• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%
  • (on) Overcrowding, standing room only

    Votes: 100 18.9%
  • (never ever got on) Chinese escalator

    Votes: 221 41.9%

  • Total voters
    528
Status
Not open for further replies.
No, it will "possibly" run faster. Seems like a bad idea to gamble on that happening.
Has happened so far with every single AMD GPU. In 6-12 months performance increased. Sometimes 20% and more.
Will it be enough to catch 1080Ti? Maybe not.

Now power consumption is not going to get better. And neither are clocks. HBM2 is another overclockers dream and 1600MHz seems to be the ceiling for GPU.

RX Vega better be priced really well. Too many factors are against it.
 
I definitely think there is some performance missing,I just have a hard time believing its drivers.



However, here is an anecdote. Intel designed a new processor called the Itanium. From a theoretical point of view the architecture is massive leap beyond the ancient X86 ISA we are stuck with. The problem is it is incredibly difficult to get the actual theoretical performance out of it. The software is too complex to make use of the hardware. So maybe AMD tried some big changes without thinking about the software side.
Could be. It all sounds to me like they are struggling with drivers. They have had a long time however. Let us hope they have it in shape for August, otherwise judging already by the way people are responding here now, they will be the laughing stock here.

My stance is, I want to see AMD succeed so we all benefit, so I am happy to buy an RX Vega even if it is around ~1080 performance (even though I have had a 1080 and sold it) as long as they do not take the **** with the price and keep it under £400. Bear in mind I do have Freesync monitor also, so that is another reason.

Plus it seems likely driver performance would improve by quite a bit over time which is good as I get free performance that way as the card would likely end up in my second rig eventually once Navi hits.

Saying all that, if AMD fail to improve on price for performance, I will just buy Volta instead. I really liked my 1070, so no problem getting the 2000 series Volta version of it if AMD do not deliver. But I would rather AMD deliver now.
 
I have an 850W supply too, so will easily be enough.

Does make you wonder about a larger Polaris though.

With some minor tweaks, they could have probably fit 5000 Polaris cores in 500mm2, and running at 1350-1400 Mhz with 11 GHz GDDR5X that would likely be a very decent GPU.


Not if it runs into the same bottlenecks, adding more and more cores doesn't scale performance if there are other limiting factors, which is why the FuryX never lived up to it theoretical potential.

Then there is the power consumption. The higher clocked rX580s are already pulling 250w.
 
Not if it runs into the same bottlenecks, adding more and more cores doesn't scale performance if there are other limiting factors, which is why the FuryX never lived up to it theoretical potential.

Then there is the power consumption. The higher clocked rX580s are already pulling 250w.
does this mean there is a possibility that when the rx vega is released a 550w psu may not be enough to power it?? or is this all just hearsay as such??
 
Not if it runs into the same bottlenecks, adding more and more cores doesn't scale performance if there are other limiting factors, which is why the FuryX never lived up to it theoretical potential.

Then there is the power consumption. The higher clocked rX580s are already pulling 250w.

Yeah very true.

There does seem to be a fundamental problem with GCN in scaling above ~3000 cores and ~1300 MHz, regardless of the transistor node.

At the risk of immediately talking about "the next thing round the corner", I hope if Vega does turn out to be a letdown they've really actually done a ground-up design with Navi and take full advantage from the 7nm node. From all official data, it really looks like AMD are being handed a gift with the 7LP IBM/GloFo process, looking to be the best of the 7nm generation. So I really hope they take full advantage of it.

At the same time though, I find it so hard to believe these FE are indicative of the RX gaming performance. The numbers are too far off the expectation of even some pessimistic napkin-assumptions, so seems to me like the drivers are very unoptimised.
 
Yeah HBM2 looks like another flop, unless there is some magic AMD can use with their cache controller and/or infinity fabric.

Only seems worth it with 4 stacks, for up to 1TB/s and 32GB, but then it's very expensive. 2 stacks is cheaper but then slower than GDDR5X at 384-bit bus.

HBM3 looks promising though. Goes up to 128GB (I think, might be 64GB) and 2 TB/s with 4 stacks, or even 512 GB/s for just 1 stack. GDDR6 at 16 GHz on a 512-bit bus will still only be 1 TB/s, and I don't think there's any roadmap for 32-40 GHz GDDR7 anytime soon?

GDDR would need to be 42.7 GHz at 384-bit (the reasonable cost bit-width) to match 2 TB/s, or 21.5 GHz to match 1 TB/s.

So when we start needing 1TB/s and above, MAYBE HBM will finally make sense :p
Interesting, thanks!
 
Interesting, thanks!

I double-checked and apparently it's 64GB at first but they're trying for higher stacks. So probably not 128GB max, as that'd need double the heigher, but maybe 80GB is plausible once its mature. And the 512GB/s per stack is correct.

This is due for late 2019, according to Samsung's roadmap.

And Micron are planning to get to 16 GHz with GDDR6 by early 2019, and have no released plans for speeds faster than that and/or the successor to GDDR6 yet (obviously there'll be something, but no idea when).

So yeah, HBM3 could be the one in the end.
 
AMD have stated the gaming version will be released at Siggraph which runs from 30th Jul - 3rd August. So those are your RX Vega gaming GPU launch dates. Then we can judge it's proper gaming capability
That's if they release it straight away. Remember the 480 was released a month after launch with the custom cards another month after.

It wouldn't surprise me with the troubling FX launch that the RX release was delayed.
 
Last edited:
would that be the part that says 45.5A 546W??? sorry for n00b questions but this is the stuff where i get totally confused about

Yeah exactly that, basically if you have a single pcie6/8 pin socket from your psu it'll have a current capability in the specifications.

In multiple rail it will look like 12v1+ 15 A, 12v2+ 15A
 
Status
Not open for further replies.
Back
Top Bottom