Intel's Kaby Lake Branding Confounds, But The Real Devil Is In The TDP Details

Status
Not open for further replies.
What I don't understand are Intel's tricks to making the TDP so low. We have the I7-7500U. 2.7Ghz turboes to 3.5Ghz. Then we have the desktop Pentium G4400 3.3Ghz with a 54W TDP. Seems like a large gap to me. Then again their TDP numbers are sometimes inaccurate I think.
 

Discorama

Commendable
Aug 25, 2016
7
0
1,510
Turbo frequencies are pushed pretty high. Too bad most of the devices these CPUs will be used in will throttle down due to heat and never use them.
But yeah, really comes down to how well the 615/620 perform as that is always Intel's weakness.
 


Why are you comparing a low power laptop chip to a desktop chip?

Intels TDP is normally the max power a CPU can dissipate before it throttles. For the i7-7500U it is a low power part so it has limitations put in place to keep it from hitting higher power numbers.

One benefit the i7 has is a more mature process. One example is the i7 4770K and i7 4790K:

http://ark.intel.com/compare/80807,75123

The i7 4790K had a 4w TDP disadvantage but a 500MHz clock speed advantage and clocked higher in overclocking. A more mature process tech and tweaks can help the TDP of a CPU.
 


I was just wondering, if you have a laptop and desktop chip of the same architecture. Let's say they have the same number of cores and threads. Let's say same frequency. Does the laptop part still perform worse and cooler for any reason other than less cache?
 

TJ Hooker

Titan
Ambassador

An educated guess:
1) Unlike desktop parts, mobile parts (especially -U and -Y) are typically only expected to be able to reach maximum boost clocks for short periods before throttling. This is why a low power mobile CPU may perform worse than a desktop chip with all the same specs (including boost clock).
2) I believe mobile parts are binned for low power consumption (i.e. low leakage) which, combined with 1), explain why they can have lower power consumption than a desktop part with seemingly equal specs.

Also, like you said, TDP isn't really an accurate measure of power consumption. It's actual purpose is to designate the general level of performance (i.e. heat dissipation capability) a heatsink would need to cool that part. The fact that a i5 6600 and 6400 have identical TDP shows that it's not exact. If I were to speculate, maybe Intel really dials in the rated TDP on their mobile parts because of how constrained cooling/power is on the kinds of systems that would use those chips, whereas with desktop they just ballpark it (and overestimate for the sake of erring on the side of caution), because it's not as important. Like I said though, this is just my own thoughts on the matter.
 

bit_user

Polypheme
Ambassador
Good reporting, Paul.

The new branding makes it a whole lot harder for the average user to determine just what Intel is inside (as it were)
Oh... Snap!

I'm not too worried about manufacturers permanently down-clocking a more expensive part, because it would cost them more than buying a cheaper SKU and overclocking it.

My guess is that people are going to spend a lot of time looking at graphs of clock speeds, but that's already the case, for mobile.
 

bit_user

Polypheme
Ambassador
On a related note, anyone who has ever puzzled over the T_case specifications of the desktop processors would do well to check out Section 8 of this:

http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html
 

Alex Atkin UK

Distinguished
Jun 11, 2012
51
2
18,545
These are great features, except manufacturers are completely useless at utilising them properly.

eg. The 1.9Ghz in my laptop throttles to 500Mhz when the battery drops to 33%. Lenovo don't seem to understand why and in fact the latest BIOS update not throttles when on AC too when it didn't before. A complete a utter mess.
 

TJ Hooker

Titan
Ambassador

Don't want to get too off topic here, but I'm a little skeptical of that guide. In Section 5, he describes how there is a 5 degree offset between Tcase and Tcore, and provides a link (and refers to a specific figure in that link). However, if you follow the link, neither that figure nor any other part of the paper backs up his claim. Using a source that doesn't support what you claim it does is a pretty big red flag for me, not only for that claim but for everything else he says.
 

srmojuze

Commendable
Jul 28, 2016
26
0
1,540
I won't comment on the ethical aspects of this. Suffice to say this has been going on for a while and it's partly due to Intel making good chips that shine (and are profitable) at the high-end, because:

Unfortunately the low-end is where the rubber meets the road. Intel has kind of made their CPUs ~so good~ that a chopped-up-high-end CPU at low clocks and low TDP is perfectly fine for most "notebook" or "ultrabook" users, particularly when paired with 8GB of RAM and decent SSDs.

For anyone expecting this trend to change, it isn't... the "Core" branding has always been indicative at best (whether you agree with their approach or not) and is quite obtuse at the low-end.

The layperson most likely will just see "Core", "i5", or "m5" (short-lived), "Intel", "Xth gen" (even then it's more of a salesperson thing - latest generation m'aam!".

OEM branding and features are more of the differentiating factor, take Lenovo "Yoga" for example - at face value one may scoff at it but to be honest that's the most innovative thing in the laptop space besides Surface 3/4, at least since the Macbook Air.

Now, to be fair, in the gaming laptop world, AFAIK the rated clockspeed (non-turbo) is considered the minimum 4-core speed the laptop is supposed to deal with. Eg. 2.6 ghz with "turbo" to 3.x ghz or what not essentially means when your laptop cooling is maxed the laptop is only rated to run at 2.6ghz. Check out ThrottleStop for all those interested. It's not an overclocking tool per se but very, very helpful to really see what's going on. Again, in the mainstream/ business space, no non-enthusiast would use ThrottleStop let alone easily grasp what TDP is.
 

ivonakiss

Commendable
Aug 30, 2016
2
0
1,510
I some earlier laptops - it was common to see laptops with cooling pads bellow them. Inadequate cooling meant you can't rely even on the "standard" clocks. These days this is more uncommon, but i think it is still valid. That is why I am so interested in the passive cooling - will it be able to really keep up - or will it just drag the whole system down
 
D

Deleted member 1353997

Guest
Accelerometers also allow the device to adjust performance based on device orientation. For instance, the device will switch to a higher-power mode when it is in a static 45-degree orientation (which indicates docking), as opposed to a 90-degree orientation, which indicates a user is holding it.
This is stupid. My Surface can have a static 45-degree orientation without being docked.
Besides, why adjust performance based on orientation and not on device temperature and battery state, like it's always been done before?
 
G

Guest

Guest
A lame CPU made by Intel for gay Surface type of devices. Stick with LGA2011-V3 platform or wait for Zen.
 

SylentVyper

Reputable
Dec 30, 2014
98
0
4,660
"What I don't understand are Intel's tricks to making the TDP so low."

Intel bases TDP only off of base frequency at load, not max turbo frequency.
 


That's the thing, though, they shouldn't have the same TDP. Taking the definitions of TDP, the 6600 should have a higher TDP since the additional power consumption from the higher clocks result in more waste heat, therefore, more heat needs to be dissipated henceforth higher TDP.
 

TJ Hooker

Titan
Ambassador

You're right, a 6600 will consume more power, and therefore technically require a beefier cooler. However, rather than have a different cooling requirement for each individual SKU, they simply specify a general requirement for CPUs in the same tier. So for a locked, Skylake desktop i5, you need a cooler capable of dissipating at least 65W. That doesn't necessarily mean that each of those CPUs (or even any of those CPUs) consumes 65W. It's just more convenient to specify a general cooling solution that works for a range of products, rather than saying you need a 55W cooler for a 6400, 60W for 6500 (making these numbers up), etc. Especially if you're an OEM who may not be using the stock Intel cooler, and you simply want a standard, minimum cost cooling solution that will work with every i5. That's my understanding of it anyway, but I don't have any official source for this so take it with a grain of salt.
 

tsteele93

Distinguished
Jan 15, 2015
2
0
18,510
I am in my 40's, I used to know a TON about PC's. I was (and still, often am) the guy that my friends and relatives came to for computer help. My first PC was an XT and then 286, 386, 486 etc... Back then it was processor number and clock speed.

Now, I don't have the faintest clue what I am buying. I try to get i7 and a great graphics card, but I've been fooled sometimes. Thinking I was buying something pretty great and finding out it was only a bit above average.

Guess what that does? It makes me wait longer to purchase. I do NOT spend more time researching. I spend less time researching and I try to make my computers last longer. Screw someone over and they don't trust you anymore and they don't like doing business with you anymore. AMD is no better. I haven't had ANY idea what their processors mean for ages and have avoided them like plague.

When I was younger, I often built my own PC's and would use AMD to get more bang for my buck. But at some point AMD quit being easy to decipher and I quit using them. Now Intel is no better, but since they are the leader and "gold standard" I still buy PC's with Intel chips. I don't build my own anymore because it is much more complex with the way that software and hardware interact and I generally prefer to just buy a PC pre-built.

My long-winded post is to say this: Crap like this makes people (like me) not buy PC's as often as we used to buy them. There is no longer an obvious migration to bigger and better. Intel has NO FRICKING CLUE about branding and selling. They SHOULD BE marketing (like automobiles) an easy-to-recognize, well branded top-of-the line processor that everyone wants, but then they can sell the next notch or two down to the average joe who wants to be associated with the premier processor they hear about. Instead, they muddy the waters so much that no one has any idea which processor is what or whether one is better than the other and there is NO MARKETING going on at all.

How can such smart people get so much so wrong?
 

bit_user

Polypheme
Ambassador
Even then, you had SX variants with half the bus width (or no math coprocessor, in the 486). And let's not forget DX2 (later DX3 and DX4) which had a clock multiplier between the CPU and Northbridge. Busses were ISA, then EISA (and MCA, for IBMs), VESA local graphics, and eventually PCI.

We could talk about modems, too. Not to mention graphics standards (CGA, EGA, VGA, SVGA, XGA, and a few more non-widescreen formats I've purged from memory), and the multitude of graphics chipsets. Disks were various flavors of IDE (anyone remember cable select?) or SCSI (with its variety of connectors, daisychaining and termination rules), and an even older standard that I forget. And I don't even know enough about DRAM and cache standards to get into that whole area.

Once you got all the hardware put together, you had to sort out IRQ and port number conflicts. Then, you had the software side of things, with autoexec.bat and config.sys. You needed enough free base memory, and then the mucking about with HiMem and EMM. Oh, and you want to network your PCs?

In many ways, it was a younger, more dynamic time. Overall, I'd say there was probably more the PC enthusiast had to know, parts were more expensive, and help was less readily available. And they were ugly (or some might say bland).

These days, PCs are much simpler to assemble. Just buy a CPU and mobo with the same socket. Add compatible RAM (I wish they'd standardized on a voltage, for the different DDR revisions & speeds, though). Plug in your storage and PSU, plug it into your monitor. Attach keyboard, mouse, network. Done. Everything (usually) just works.

As for performance, the biggest difference between then and now is the CPU core count, but many would argue that dual-core is still fine for typical home users. The other big variable you didn't have before is storage. But, as long as it's SSD, a typical user will tend to be happy with it.

I think we can end, right here. If you care what you're buying, you need to do a certain amount of research. That goes for cars, washing machines, TVs... pretty much anything you can think of. I don't know why you expect to look at one or two specs, on something as complex as a computer, and assume it tells you everything you need to know. You wouldn't even do that with food!

There are lots of people who make their money selling average (or worse) products to the unsuspecting or uncaring public. We all know this.

I will agree with you that Intel started with this oversimplified naming scheme: Celeron, Pentium, i3, i5, i7. Then, they failed even to follow it in a consistent way. The latest move doesn't help.

However, the information you need to make basic purchasing decision has never been more abundant or accessible. It still takes some work, but if you care about having a fast PC, then it's probably easier than in the "good 'ol days".

It seems like they had a good intention, with the current market segmentation scheme. But the seeds of its demise were sewn in it oversimplification. Then, the waters got clouded by too much short-term thinking and wanting to push more of the latest generations and cheaper SKUs on a public that sort of learned the naming scheme.

Perhaps it really comes down to a lack of discipline and strategy. That said, marketing is about oversimplification. What you want is transparency, which is decidedly not a marketing virtue.
 

bit_user

Polypheme
Ambassador
Status
Not open for further replies.