Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price chart

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Johnny5k
Posts: 5
Joined: Mon Feb 01, 2016 1:41 pm

Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price chart

Post by Johnny5k » Tue Feb 09, 2016 2:17 am

I made a spreadsheet comparing all the available Haswell & Skylake Intel Core i3, i5 and i7 CPUs. My goal was to find the best low-power processor for use in a low-profile HTPC. I thought I'd post it in case anyone else finds it useful. Let me know if you do!

https://goo.gl/LKckKP

I didn't include Celeron or Pentium CPUs. If people are interested, I could add those, to make it a complete chart; but for my needs, I was concentrating on the Core i- lines.

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by xan_user » Tue Feb 09, 2016 7:46 am

i5 4430 would be a better choice over the intentionally crippled "t" part. you could always downclock it a little if you found it to be too hot in your system.

Johnny5k
Posts: 5
Joined: Mon Feb 01, 2016 1:41 pm

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by Johnny5k » Tue Feb 09, 2016 3:26 pm

xan_user wrote:i5 4430 would be a better choice over the intentionally crippled "t" part. you could always downclock it a little if you found it to be too hot in your system.
Thanks for the suggestion. The i5-4430 has a PassMark score of 6269 @ 84W TDP, vs. 6700 for the 4690T @ 45W. So while it might already be intentionally crippled (underclocked/volted), it's still getting better benchmark speeds than the un-crippled 4430, and they're about the same price. Since this is for a low-power (quiet/cool) system, I would have to uc/uv the 4430 quite a bit to get it down to 45W TDP to match the 4690T, in which case it would be even slower. I may be able to uc/uv the 4690T a little more to get it to 35W, but to get the 4430 down that low I'd need to uc it by about 37%, and uv it by about 0.2V, which sounds like a lot. I've never under- or over-clocked before, so Idk how much you can push – or in this case, pull – the values before leading to an unstable system. And since I'll never want it higher than 45W, and Intel has already tested the 4690T to be stable at 45W, I think it's the better choice for me.

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by Vicotnik » Tue Feb 09, 2016 3:43 pm

What beast of an HTPC are you building that you need an i5? Why isn't a Pentium enough?

And this TDP thing again.. ;) The factory underclocked parts are nothing special. The T-part is not more efficient, it's the same CPU with different settings. Get the 4690T if you want, but please do so because you like the turbo frequency or something real. Not because you believe that it draws 45W while the 4430 draws 84W doing the same work because it doesn't work like that.

Forget about TDP. Forget about PassMark. Or at least push them aside a bit.

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

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by CA_Steve » Tue Feb 09, 2016 10:02 pm

Vicotnik wrote:And this TDP thing again.. ;) The factory underclocked parts are nothing special. The T-part is not more efficient, it's the same CPU with different settings. Get the 4690T if you want, but please do so because you like the turbo frequency or something real. Not because you believe that it draws 45W while the 4430 draws 84W doing the same work because it doesn't work like that.

Forget about TDP. Forget about PassMark. Or at least push them aside a bit.
+1

Intel TDPs are just vague bins for mostly marketing purposes. Take a T and non-T part, set them at the same freq and voltage and they'll perform exactly the same. So, figure out your use conditions and pick a part that meets it.

Johnny5k
Posts: 5
Joined: Mon Feb 01, 2016 1:41 pm

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by Johnny5k » Thu Feb 11, 2016 1:50 am

Vicotnik wrote:What beast of an HTPC are you building that you need an i5? Why isn't a Pentium enough?

And this TDP thing again.. ;) The factory underclocked parts are nothing special. The T-part is not more efficient, it's the same CPU with different settings. Get the 4690T if you want, but please do so because you like the turbo frequency or something real. Not because you believe that it draws 45W while the 4430 draws 84W doing the same work because it doesn't work like that.

Forget about TDP. Forget about PassMark. Or at least push them aside a bit.
The system will be primarily for watching video, but I'll use it for converting video as well. While a non-T CPU would encode that video faster, it would also create a lot more heat – too much heat for the cooling available in a system of this size (80mm tall case with a 3.5" drive above the cpu/cooler – so a sub-30mm cooler is necessary; and no case fan, if temperatures allow). So I would have to underclock the non-T CPU to make sure it doesn't get too hot. So isn't the T CPU a better choice, because it still has Turbo mode, whereas a manually underclocked non-T version would lose that advantage?

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by xan_user » Thu Feb 11, 2016 3:36 am

Johnny5k wrote: So I would have to underclock the non-T CPU to make sure it doesn't get too hot.
I bet a beer, that you dont even have too.




besides.....it will downclock on its own if you do occasionally run it too hot....

lodestar
Posts: 1683
Joined: Fri Aug 05, 2005 3:29 am
Location: UK

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by lodestar » Thu Feb 11, 2016 3:47 am

Johnny5k wrote:So isn't the T CPU a better choice, because it still has Turbo mode, whereas a manually underclocked non-T version would lose that advantage?
Yes the loss of Turbo modes is the major reason why under clocking down to T levels is not the same as using a T CPU. But bear in mind that all Haswell and SkyLake CPUs idle by default at around 0.8Ghz and 0.7V, and under low to medium levels of CPU utilisation can increase clock speed and voltage without necessarily reaching the level where Turbo modes kick in. With the SkyLake 6700T for example it will not exceed 2.8Ghz in this scenario when contributes partly to its lower calculated TDP. There is also a contribution from the Turbo modes scheme for the 6700T, 3.6Ghz for 1 core, 3.5Ghz for 2 cores, and 3.4Ghz for 3 or 4 cores which is less aggressive than the vanilla 6700. If you are leaning towards a T processor such as the 6700T it may help that it is priced somewhat less than both the 6700 and 6700K. There is a performance differential but for non-gaming purposes it may be of no real significance.

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by quest_for_silence » Thu Feb 11, 2016 5:31 am

Johnny5k wrote:The system will be primarily for watching video, but I'll use it for converting video as well. While a non-T CPU would encode that video faster, it would also create a lot more heat – too much heat for the cooling available in a system of this size (80mm tall case with a 3.5" drive above the cpu/cooler – so a sub-30mm cooler is necessary; and no case fan, if temperatures allow). So I would have to underclock the non-T CPU to make sure it doesn't get too hot. So isn't the T CPU a better choice, because it still has Turbo mode, whereas a manually underclocked non-T version would lose that advantage?

Just some data for your reasonings, use some pinches of salt.

http://en.gecid.com/cpu/intel_core_i5-4 ... s#graphics

As I told you elsewhere, probably a Core i5 6500 could be a somewhat better option.

Vicotnik
*Lifetime Patron*
Posts: 1831
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by Vicotnik » Thu Feb 11, 2016 8:38 am

Johnny5k wrote:The system will be primarily for watching video, but I'll use it for converting video as well. While a non-T CPU would encode that video faster, it would also create a lot more heat – too much heat for the cooling available in a system of this size (80mm tall case with a 3.5" drive above the cpu/cooler – so a sub-30mm cooler is necessary; and no case fan, if temperatures allow). So I would have to underclock the non-T CPU to make sure it doesn't get too hot. So isn't the T CPU a better choice, because it still has Turbo mode, whereas a manually underclocked non-T version would lose that advantage?
If the load is 100% then yes, the T-version would take a bit longer but keep within a smaller power envelope than the non-T version. But if you are doing something not so heavy in real time I think that the two CPUs would perform the same. With the non-T version you would have more headroom; if you experience heat problems, which I like xan_user think is unlikely, you could limit the max multiplier or something and have pretty much the same CPU as the T-version sans better turbo mode.

In your case I would probably also consider the T-version. That turbo mode is sweet and Intel knows it. But cost would probably be the number one consideration, which leads me to the Pentium. It can convert video as well you know. ;) And despite what the TDP values says it would probably be cooler as well.

In either case, those TDP numbers should be taken with a grain of salt. The T and non-T are usually closer than the official numbers suggests. And PassMark is only one benchmark among others, it doesn't reflect real life performance on it's own. Also it seems to favor the T-version and its turbo mode, while you hardly need great single thread performance for your use case.

Wild Penguin
Posts: 83
Joined: Mon Jul 01, 2013 3:30 pm

Re: Intel Core i3-i5-i7 Haswell/Skylake TDP/PassMark/Price c

Post by Wild Penguin » Fri Feb 26, 2016 6:14 am

CA_Steve wrote:
Vicotnik wrote:And this TDP thing again.. ;) The factory underclocked parts are nothing special. The T-part is not more efficient, it's the same CPU with different settings. Get the 4690T if you want, but please do so because you like the turbo frequency or something real. Not because you believe that it draws 45W while the 4430 draws 84W doing the same work because it doesn't work like that.

Forget about TDP. Forget about PassMark. Or at least push them aside a bit.
+1

Intel TDPs are just vague bins for mostly marketing purposes. Take a T and non-T part, set them at the same freq and voltage and they'll perform exactly the same. So, figure out your use conditions and pick a part that meets it.
+1... I've been in a similar situation and made the same misevaluation previously as the OP. I think one way to understand this, is to read about intel P-states and understand what they are used for. The way I think about this is, that for any CPU time consuming task, it is usually most power-efficient (which also means most efficient in terms of generated heat and noise) to do a task as fast as possible.

TDP can not be used as an efficiency meter in terms of [ CPU work done ] / [ Wh of electricity used ] (or vice versa, as in your spreadsheet). Think of TDP more like as a rough guideline for choosing adequate cooling solution if you want to cool your CPU in cases of 100% usage for prolonged time.

What really matters in choosing a CPU is your use case - constant, heavy usage, very little CPU usage, or burst of high CPU usage - or combinations of these. Also the amount of threads affects the choice (roughly as: normal UI and light office / browsing -> 2 cores is enough, heavy gaming rig and enthusiast hobby projects -> 4 cores, workstation for rendering, folding or similar etc -> more than or at least 4 cores). Dividing the same workload on a longer period of time means that the heat can be dissipated slower, although more W of heat will be generated in total in the end.

Post Reply