PCTuning fórum

flanker píše:vždy se osazují vzdálenější sloty u ASUS desek.....

Tralalák píše:Zdar .flanker

Kľúdne si to skomparuj aj s tvojou recenziou AMD Athlon 5150 1.6GHz (Jaguar, Kabini):

Cinebench 11.5
• Piledriver: 1,33 pts ( 79,17 % - - - 100 % )
• Steamroller: 1,54 pts ( 91,67 % - - - 115,79 % )
• Jaguar: 1,64 pts ( 97,62 % - - - 123,31 % )
• Excacator: 1,68 pts ( 100 % - - - 126,32 % )

Carrizo is definitely not designed for heavy gaming or professional applications such as modeling / rendering or video encoding.

It would be the ultimate chip to throw in a ITX sized HTPC, however so far it seems that it´s not going to happen until Bristol Ridge.

Trying to squash a 15h architecture based CPU into a power envelope of 15W is IMO a ridiculous idea to begin with, especially when you remember that the same power envelope is shared by five different entities. Each three of the total five entities can alone consume the whole power envelope, if allowed to.

The thing is that at 15W power limit the performance is just not there. None of the domains can operate at their full clocks alone, let alone all together simultaneously. At higher power limits the performance picks-up nicely, however especially in this segment even small amounts of heat and power draw (on desktop standards) mean annoyingly loud noise, high temperatures and naturally very poor battery life.

There is no magic in Carrizo: The design itself isn´t more power efficient at hardware level than the previous generations were. Carrizo simply has much more advanced, effective and most importantly finally functional power management. In fact at hardware level Carrizo is less power efficient than the previous generations due the changes in the design and the manufacturing process. At higher frequencies (>3200MHz) Kaveri / Godavari can easily outperform Carrizo in power efficiency.

At >3200MHz frequency range the frequency scaling probably isn´t perfectly linear on either of these 28nm process variants, so I need to do a test on both of them at lower speeds. However when we look at the changes AMD made in Carrizo, it is clear that they knew it would never reach high frequencies. Otherwise they wouldn´t have swapped to cheaper manufacturing process (28nm custom to 28nm HPP), implemented HDL for the CPU cores or reduced the latency on the L2 cache, which already was the ultimate Fmax limit to begin with.

The 2MB cache block on 15h designs is responsible for around 10% of the total power consumption of the compute unit. In Carrizo it is half the size compared to Kaveri / Godavari, but features lower latency. Given that they have implented HDL and various other gimmicks in Carrizo, I would expect it to murder Kaveri / Godavari in power efficiency. However since it clearly does not, it must be because at >3200MHz the design is already operating at several hundreds of MHz higher than it´s optimal Fmax range.

I think I´ll clock both Godavari and Carrizo to 2.1GHz and check the power figures once again.
That way Carrizo should still operate within the "optimal" Fmax range of the design / process.

To be honest, I don´t think Bristol Ridge will able to match Kaveri / Godavari in CPU performance.
It probably offers a slight boost to the graphics performance (due FBDCC in GCN1.2), but not much more.

velikost L1 cache, write trough cache, 8MB L3 cache

(až na tu (ne)sdílenou FPU a 8 SMT či dual Issue decode. Samo i L2 je menší)

Je poměrně zajímavé, že pouhý dual issue decoder dokáže krmit jádro, které podporuje až osmicestný SMT.....

The question is whether or not Bristol Ridge will also release on the DDR4 supporting AM4, to get that infrastructure warmed up in time for Zen, and also for the DDR4 to unleash a bit more of the GPU's potential.

This BIOS release is a good confirmation that Bristol Ridge will also be available for legacy DDR3 platforms, which is nice for upgraders, and OEMs and motherboard manufacturers with excess stock to shift.