Ooops, I should never post here in the middle of the night...
I meant you get 50% more computation units (obviously) for 150%
cost.
On Jun 21, 10:24*am, AGross <farra...@gmail.com> wrote:
> Hello Daniel,
>
> Well, you get 50% more cores in the same package and the same TDP.
>
> Other than that, the price might be an anomaly caused by the resellers
> in your country.
> For example, in Newegg, the 6128 is 300 USD and the 6168 is 759 USD.
>
> So you get something like 150% more (independent) computation units
> for 250% price. You can look at it like that, just at the CPU cost,
> but, to make better sense, you can look at it from the POV of a blade/
> server integrator, where you take into account the motherboard, the
> space and heat exhausted, etc. If you take all those into account, the
> variation of delta-cores/delta-cost isn't that bad as it might look
> initially.
>
> If you look at it as a desktop 'consumer' then, yes, the price is
> skewed, but it is not meant for desktop users, but for blade
> integrators where other costs factor in that you do not see just by
> looking at the CPU alone.
>
> A really simple example would be 2x G34 sk each with a 12 core Opteron
> = 24 cores @ ~ 250W TDP (maximum, the real draw is far lower)
> However, if you go with the 8 cores, you can put at mosty 2x 8 = 16
> cores @ ~ 250W TDP and, if you want to jump to 24 you have to buy a 1S
> pr 2S system with one more 8 core Opteron that draws ~ 115W more (plus
> the motherboard/chipsed, drives, memory, psu efficiency etc).
>
> You get far more in the same 'space' with the 12 core Opty than with
> the 8 x one, not only as core count, but power usage, ease of
> integration and management (less blades to manage) and so on.
>
> On Jun 14, 2:50 pm, Daniel Meszaros <s...@meszi.de> wrote:
>
> > Hi,
>
> > I am astonished about the price differences of current G34 Opterons:
>
> > - 6128, *8x 2000 MHz, *8x 512 kB L2, 2x 6144 kB L3: 248,77 ¤
> > - 6168, 12x 1900 MHz, 12x 512 kB L2, 2x 6144 kB L3: 681,42 ¤
>
> > This is more than 25 ¤ difference per core.
>
> > Of course, put on a dual mainboard one makes "just" 16 cores, the other
> > makes 24 cores possible. But are there other reasons?
>
> > Does anyone of you know performance benchmark results? I could not findany.
>
> > CU,
> > Mészi.