Official OcUK RX480 4GB and 8GB review thread

[humbug]

Usual suspects...

Anyone who doesn't follow the line some in here want are AMD bias.

 

[Angilion]

So what would be better. Undervolting to reduce thermal throttling or turning the power usage up to just give it sheer oomph. Maybe increase power limit and fan speed up? Would it still throttle?

I'd love to see some results using a watercooler at stock with the power upped. Seems like it's suffering in a similar manner to the Nano, which should be more or less Furyx performance.

Thermal throttling doesn't seem to be the first bottleneck, so I think that reducing temps won't do much by itself.

Turning the power usage limit up wouldn't help much if at all on a reference card either because it already exceeds the maximum specified power draw from the sources allowed to it. 75W from PCI-E slot + 75W from 6-pin connector = 150W and the card is pulling 160W at stock power usage limit. You really aren't going to want it pulling any more through the PCI-E slot in particular (6-pin PCI-E cables are usually actually rated a long way over spec, so that's not a major problem).

Undervolting would reduce power consumption and thus reduce power throttling, so that should help even on a reference card at stock power draw limits, but I think to really benefit you'd need a 3rd party card that draws less from the PCI-E slot and more directly from the PSU (either with an 8-pin connector or 2 6-pin ones). That would allow the card to draw up to 225W and still remain in spec, so you could then increase the maximum allowed power draw and thus reduce power throttling by a greater amount. My guess is that the RX480 is actually a 200W card, not a 150W card.

If I had such a card, my first course of action would be to undervolt it as far as possible and increase the maximum power draw to +20% while keeping speeds at stock. I think that would probably give the best performance increase and reduce temps (due to the reduced voltages). Then I'd do some testing with increasing clocks a bit at a time and voltages by just enough for stability at the higher clocks, while monitoring how much power throttling occurs (and monitoring temps, of course).

To give a real world example that should be relevant to some extent, I increased the performance of my 7950 by 36% at stock clock speeds by reducing power throttling (by undervolting and increasing maximum power draw). In addition to the huge performance increase, the temperature delta on the GPU decreased by 19C!

 

[D13]

Thank you for doing this Cat it really helps

 

[Orangey]

Why do people say "bias" instead of "biased" so much lately.

I blame American TV/movies.

 

[CAT-THE-FIFTH]

Thank you for doing this Cat it really helps

No problems - just added another review as I missed it!!

 

[Freddie1980]

Tad harsh imo. They had some difficult choices to make ~3/4 years ago which we have been seeing play out with tonga/fiji/polaris. I am withholding judgment until we can see the threads of the last few years of effort coming together for 2017. (R&D down the product development pipeline 'from scratch' so to speak).

IMO R&D is a big problem for a company like AMD especially when it comes to graphics development. If we take a look at what's happening in the tech sector as a whole companies are ploughing development into graphics the likes of Samsung, Apple, Qualcomm, Intel have all significantly stepped up spending on graphics over the last few years. Since there's bigger increase in demand for graphics talent the most talented are going to get head-hunted and switch jobs.

It would help explain why AMD’s cards aren’t anywhere near as efficient as the Nvidia equivalents (ever since the 7970) and the lack of a high end Polaris card.