Turbo modes ...

[Massman]

This topic came up at the last staff meeting: should we set the stock frequency of a CPU to the stock rating or the turbo rating of this CPU? I'm not against the idea of doing this ... after all, at stock settings the turbo mode will be enabled anyway.

 

I vote 'set turbo mode as stock frequency'.

[Eeky NoX]

Same here I would vote for But I'm not sure of what you meant exactly. Is to reduce the OC % for each CPU turboproof ?

 

Anyway, where is the poll mate ? ^^

[knopflerbruce]

IMO the "stock speed" is the speed you get when you use the resources the CPU has to offer, and that means loading all cores. If you get that frequency then, use it as stock.

 

You could also have two stock speeds, separated with a "/" or something...

[S_A_V]

What frequency you see on the box of the CPU?

What frequency you see in CPUZ (Specification field)?

What frequency CPU working when all cores/threads 100% loaded (on motherboards that not support fixed CPU multi for Turbo Boost)?

The stock rating, not the turbo rating.

You could also have two stock speeds, separated with a "/" or something...

At least four:

1. Stock speed

2. Turbo with only 1 core enabled

3. Turbo with only 2 cores enabled

4. Turbo with 3 or more cores enabled

[Antinomy]

What frequency CPU working when all cores/threads 100% loaded (on motherboards that not support fixed CPU multi for Turbo Boost)?

So if we take a Prescott-2M or Smithfield which overheat on a box cooler under 100% load (like S&M) and thus throttle via TM2 to 14x multi - then it's their stock?

They'll work on this frequency under load every time unless you change the cooler.

 

OTOH, the turboboost is just an expansion of C1E which lowers the multi at idle. This one only pumps it up under load - the same stuff. So I don't see why we have to change the definition of stock frequency because one technology was expanded. If so, we should re-define the stock speed for idle and because CPU-Z valid is made in idle, we should count it from the low-speed mode

 

The same logic.

[Massman]

If a CPU is rated to work at 3.4GHz using turbo mode ... then why should we treat is as if it were a 2.8GHz processor?

[Antinomy]

If it's rated to work at 1.6GHz using idle mode. Then why should we treat it as if it were a 2.8GHz processor? Were making a CPU-Z validation in idle, aren't we?

[Massman]

Because the manufacturer says this CPU is rated to work at 2.8GHz and warrants it will be doing that for ever.

[Antinomy]

If a CPU is rated to work at 3.4GHz using turbo mode ...

Because the manufacturer says this CPU is rated to work at 2.8GHz and warrants it will be doing that for ever.

M..?

[Massman]

Oh, you're still talking about the same example.

 

The manufacturer warrants that this CPU will be capable of running 3.4GHz, so why treat it as 2.8GHz?

[Antinomy]

The manufacturer claims 2.8GHz as stock frequency. Do we agree with the manufacturer:

The manufacturer warrants that this CPU will be capable of running 3.4GHz

The manufacturer claims 2.8GHz as stock frequency.

or not?

I can warranty that it's capable of running at 3,5GHz and I say it's stock is 2,93.

Not even taking stuff like overclocked samples of videocards in account

[Massman]

You can warrant anything, but the manufacturer makes the product.

 

The manufacturer says that stock settings will give you 3.4GHz.

[Antinomy]

Damn, I didn't want to do this

[Massman]

It's a pointless argument, because we're just both explaining how we look at it. There's no absolute correct way, it's just two opinions.

 

This can go on forever.

[Antinomy]

The processor supports ACPI P-States. A new feature is that the P0 ACPI state will be a

request for Intel Turbo Boost Technology. This technology opportunistically and

automatically allows the processor to run faster than its marked frequency if the

processor is operating below power, thermal, and current specifications.

 

That's what I didn't want to - to quote the datasheet

[knopflerbruce]

If we need turbo modes to be represented in the specs, we should add the other speeds, not replace the actual stock speed.

[Bobnova]

Personally, I think that HWBot should work on securing the benchmarks from cheating before they start chasing after the golden samples.

How many golden 2d CPUs are actually creative cheats? I bet there are a few, at least.

 

It seems to me that the proposed system makes it way more profitable to cheat, and no easier to prevent cheating at all.

[Antinomy]

Bobnova, you posted in the wrong thread

knopflerbruce, nice catch. Turbo can be represented on the specs page. For each core count load.

[Trouffman]

I'll definetly not agree to put he 'turbo mode' as the official frequency, some turbo, only run on a few core, not all of them at the same time.. then to get absurd why not putting a 2.8Ghz quad creo with only 2 core and 1 core with turbo @ 3.4 ? what is taht ? quad / dual / 2.8 / 3.4 ?

 

We should stay by the 'official stock frequencies' and if you want to display turbo mode, could be interesting to do so in the spec , or as an extra detail.

[Christian Ney]

Yep,

 

We can't say it is running at 3.4 Ghz if only one or two cores are really running at 3.4 and others at 2.8.

 

So Turbo mode should be banned ( only for CPU-Z submissions ) :D:D

Edited August 28, 2010 by Christian Ney

[knopflerbruce]

Turbo is a feature, no reason to ban it

[Antinomy]

Yep

it's like banning HT

[Massman]

How can we ban turbo mode ... this is an overclocking website

[Christian Ney]

, only for CPU-Z submissions, or an another cathegory should be added.

( like 1x, 2x, 3x, CPU for WPrime, add With and Without Turbo for CPU-Z )

[Massman]

You do realize that turbo mode is 'just' an extra multiplier, right?