Fake cores vs Real Cores

[Wardaddy]

 Hey guys, I have a hypothetical question. If there are two processors- A has 2 physical cores but 4 logical cores and B has 4 physical cores. Both of them have same IPC and clock speed. So, under similar conditions and stress which will perform better especially in gaming.

[The Official Czex]

The real quadcore.

[SeanAngelo]

 Hey guys, I have a hypothetical question. If there are two processors- A has 2 physical cores but 4 logical cores and B has 4 physical cores. Both of them have same IPC and clock speed. So, under similar conditions and stress which will perform better especially in gaming.

B cause it has 4 REAL CORES

 

/thread

[Enderman]

4 physical

[JuztBe]

Basically you are comparing same clocked i3 vs i5

[Wardaddy]

The real quadcore.

 

 

B cause it has 4 REAL CORES

 

/thread

 

 

4 physical

Why???

[TheRandomness]

Why???

Hyperthreading doesn't make physical cores, it only streamlines the delivery of instructions iirc.

[Enderman]

Why???

because youre making 2 cores do the work of 4

obviously 4 cores doing the work of 4 is more efficient than two cores doing the work of 4

[The Official Czex]

Why???

 

HT is essentially making one core process two threads almost simultaneously (one after the other very quickly).

[Sam Z Man]

Hyperthreading doesn't make physical cores, it only streamlines the delivery of instructions iirc.

THIS

[Wardaddy]

The situation is hypothetical because i don't think i3, i5, i7 even of same generations can have the same IPC otherwise a 6th generation i5 wouldn't perform better than a 6th generation i3 even at same clock speed, which it does. Correct me if I am wrong

[Gikero]

https://www.youtube.com/watch?v=wnS50lJicXc

Hyper Threading Fast as possible.

 

[Wardaddy]

Ok! so is there any method  to quantify the difference in performance approx.

[don_svetlio]

The situation is hypothetical because i don't think i3, i5, i7 even of same generations can have the same IPC otherwise a 6th generation i5 wouldn't perform better than a 6th generation i3 even at same clock speed, which it does. Correct me if I am wrong

i3s actually have much higher clockspeeds (i3 4170 has 3.9 vs the i5 4460 at 3.2) but the i5 still performs better due to HT. The same reason even mobile i7s outperform desktop i5s

[don_svetlio]

Ok! so is there any method  to quantify the difference in performance approx.

a HT is about 20-40% of an actual core depending on your use.

[Wardaddy]

i3s actually have much higher clockspeeds (i3 4170 has 3.9 vs the i5 4460 at 3.2) but the i5 still performs better due to HT. The same reason even mobile i7s outperform desktop i5s

i3 4010u and i5 4210u both with 4 logical cores at 1.7ghz... but i5 performs better, that's why i was guessing the IPC of i5 must be greater

[don_svetlio]

i3 4010u and i5 4210u both with 4 logical cores at 1.7ghz... but i5 performs better, that's why i was guessing the IPC of i5 must be greater

Mobile is different

"U" means low voltage so lower clockspeeds

i3 is limited to 1.7GHz? i5 starts at 1.7GHz but goes to 2.7GHz if thermals allow. It's called "Turbo Boost" and is only present on i5s and above.

[runit3]

The situation is hypothetical because i don't think i3, i5, i7 even of same generations can have the same IPC otherwise a 6th generation i5 wouldn't perform better than a 6th generation i3 even at same clock speed, which it does. Correct me if I am wrong

Every generation (at least on the Intel side) maintains the exact same IPC across all model lines of the CPU's. There is nothing that differentiates them aside from active cores/threads and cache size.

 

If you took an i5-4690K and an i7-4790k with HT turned off, clocked them exactly the same under the same hardware circumstances, they will receive identical scores/results except those results which differ as a consequence of the different cache size between the two chips.

[don_svetlio]

Every generation (at least on the Intel side) maintains the exact same IPC across all model lines of the CPU's. There is nothing that differentiates them aside from active cores/threads and cache size.

 

If you took an i5-4690K and an i7-4790k with HT turned off, clocked them exactly the same under the same hardware circumstances, they will receive identical scores/results except those results which differ as a consequence of the different cache size between the two chips.

Cache is a big factor.

[Glenwing]

Ok! so is there any method  to quantify the difference in performance approx.

 

Yes, here:

 

Columns are

3.0GHz

3.0GHz with hyperthreading

4.5GHz

4.5GHz with hyperthreading

 

[runit3]

Cache is a big factor.

Depends on the test, but yes, it can be. Why people still try to use a 5960X chopped down with inactive cores for comparisons is beyond me.

[don_svetlio]

Depends on the test, but yes, it can be. Why people still try to use a 5960X chopped down with inactive cores for comparisons is beyond me.

Cause Linus is lazy

[runit3]

Cause Linus is lazy

If I had a stack of sponsored Intel chips that I could throw into any bench I would I'm silently waiting for the day we get a 40MB cache on a reasonably priced consumer chip.

[don_svetlio]

If I had a stack of sponsored Intel chips that I could throw into any bench I would I'm silently waiting for the day we get a 40MB cache on a reasonably priced consumer chip.

1GB HBM cache?

[Darkseth]

To be more precisely on Hyper Threading: This technology utilizes the "parts" of a Core, that are not fully used (because not all instruction sets, etc can be used), and makes a "virtual core" of that.

 

Thats why even if an Application can use HyperThreading fully, the virtual Core will have 20-30% of the real core.

 

its basicly, telling a human to do some small stuff with his feet, while he focuses on Working with his Hands. (hope that comparison isnt too off lol)