Return to Level1Techs.com

Has intel removed Multithreading before?

#1

The move from 8700K to 9700K, case and point.

Has this happened before? I’m asking because it gets me thinking about certain things.

0 Likes

#2

are you asking if they’ve reduced functionality on equivalent product lines or if they’ve changed the way they’ve marketed products?

every company does both

0 Likes

#3

For whatever reason. I suppose it’s reduced functionality. So yeah. Same product line. i7 with no multithreading. :unamused:

0 Likes

#4

I don’t think you know what multithreading means

do you mean hyperthreading (or SMT on amd products?)

0 Likes

#5

More than one thread per core?

0 Likes

#6

so this is the thing right

8 threads to 12 threads back to 8

Except 8 cores without SMT will function better than 4 cores with smt in most cases on the same intel uarch or revision thereof.

the 8700k was a deviation from their normal marketing and the 9700k will probably perform better than the 7700k

it’s mostly a marketing difference.

also “multithreading” generally refers to a processor delivering multiple threads, which can be done in a number of ways, whereas SMT/HT is a technology that enbles slightly more throughput by letting a single core execute on multiple threads at once.

0 Likes

#7

I don’t get it? Isn’t that an actual part of the processor’s design? More than one thread per core? 9700 should perform better than an 8700K simply because it’s the next one in line; it’s newer. :thinking: But regardless. Just a weird move. It’s like they wanted it reserved for people who are paying more.

1 Like

#8

Difference in enabled features, but you get better ipc on 1 thread per core than 2 threads on half the cores on most applications as well. even on the same uarch.

Functionally doesn’t matter but it is definitely a change in their marketing.

0 Likes

#9

Enabled. Implying it can be RE-ENABLED?

0 Likes

#10

probably not no, but most CPUs in a given product line are the exact same physical dies with sections enabled/disabled according to application needs.

Once upon a time AMD released some chips you could enable cores on, but that time is long past.

0 Likes

#11

I mean it happens all the time in the mobile line-up; good luck keeping track of which i5’s are dual-cores with hyperthreading, and which are native quad-cores without HT.

But in the standard desktop i7 lineup specifically, no, the Core i7’s have always had hyper-threading. However, there are no hard rules about the i3/i5/i7 line-up. The best way to think about it is generally the i3/i5/i7 is good/better/best respectively with the market segment (considering budget, power, and thermal constraints and performance needs, etc.). However, with the introduction of i9 series, the i7 is no longer the best, so they need a way to distinguish the i7’s from i9’s, with cannabilizing other product lines, and Intel decided HT would make the most sense.

I honestly the Intel product line up is too segmented, with your various colours of Pentium (e.g. silver, gold), I think Celeron is still a thing, and now your Core i3/i5/i7/i9. Do we really need that many performance tiers?

0 Likes

#12

So it’s like disconnecting part of a brain? Actual physical parts of the CPU? Would be neat if there was a good diagram or illustration; something for those of us that aren’t that deep into CPU microarchitecture.

0 Likes

#13

sort of.

It’s really expensive to design multiple chips, and you’ll have some defects anyway, so if you laser off or firmware disable certain areas on a defective chip, or a chip that isn’t selling well in its highest end iteration, you can get more out of your production line

where if you make physically different chips for every model, you multiply your production costs by the number of product SKUs.

1 Like

#14

I thought miulti-threading was more of an OS thing
“Win95 is both multi-tasking and multi-threading” was the talking point in 95 when the 486 was livin high on the hog

0 Likes

#15

As a more digestible real world example:

Nvidia only produces 3-4 chips per GPU generation

Usually code-named the first letter of the uarch name, and a 3 digit number.

The Titan X, Titan XP, and 1080Ti are all the same GP102 chips with varying levels of feature cutoffs on the die

The 1080, 1070 Ti and 1070 are all the same GP104 chips

The 1060 are a mix of ultra defective GP104 and regular GP106 chips

and the 1050/1030 have their own chips because the die is so much smaller that the yields are high enough not to concern themselves with defects.

so 5 chips for 12-14 different card models (depending on how you feel about the 9gbps 1060’s and the DDR 1030’s)

2 Likes

#16

its got a parallel meaning in the context of software. same word means 2 different things.

any multi core processor is “multithreaded” or has “multithreading”

HT/SMT/CMT are ways to shove more threads onto one core with various tradeoffs.

Multithreaded software is software designed to take advantage of multiple cores/threads

4 Likes

#17

this is just a way to keep the lesser chips from being waste and a way to force segmentation. it is done by all tech companies. amd does the same thing to the ryzen dies by using the ones with defects in the lower tier chips and using the perfect chiplets for the higher end chips.

0 Likes

#18

AMD probably has the least cost efficient approach because they have to keep so many different dies in production at once with their refresh cycle, funnily enough (for their gpus)

they do the same thing with CPUs though yeah.

1 Like

#19

You forgot the U/H/M/T/K/F subsku stuff

0 Likes

#20

But I though that per generation AMD had one chip design for their CPUs and every thing there after was multiples of chips or cut down chips depending on the direction the product was needing.

0 Likes