FX-8320 vs i5/i7's on DX12?

[Baionoido]

-How do you guys think the FX-8320(50) will perform when DX12 (multicore) arrives against i5/i7's?

-Do you think it will be bottleneck 290/970's with DX12 (high-end cards)

[SeanBond]

On dx12 I think an 8320 should be fine. At least for a little while but if you are looking for a new pc just wait for the next set of i5's to come out because even compared to an i5 now full utilization on all cores and 8320 is barely faster.

[lani100]

it won't get a huge performance increase, and by the time dx12 is the norm the amd fx 8xxx are outdated already.

[Tim Drake]

It will help with AMD CPUs in games (in theory) but that doesn't stop them from sucking.

 

Intel WILL still be better.

[Samsterstorm]

The 4770k and 4790k are all 8 cores in a way (4 cores 8 threads) so i reckon the 4790k will kick ass on DX12!

[SuperPug]

Just because DX12 supports more cores doesn't mean that the devs are gonna use them.

[D2ultima]

The 4770k and 4790k are all 8 cores in a way (4 cores 8 threads) so i reckon the 4790k will kick ass on DX12!

Nah. Believe me when I say hyperthreading is nothing like an extra core. It's about 30% of an extra core, if a program uses it right.

 

Just because DX12 supports more cores doesn't mean that the devs are gonna use them.

It's not exactly "supporting" more cores... it's more that it lets the programs talk more directly to the GPUs without the CPU pre-rendering soft frames. This means the GPUs are not bottlenecked by the CPU (though the CPU will still perform tasks like AI calculations or CPU-based Physics such as what the Havok engine uses; so it will still be used) but also that the GPUs work a lot harder. It means they can also be used in different ways too; such as combining vRAM instead of mirroring it in multi-GPU configurations. We'd lose the memory bandwidth benefit from multiGPU in those instances, but then they might be able to cook something else up... like basically doing Raid 0 with vRAM. Just speculation there though.

[SuperPug]

It's not exactly "supporting" more cores... it's more that it lets the programs talk more directly to the GPUs without the CPU pre-rendering soft frames. This means the GPUs are not bottlenecked by the CPU (though the CPU will still perform tasks like AI calculations or CPU-based Physics such as what the Havok engine uses; so it will still be used) but also that the GPUs work a lot harder. It means they can also be used in different ways too; such as combining vRAM instead of mirroring it in multi-GPU configurations. We'd lose the memory bandwidth benefit from multiGPU in those instances, but then they might be able to cook something else up... like basically doing Raid 0 with vRAM. Just speculation there though.

 

Ah.. didn't know that i just assumed it had something to do with multicore optimizaion because of the FX mention.

[D2ultima]

Ah.. didn't know that i just assumed it had something to do with multicore optimizaion because of the FX mention.

Nah. It removing the load off the CPU for the GPUs simply helps those weaker CPUs shine. It's why Mantle is less beneficial the stronger the CPU you have. Ever notice those "60% gains!" they advertised were using APUs with midrange GPUs? Mantle on a 4.8GHz FX-8350 + R9 290 isn't worth the bit of stutter/colour change for my friend, even (BF4).

 

It'll help new games though. CPU requirements will fall immensely, and streaming/CPU-based recording would become IMMENSELY easier.

[dandrix]

DX12 is far away, but FX 8320 is ok if not decent for what it is.

[chopdok]

Considering that i5 will benefit greatly from DX 12, and also the fact that even after DX 12 comes out, will be a while until real games, and not benchmarks, will start utilizing it - i5 is still a superior choice to FX 8xxx series for gaming.

[mathijs727]

Nah. Believe me when I say hyperthreading is nothing like an extra core. It's about 30% of an extra core, if a program uses it right.

 

It's not exactly "supporting" more cores... it's more that it lets the programs talk more directly to the GPUs without the CPU pre-rendering soft frames. This means the GPUs are not bottlenecked by the CPU (though the CPU will still perform tasks like AI calculations or CPU-based Physics such as what the Havok engine uses; so it will still be used) but also that the GPUs work a lot harder. It means they can also be used in different ways too; such as combining vRAM instead of mirroring it in multi-GPU configurations. We'd lose the memory bandwidth benefit from multiGPU in those instances, but then they might be able to cook something else up... like basically doing Raid 0 with vRAM. Just speculation there though.

OFFTOPIC:

I dont get why people thing DX12 is gonna magically make the PCI-e bus so much faster that we dont need mirrored VRAM. The only thing they said that developers can get control over both the vRAM seperatelly, not that it will make the PCI-e bus fast enough to use it to transfer gigabytes of data for every frame...

 

ONTOPIC:

At the time DX12 will be used in games (which will take a while) the FX8350 wont fall that much farther behind Intel as they're focussing on power efficiëncy instead of performance. The question is if real games will use those 8 cores because not all games will be that easy to parallelize as the starswarm demo's that they're showing now (which are pretty much the most optimal case).

[vembutech]

As said the dx12 should be good with 8320 and can go with that hopefully it will work fine.