Intel now offering hyperthreaded Pentium CPUs!

[Energycore]

1 minute ago, Stefan1024 said:

The size of every single one not the amount : )

Yeah having too many would definitely weird out your partner

[zMeul]

6 minutes ago, Stefan1024 said:

When will people learn that hyperthreading is only a way to utilize the coes you have a bit better (about 20% in best case szenario)? A dualcore with HT is worse than a 3 core without HT.

 

Only advantage I can see:

- get around the "low threadcount lock", but a proper coded application (games tend to be broken all over the place) should not have such a requirement but scale smoothly from 1 to 256 cores (it may have negative speed up but should just run)

- just show off you thread count in task manager (or don't be ashamed of showing two)

huh!?!?

 

both the Kaby Lake Pentium and i3 (6100) have two cores with HyperThreading

[Stefan1024]

2 minutes ago, zMeul said:

huh!?!?

 

both the Kaby Lake Pentium and i3 (6100) have two cores with HyperThreading

It's a general statement towards people overestimating the impact of HT or specific pentium to i3 related.

[leadeater]

3 minutes ago, zMeul said:

huh!?!?

 

both the Kaby Lake Pentium and i3 (6100) have two cores with HyperThreading

Yea but the point was the previous Pentium G series did not have HT and adding HT barely does anything at all, 3 core no HT is better than 2 core with HT by a significant margin.

[zMeul]

Just now, Stefan1024 said:

It's a general statement towards people overestimating the impact of HT or specific pentium to i3 related.

does having HT helps? yes it does (for software that actually knows how to use it)

[zMeul]

2 minutes ago, leadeater said:

Yea but the point was the previous Pentium G series did not have HT and adding HT barely does anything at all, 3 core no HT is better than 2 core with HT by a significant margin.

there's a problem with that logic: there's no 3 core CPU available on the current market

 

the problem should be put this way: does HT hurt performance? in a meaningful way - yes / no

then why shouldn't it be there? there are cases where games won't even start on 2 core CPUs

[leadeater]

1 minute ago, zMeul said:

does having HT helps? yes it does (for software that actually knows how to use it)

But a second thread on a fully utilized core can do little to add performance. HT is designed to make better use of a core since it is common for a core to not be fully utilized even when the thread CPU time is 100%, this is where HT helps.

[zMeul]

6 minutes ago, leadeater said:

But a second thread on a fully utilized core can do little to add performance. HT is designed to make better use of a core since it is common for a core to not be fully utilized even when the thread CPU time is 100%, this is where HT helps.

that's not exactly how HT works - that's an oversimplification to the extreme

 

there's actual HW logic behind SMT, but the software needs to be fully aware for it to produce gains

[Stefan1024]

1 minute ago, zMeul said:

does having HT helps? yes it does (for software that actually knows how to use it)

It does if your workflow allows parallelism and your programm tends to stall the pileline. But an additional core is just way better.

 

4 minutes ago, zMeul said:

there's a problem with that logic: there's no 3 core CPU available on the current market

Wasn't there one from AMD at one point? (Quadcore with one core diabled)

But that's only to measure the speed up, a quadcore is even better than a dualcore with HT.

3 minutes ago, zMeul said:

that's not exactly how HT works - that's oversimplified 

Actually it's pritty close. HT just uses the gaps in the pipeline if your primary thread would create a stall or only utilize a part of the CPU (e.g. very FP heavy and interger units are free).

[NumLock21]

Will the cpu fans have RGB more importantly will the cpu itself have RGB?

[leadeater]

4 minutes ago, zMeul said:

that's not exactly how HT works - that's an oversimplification to the extreme

Quote

For each processor core that is physically present, the operating system addresses two virtual (logical) cores and shares the workload between them when possible. The main function of hyper-threading is to increase the number of independent instructions in the pipeline; it takes advantage of superscalar architecture, in which multiple instructions operate on separate data in parallel. With HTT, one physical core appears as two processors to the operating system, allowing concurrent scheduling of two processes per core. In addition, two or more processes can use the same resources: if resources for one process are not available, then another process can continue if its resources are available.

No pretty much exactly how I explained it, yes simplistically but not overly so. A core can only do so much work and a single thread can fully use that core making HT utterly redundant, this is just uncommon in the desktop computing space.

[NumLock21]

2 minutes ago, Stefan1024 said:

It does if your workflow allows parallelism and your programm tends to stall the pileline. But an additional core is just way better.

 

Wasn't there one from AMD at one point? (Quadcore with one core diabled)

 

Yes it's was the AMD Athlon II X3 a triple core cpu. During those days, AMD disabled them cores, but mobo makers let you unlock them call cpu unlocker or something through the bios. So some buys a cheap dual core and unlock it to the more expensive quad core.

[Vode]

Haha isn't this the first time since the Netburst based Intel Pentium 965 EE that we see Hyperthreaded Pentiums? Has been over ten years since such a SKU has been released.... 

[zMeul]

@leadeater @Stefan1024 you are both oversimplifying SMT and I'm too tired to dig up Intel's documentation

HyperThreading can deliver 20% gains from workloads that are SMT aware

[Vode]

Just now, zMeul said:

@leadeater @Stefan1024 you are both oversimplifying SMT and I'm too tired to dig up Intel's documentation

HyperThreading can deliver 20% gains from workloads that are SMT aware

Yeah, the short version is Hyperthreading can bring up to 20% performance gains depending on the workload. This can also mean 0% in a lot of cases.

[Stefan1024]

5 minutes ago, zMeul said:

@leadeater @Stefan1024 you are both oversimplifying SMT and I'm too tired to dig up Intel's documentation

HyperThreading can deliver 20% gains from workloads that are SMT aware

Exactly what I said: 20 improvement in best case szenario:

Dualcore with HT: 1.2 x 2 = 2.4

3 coes: 1 x 3 = 3

Quadcore: 1 x 4 = 4

 

More cores allway winns compared to HT.

[leadeater]

1 minute ago, zMeul said:

@leadeater @Stefan1024 you are both oversimplifying SMT and I'm too tired to dig up Intel's documentation

HyperThreading can deliver 20% gains from workloads that are SMT aware

I'm not saying it can't increase workload performance, I'm saying there are cases where it can't.

 

Here's a real world case for an actual application that does benefit quite well from HT.

https://support.microsoft.com/en-us/kb/322385

[zMeul]

1 minute ago, Vode said:

Yeah, the short version is Hyperthreading can bring up to 20% performance gains depending on the workload. This can also mean 0% in a lot of cases.

as I said, if the workload is SMT aware - that was the whole point of the debate wasn't it!?

[leadeater]

1 minute ago, zMeul said:

as I said, if the workload is SMT aware - that was the whole point of the debate wasn't it!?

No it was 3 core = better than 2 with HT every time.

[porina]

26 minutes ago, Stefan1024 said:

When will people learn that hyperthreading is only a way to utilize the coes you have a bit better (about 20% in best case szenario)? A dualcore with HT is worse than a 3 core without HT.

Historically the best throughput I've seen from having HT on compared to off, with all else the same, was 50%. So in best case, a dual with HT is about equivalent to 3 cores without. As a recent example of software doing this, try the AMD Ryzen Blender benchmark. I was seeing above 50% improvement from turning on HT, although it is within margin of error I hesitate to say it was actually greater.

[zMeul]

4 minutes ago, leadeater said:

I'm not saying it can't increase workload performance, I'm saying there are cases where it can't.

then what's the point of the whole debate?

it's not like you can go and buy a 3 core CPU with the same price / perf

 

AMD did had a 3 core CPU a while ago and it went as fast as it appeared - 8y ago

[leadeater]

1 minute ago, zMeul said:

then what's the point of the whole debate?

it's not like you can go and buy a 3 core CPU with the same price / perf

You know Intel could release a 3 core SKU if they wanted to right...

[zMeul]

Just now, leadeater said:

You know Intel could release a 3 core SKU if they wanted to right...

yes, I'll go on a acid trip too 

[Stefan1024]

2 minutes ago, porina said:

Historically the best throughput I've seen from having HT on compared to off, with all else the same, was 50%. So in best case, a dual with HT is about equivalent to 3 cores without. As a recent example of software doing this, try the AMD Ryzen Blender benchmark. I was seeing above 50% improvement from turning on HT, although it is within margin of error I hesitate to say it was actually greater.

That's interesting to see.

But this implies the software produces terrible pipeline stalles and is probaly badly optimized.

[porina]

2 minutes ago, Stefan1024 said:

That's interesting to see.

But this implies the software produces terrible pipeline stalles and is probaly badly optimized.

Maybe, but it isn't alone. I don't have the exact values to hand, but Cinebench was over 30% in similar situation. Put it another way, if it didn't give any advantage, there wouldn't be much point to having it. Competitive benchmarkers will also go for the HT model too, as it generally gives some advantage in many benchmarks.

 

I've not come across the 20% value attributed to Intel being thrown around here. I'd guess they may be indicating an average, but specific cases can be far above or below that. I know of tasks where the benefit is approximately zero also...