has shown me that his answer is the most accepted answer on the web.

And yet, I do have to ask, according to your answer, why is Intel running two wafer dies? It doesn't make sense. All they would have to do is disable 2 cores, HT, etc..... on the I3, Celeron, and Pentiums, to get the processors they want. And there has been multiple times that Intel has shown that they do exactly that. Why would they change it?

That question just bugs me if in fact you are correct. Because that would increase the cost to Intel by MILLIONS. If they have one die, as I was stating and having high QA, and they need more I3 or Pentiums, then they do what they have also done before. They lock the HT or cache, or whatever and make an excess I5 into and I3, Celeron, or Pentium, or whatever they need to. This costs less.

So, basically we are using the same source (Current Intel person vs former Intel Engineer) for 2 different sources on this issue. So I do have to ask the question of why so much disparity between your answer and his answer?

I also found THIS Youtube video which supports the Engineer and his description of Binning. That would be some serious FUD if they are both wrong. And this would indicate that someone is spreading a massive amount of false information all over the web, yet no one is correcting it. Weird.

Liam m · Jun 1, 2016

Interesting discussion..
It makes sense if the i3 is the top of the dual cores, it runs at highest frequency like "Rogue Leader" said.
But indeed, why does techquickie say otherwise, just like the first link in the comment from "Groveling_Wyrm" .

synphul · Jun 1, 2016

For the i5 6600 (non k) the stock speed is 3.5ghz. Turbo boosts up to 3.9ghz at the following intervals
1 core fully loaded, 3.9ghz
2 cores fully loaded 3.8ghz
3 cores, 3.7ghz
4 cores, 3.6ghz
http://www.intel.com/content/www/us/en/support/processors/000005647.html

It's called turbo boost because it raises the clock speed above 3.5ghz which is where intel placed the stock base speed. Has nothing to do with an i3. A faster speed than the stock base speed is a 'turbo clock'. If you'll notice the i5 does run at 3.8ghz when 2 cores are loaded, same as that i3. Dropping speed 100-200mhz is the tradeoff for 2 extra physical processing cores.

My bet is that it's simpler to have a flat stock speed which is part of the reason the i3 is less expensive. Less features. It's allowing people to get a 3.5ghz quad core that is equally capable to the i3's higher clock speed under the same conditions, 2c fully active. There's nothing misleading about it, the i5 is capable of the same speeds (more if overclocking a k series) than the i3. It has more cache, twice the physical cores, along with a handful of other features like trusted execution, sipp, vpro and tsx-ni.

Even if we assumed people were 'duped' by 'turbo' in the name they can see one is 3.8ghz and the other is 3.9ghz. 100mhz isn't exactly going to win people over or wow the crowd. It's just a matter of not having to choose between core speeds and core count. I think you'd be hard pressed to start asking the masses why intel did what they did since none of us here (to my knowledge) were part of the decision making process. I wasn't asked for my input on the matter by them lol.

Honestly I don't think I'd worry about it unless you feel like you bought an i5 for nothing. Are you unhappy with your cpu? They have so many models there's literally one for just about everyone. They have cheaper i3's, even cheaper pentiums, unlocked i5's, more expensive and faster i7's. If they in fact changed the i3 and enabled turbo boost (an additional feature which takes tech to make it happen) the i3 would cost more. Assuming they advertised the i3 with turbo boost and went to the trouble to enable it and it turbo'd up to 3.8ghz how would that change anything from the way it currently is?

Liam m · Jun 1, 2016

synphul :

Little correction: the i6 6600 is clocked at 3.3ghz and the turbos you mentioned are correct.
I just asked the question out of curiosity. I am very happy with the performance of the cpu, although if I had the money I would definitely go with the i7 6700K.

And yes, you have a good point there, comparing two cores from i5 with the i3. (if you exclude the fact that the i3 has HT with those 2 cores at 3.9ghz, but that is not my concern.)
It was not my intention to hint that i5 owners are duped by the fact they have turbo but lower base clock, apologies for that misunderstanding.

Thanks for the answer.

synphul · Jun 1, 2016

You're right I made a typo on the base stock speed. Honestly I don't know unless it has something to do with power savings though that's what speedstep is. The i3's cores are hyper threaded but still only 2 processing cores vs 4. Ht makes the existing cores a bit more efficient but doesn't replace true cores obviously. I'm sure part of the answer lies in the details of production while others are a matter of marketing. Especially given all the various cpu's in a family.

I've never personally understood why there's a 4690k, 4690, 4590, 4460, 4440 all separated by just a couple hundred mhz and priced $10-20 apart from one another. Pretty soon we'll be able to custom order our cpu's perhaps. Be able to tell them no, 3.3ghz doesn't seem like enough, 3.5ghz seems like too much, I'd like to save $3.50 if I could just get it at 3.42ghz? A low end i5 (slower cheaper) a basic i5 (slightly more) and a k series (extra cost for unlocked feature) seems the easiest route to go but that's just my thought on it. Too many options can get confusing.

Liam m · Jun 1, 2016

Yes, I totally agree with you... Intel could better only sell a low-mid-high end i5... If you buy a processor, you almost regret not buying the one above it for 15$ more... At least that's how I am..

And I just tested cinebench r15 to check the difference between 1 core and 1 core with HT, but that's impossible to check with cinebench? I have i7 4700mq (Locked at 3ghz max with intel extreme tuning utility for testing purposes because if I use all cores that's also at 3ghz max) and I get 118cb with one thread (thus 1core) and 235cb with 2 threads selected in options, which is just about the double score of one core, so I assume it just used 2 real cores when entering 2 threads in the settings? This question is a little bit off topic I think, but maybe you or somebody else who sees this knows this

(I must say these results are not completely representative because I only tested one run, and I had a ppt presentation open for studying because I have exams)

Rogue Leader · Jun 1, 2016

Groveling_Wyrm :

Intel doesn't have only one machine producing chips. They have hundreds maybe even thousands. It makes sense to figure out the most efficient combinations and then devise those recipes and put them to the machines for what they need. Now what I did say about separating between 2 and 4 cores was not always the case. I know for sure in the past they just cut chips down should they need to. But in fact having separate setups for 2 and 4 cores makes a lot of sense, when you are producing on the scale that they are.

Don't forget they also make CPU's that have 8, 10, 20, 28 cores and they are all dimensionally the same, Some combos of those would also need to have their own recipes and produced differently than their bread and butter 2 and 4 core chips. I don't know what the breakdown is there, but I am sure there is one.

Your former engineer is correct, and that is what was done, however I know Skylake is done this way, maybe even some other gens before it (I don't know for sure and Im not going to speculate or keep calling my friend) are done this way.

For my own self interests I hope they keep working this way, it means they keep buying machines from companies like mine, and need service from our folks in configuring the machines

You are correct though I have no data, because my friend is a real life friend with whom I discuss these things with (amongst other things like women cars and beer), and my company whose name I will not share, would not like me to disclose anything more than basic info that I know (and as well I'm not an engineer so I only have secondary knowledge, TBH I probably don't know anything that isn't public anyway).

Liam m :

I'm not 100% on this but I don't think you can force a hyperthread to work as a core. It works more along the lines of "ok all 4 real cores are occupied, time to hyperthread 4 more".

Liam m · Jun 1, 2016

Rogue Leader :

Groveling_Wyrm :

Intel doesn't have only one machine producing chips. They have hundreds maybe even thousands. It makes sense to figure out the most efficient combinations and then devise those recipes and put them to the machines for what they need. Now what I did say about separating between 2 and 4 cores was not always the case. I know for sure in the past they just cut chips down should they need to. But in fact having separate setups for 2 and 4 cores makes a lot of sense, when you are producing on the scale that they are.

Don't forget they also make CPU's that have 8, 10, 20, 28 cores and they are all dimensionally the same, Some combos of those would also need to have their own recipes and produced differently than their bread and butter 2 and 4 core chips. I don't know what the breakdown is there, but I am sure there is one.

Your former engineer is correct, and that is what was done, however I know Skylake is done this way, maybe even some other gens before it (I don't know for sure and Im not going to speculate or keep calling my friend) are done this way.

For my own self interests I hope they keep working this way, it means they keep buying machines from companies like mine, and need service from our folks in configuring the machines

You are correct though I have no data, because my friend is a real life friend with whom I discuss these things with (amongst other things like women cars and beer), and my company whose name I will not share, would not like me to disclose anything more than basic info that I know (and as well I'm not an engineer so I only have secondary knowledge, TBH I probably don't know anything that isn't public anyway).

Liam m :

I'm not 100% on this but I don't think you can force a hyperthread to work as a core. It works more along the lines of "ok all 4 real cores are occupied, time to hyperthread 4 more".

Thanks for the reply on the binning process!

And for the HT: Yes that could be the case indeed. I guess I can test the HT effect then when I put 4 threads vs all (8) threads, or just go to msconfig and set up 2 cores, because then windows starts up with 1 physical core and two logic cores I think and compare that to single core percormance... But that's for another time, now back to study.

teknobug · Jun 1, 2016

Rogue Leader :

This reminds me of Unix's "make -j" flag, -j for jobs (technically like threads) so compiling a Linux kernel with -j4 on a single core would choke the system and fail to complete the compile, I used to use -j4 or sometimes -j6 on a dual Pentium Pro system and it would fly, but would fail on my AMD Thunderbird system, today I can probably get away with make -j16 on an i7.

Liam m · Jun 2, 2016

teknobug :

That's something I don't know nothing about...

Liam m · Jun 3, 2016

Is the reason the TDP's of different i5's being the same, despite having clock speed differences also because the higher clocked i5's (i5 6500 and i5 6600) were more efficient at the end of the production process than lower clocked i5 cpu's (i5 6400) and thus having more or less same TDP?

i3 and i5: Turbo boost useless on i5??

Commendable

Don't

Commendable

Distinguished

Illustrious

Commendable

Distinguished

Commendable

Commendable

Distinguished

Distinguished

It's a trap!

Distinguished

It's a trap!

Distinguished

Commendable

Champion

Commendable

Champion

Commendable

It's a trap!

Commendable

Distinguished

Commendable

Commendable

Share this page