The idiocy of Intel's T desktop processors

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

The idiocy of Intel's T desktop processors

Post by CA_Steve » Tue Apr 03, 2018 6:30 am

Intel expanded the Coffee Lake U CPU offering, including a couple of 35W T parts. Here's the fun marketing spin bit via Anandtech:
The most interesting element to these values are the 35W low-powered T processors. In each case, the all core turbo is much, much higher than the base frequency. For example, the Core i5-8400T has a base frequency of 1.70 GHz, but the all-core turbo is set at 3.0 GHz - almost double. Given the fact that TDP is defined at the base frequency, it is quite clear that the all-core turbo mode suggested to motherboard manufacturers is going to blow that 35W limit on the i5-8400T.
In the past, T parts were basically nerfed non-T parts with lower base freq and tighter control over the turbo freq. Not so much, now. Buy a T part and you have to disable turbo in order to stay within the 35W TDP. I'm guessing the additional two cores ate up whatever headroom there was on TDP.

So, T parts: used to be fairly useless - now are completely useless. :)

HFat
Posts: 1753
Joined: Thu Jul 03, 2008 4:27 am
Location: Switzerland

Re: The idiocy of Intel's T desktop processors

Post by HFat » Sat Apr 07, 2018 4:47 am

Clueless reviewer strikes again. The TDP value isn't defined for any individual CPU, nevermind for a particular frequency. Typically, CPUs never consume as much power as their TDP.
Measurements would be required to confirm that this part is simply a factory-nerfed version of a similar non-T CPU but even so, there's no reason to suppose that the TDP needs to be exceeded just because the CPU is allowed to run faster. The CPU might want to draw more than its TDP depending on the code you ask it to run but I have to assume motherboards would be able to limit the CPU's power draw since it's something Intel mobos have let us configure a while. So there's in principle no need to disable turbo to make the CPU stay within its TDP.
Also, some people might actually want the CPU to exceed its TDP as long as it's not too hot.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Re: The idiocy of Intel's T desktop processors

Post by Olle P » Mon Apr 09, 2018 12:16 am

HFat wrote:Clueless reviewer strikes again... Typically, CPUs never consume as much power as their TDP.
As a blanket statement that's plain wrong nowadays, if one discuss power consuption at 100% load. (Admittedly, most CPUs are not demanded to run at 100% load most of the time.)

The TDP is pretty much defined as the power consumption when all cores run at the base speed.
At idle the CPU will draw less, but if then suddenly being asked to provide maximum ("turbo") performance it will draw more power for a couple of seconds before the motherboard is supposed to restrict the power usage to keep it within TDP.

Higher end motherboards are often more relaxed in their settings, allowing the CPU to use more power as long as the temperature is acceptable.
Source here.

AKM76
Posts: 19
Joined: Wed Mar 07, 2018 8:22 am

Re: The idiocy of Intel's T desktop processors

Post by AKM76 » Sun Apr 22, 2018 11:41 am

I totally agree that Intel's T marked processors is something confusing, to say the least.

I would understand if they'll have, say, a normal i5 8400 and then an i5 8400T that will have exactly the same frequencies and all other specs except being able to run on a lower wattage, for a somewhat higher price. That would be logical. Actually the perfect solution would be every individual processor having it's own TDP rating.

From what I was able to see around on the Internet, I came to a conclusion that there is nothing special about the T processors. For the same generation they are just all use the same technology, just different clocks. I would rather prefer to see another markings for them, in digits somehow, without the "T". I have no idea why they name them, say 8700 / 8700T? Basically simply looking at the clocks you can more or less guess how hot they will run if you have a some measurements to compare, no magic here.

Ah, I think "T" processors will throttle at lower temps... Maybe I miss something else...

BTW, I'm tremendously grateful to CA_Steve for recommending me the normal i5 8400 for the HDPlex H3 V2 build - it was exactly spot-on. I was VERY nervous and full of regrets after ordering it, thinking that I'd better have choose some T processor. Now glad I did not. The review is coming.

AKM76
Posts: 19
Joined: Wed Mar 07, 2018 8:22 am

Re: The idiocy of Intel's T desktop processors

Post by AKM76 » Mon Apr 23, 2018 11:18 am

I really wonder what am I missing. I'm looking at this:
http://cpu.userbenchmark.com/Compare/In ... 9vsm475176

i5 8400 vs. i5 8400T. To me it looks just like two 6-core processors, one of them is much slower. What really surprises me is that the slower one is more expensive. Why? What am I missing? I guess it should be 3 times colder to justify the higher price. Maybe it is of course, but I seriously doubt it.

Derek Semeraro
Posts: 124
Joined: Sat Sep 23, 2017 6:49 am

Re: The idiocy of Intel's T desktop processors

Post by Derek Semeraro » Tue May 01, 2018 5:18 pm

T processors are for the all-in-one desktop market that seeks to accomodate desktop processors at significantly lower power consumption to achieve a more compact size.

I'm assuming the T version is more expensive because they're selling less of them and it has a higher average cost to produce

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: The idiocy of Intel's T desktop processors

Post by CA_Steve » Wed May 02, 2018 9:23 am

Derek Semeraro wrote:I'm assuming the T version is more expensive because they're selling less of them and it has a higher average cost to produce
They aren't a different die or placed in a different package than the non-T parts. The laser part number marking is done post package level test...and the T parts are more likely to be parts that didn't meet the non-T part test limits for turbo rather than having to meet a tougher spec. So, at best they cost the same, at worst they are fallout parts and are a way to get rid of otherwise useless devices.

Derek Semeraro
Posts: 124
Joined: Sat Sep 23, 2017 6:49 am

Re: The idiocy of Intel's T desktop processors

Post by Derek Semeraro » Thu May 03, 2018 6:44 pm

It's a problem of demand - not supply. Intel sells these by the thousands to companies like HP for pre-built AIO's. Very few consumers actually want these, and there's not a great need for them since the DIY desktop market is already a niche and most DIY PC builders don't see the energy efficiency of desktops as a problem at all. Selling these to a very small niche was probably not too far of a decision from not selling them at all.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: The idiocy of Intel's T desktop processors

Post by CA_Steve » Thu May 03, 2018 8:38 pm

Regardless of their target market, maybe 1 out of 10 of the DIY'ers that come here for advice think the T processors are good thing. I get to spend time explaining why they aren't. :)

AKM76
Posts: 19
Joined: Wed Mar 07, 2018 8:22 am

Re: The idiocy of Intel's T desktop processors

Post by AKM76 » Mon May 07, 2018 5:35 pm

I think adding this "T" is a clever marketing decision. Otherwise they'll be just entry level, inexpensive, low demand items. Marketing them as T-graded make them look like something special, purposely engineered for the specific needs, rising their value for some niche markets (fanless PCs). That is really clever, almost tricked me to buy one. Again, they are indeed low TDP ones, which is not surprising given their clocks. I would say that normal processors like i3 8100, i3 8300, i5 8400 could be easily marked as T also, because they are far below 35W power consumption at non-turbo clocks, and also far below 65W at full turbo synthetic load. I know that power consumption is not exactly the same as TDP values, but they are really closely related. What I'm very curious is to compare the i7 8700T to i5 8400, both having relatively close clocks and number of cores. Something tells me the 8700T will be a tad hotter, because of some additional quirks they have, I'll be surprised to be wrong.

BrianF
Posts: 151
Joined: Sat Apr 16, 2005 2:59 am

Re: The idiocy of Intel's T desktop processors

Post by BrianF » Wed May 09, 2018 3:54 am

AKM76 wrote:...and also far below 65W at full turbo synthetic load.
Slightly OT, but I'd love to see that data. So far everything I've read shows that even a strait 8400 will, if allowed, continuously draw well over 100Watts during synthetic benchmarks (see Tom's review, specifically the Prime95 with AVX enabled test loop).

AKM76
Posts: 19
Joined: Wed Mar 07, 2018 8:22 am

Re: The idiocy of Intel's T desktop processors

Post by AKM76 » Thu May 10, 2018 8:05 pm

BrianF wrote:
AKM76 wrote:...and also far below 65W at full turbo synthetic load.
Slightly OT, but I'd love to see that data. So far everything I've read shows that even a strait 8400 will, if allowed, continuously draw well over 100Watts during synthetic benchmarks (see Tom's review, specifically the Prime95 with AVX enabled test loop).
Hi. I borrowed a watt meter from my friend to check what it draws from the wall. Also I use Core Temp software to monitor the internal data. By default there are a lot of things enabled for power saving in BIOS and Windows (10 is very much recommended as it use the most advanced latest Speed Shift/Speed Step technologies), and it is what 99.9% of users should use. I have a Mini-ITX Asrock z370 board, 16 gigs of RAM and 2 SSDs, plus USB audio interface (RME Babyface Pro) and a smartphone charging. At idle it draws about 25-27 watts from the wall. On a processor level at idle it draws about 2 watts. With disabled C-states, disabled core parking, and minimum limit for the clocks @ 2000 MHz (settings I need for audio work) it draws 6-7 watts at idle and Internet browsing (processor only). Watching some video adds about 2-3 watts extra (10 watts overall).

I'm not a big synthetic tests expert, I've run something in Prime 85 and also there is a "stress" button in the CPU-Z software. To me they do about the same thing. All cores are loaded 100%. Internally the software shows 46.6 Watts for processor package in this case (Core Temp utility). So, from the wall it'll be, add 40 to 27 (idle from the wall) - so about 67W from the wall at full synthetic(!) load. In reality with what I do (working with audio) I stay at about 10-12 watts internally 98% of time, with occasional load up to 20 watts for some short period audio rendering, files unpacking, etc. The processor temps stays at about 37-40 C for now (max at 51 C), with usual comfortable spring temperatures outside. Although it'll be very hot soon in the summer where I live. I use HDPlex H3 V2 Passive case, making an extensive review right now.


ADDED: sorry, as I said, I'm not a testing expert, just checked with the Prime85, set it to small FFT this time (I don't know what it is, also did not find where are the AVS settings and what it is) - so yes, 85 Watts on a processor level this time, which means slightly over 100 Watts from the wall.

Krall
Posts: 3
Joined: Fri Feb 28, 2020 3:03 pm

Re: The idiocy of Intel's T desktop processors

Post by Krall » Fri Feb 28, 2020 3:09 pm

CA_Steve wrote:
Thu May 03, 2018 8:38 pm
Regardless of their target market, maybe 1 out of 10 of the DIY'ers that come here for advice think the T processors are good thing. I get to spend time explaining why they aren't. :)
That would be me :) I've been looking around for an Intel Core i5-9600T for my mITX 'server', but they are like vaporware.

Is it possible to buy a the same CPU as a "K" chip and disable the turbo to have it produce less heat/consume less wattage?

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: The idiocy of Intel's T desktop processors

Post by CA_Steve » Sat Feb 29, 2020 6:04 am

Welcome to SPCR.

If you are buying an aftermarket mobo, there should be options to disable turbo, set a lower clock speed, and also under volt the CPU. Note that some of the lowest priced SKUs in the stack may not provide all of this functionality (like under volting).

That said, if you are using the server for video transcoding or other high horsepower tasks, you may wish to let the CPU run as is. It'll ramp up, do the task and ramp down quicker and the task power use should be in the same ballpark.

Krall
Posts: 3
Joined: Fri Feb 28, 2020 3:03 pm

Re: The idiocy of Intel's T desktop processors

Post by Krall » Sat Feb 29, 2020 9:27 am

CA_Steve wrote:
Sat Feb 29, 2020 6:04 am
Welcome to SPCR.

If you are buying an aftermarket mobo, there should be options to disable turbo, set a lower clock speed, and also under volt the CPU. Note that some of the lowest priced SKUs in the stack may not provide all of this functionality (like under volting).

That said, if you are using the server for video transcoding or other high horsepower tasks, you may wish to let the CPU run as is. It'll ramp up, do the task and ramp down quicker and the task power use should be in the same ballpark.
Thanks Steve and thanks for your help!

I'm not familiar with undervolting, I'll have to read up on it.

I went with a ASrock Z390M-ITX/ac MB. It'll run an m.2 drive and possibly a single HDD (for backup) in a older Silverstone ML05 mini-ITX case. No video transcoding, it's basically a box that'll run Quickbooks, MS Office, some scanning software and Anti-virus. It needs to be on 24/7 (though I think I can use wake on LAN) so I was really just looking for energy efficiency. I think the CPU is overkill, but there weren't a lot of in-stock options.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: The idiocy of Intel's T desktop processors

Post by CA_Steve » Sat Feb 29, 2020 12:39 pm

Honestly, All of the i5's will use the same amount of power for lower level loads. One thing you can do is go into device manager and turn off devices you aren't using (like the wifi if it's LAN-connected and vice versa).

Krall
Posts: 3
Joined: Fri Feb 28, 2020 3:03 pm

Re: The idiocy of Intel's T desktop processors

Post by Krall » Sat Feb 29, 2020 1:25 pm

CA_Steve wrote:
Sat Feb 29, 2020 12:39 pm
Honestly, All of the i5's will use the same amount of power for lower level loads. One thing you can do is go into device manager and turn off devices you aren't using (like the wifi if it's LAN-connected and vice versa).
Thanks again. I was mainly concerned that it was a K chip and listed at 95 watts compared to the T chips which are listed at 35 watts. I know in the environment I’ll be using it in it won’t climb to those figures or at least not for very long.

I’ll definitely go in there and disable unused devices. I need to figure out how to strip the online bloat of Win 10 where it’s always fetching apps for stuff.

Post Reply