RX VEGA AIB Cards Information Needed?

[Spyhop]

And Nvidia are putting out decent products (unlike apple)

If we look at strictly performance then the Apple SoCs murder the competition in both peak performance and perf/watt.

https://browser.geekbench.com/ios-benchmarks
https://browser.geekbench.com/android-benchmarks

Apple as a whole certainly does a lot of annoying stuff, but their in-house chip design is as dominant as NVIDIA in their respective market, perhaps even more so.

 

[JediFragger]

it would be price as much if not more.

Yup, AMD no longer in the value game.

 

[nashathedog]

You never know, as AMD aren't producing then anymore now, so they may stay as priced, as they want em gone.

I wonder how many of the big bundles that included Ryzen and monitors actually sold? From what I gathered the only people really waiting on Vega were doing so because they had Freesync monitors already.

Remember RTG are in flux atm, and AMD are taking their sweet time appointing someone to replace Raja!! Not good for the medium-term, not good at all (things will slip no doubt, and nVidia are solid as a rock leadership wise)

I imagine Mrs Su is more than happy to fill in for now, She messed up with Raj so there's no rush and hopefully with her at the reigns things will get better. It'll be interesting to see what replaces Polaris. Will it also be Navi or is Navi meant to be what Polaris is? Maybe they'll just do Navi on both Gddr and HBM, I can't imagine they'll want to risk another Vega launch.

 

[LoadsaMoney]

Vega 11 is supposedly going to replace Polaris, and its supposedly using GDDR5 as well.

 

[Freddie1980]

Vega 11 is supposedly going to replace Polaris, and its supposedly using GDDR5 as well.

I wonder if it will have die space for HBCC.

 

[ubersonic]

Yup, AMD no longer in the value game.

I wouldn't be so fast to call that TBH, the Ryzen CPUs noticeably undercut Intel's equivalent so they are still focused on bang for buck. I think the reason Vega only price matched Nvidia's equivalents is because it was expensive to make (which is another issue entirely).

From what I gathered the only people really waiting on Vega were doing so because they had Freesync monitors already.

You'd be surprised, I.E the workstation I use my twin Vegas in doesn't even have an adaptive sync monitor, and I know a couple of people who are waiting for the AIB Vegas before ordering new monitors. Then of course there's the mining farms lol, they can't get enough of Vega.

 

[RavenXXX2]

 

[southernorth]

Incredible that they've missed that, especially for a reviewer sample which many speculate are 'chosen with extra care'.

Shown later on in the video to be fixed, but still.

 

[Hotwired]

Nah it's just a reviewer sample, many speculate they are rush jobs just for a quick review.

 

[RavenXXX2]

Nah it's just a reviewer sample, many speculate they are rush jobs just for a quick review.

Apparently their solution is to spade on the paste to make up the gap.

 

[shankly1985]

So now I guess we can put to bed AIB partners didn't want to touch Vega! Its take a long time though.

 

[ubersonic]

Apparently their solution is to spade on the paste to make up the gap.

Well that is exactly what the paste is for.

I mean we're talking about a 0.04mm gap, you get bigger than that from one side of a CPU IHS to the other.

 

[Steampunk]

Well that is exactly what the paste is for.

I mean we're talking about a 0.04mm gap, you get bigger than that from one side of a CPU IHS to the other.

So it's kind of amazing that on their performance air-cooled model, Powercolor managed to put a layer of paste that didn't connect with the HBM, but did connect with the GPU. You can see the paste is there on the heatsink, but it doesn't mate with the HBM modules.

 

[ubersonic]

So it's kind of amazing that on their performance air-cooled model, Powercolor managed to put a layer of paste that didn't connect with the HBM, but did connect with the GPU. You can see the paste is there on the heatsink, but it doesn't mate with the HMB modules.

My guess is they used the normal amount instead of adapting their process for the Vega layout, it's the kinda thing you expect for a prototype however you would have thought somebody would actually test the card prior to shipping to a reviewer and noticed higher than expected HBM temps lol. Then again ASUS (I think) shipped out Strix prototypes with the wrong BIOS settings earlier this year so it's not like Powercolor are alone in the mess up department

Manufacturers are usually quite erratic with the TIM, sometimes they are so stingy it doesn't completely cover the core, sometimes they use so much you could butter a cracker with the excess

 

[hyperseven]

Well that is exactly what the paste is for.

I mean we're talking about a 0.04mm gap, you get bigger than that from one side of a CPU IHS to the other.

excuses, excuses

Why didn't they just machine a recess into the heatsink/add a shim so they both will be flush? Are shims still a thing?

Powercolor messed up, now there's some poor minimum wage guy unscrewing all the heatsinks to put an extra splodge of tim on, poor guy

 

[Davedree]

excuses, excuses

Why didn't they just machine a recess into the heatsink/add a shim so they both will be flush? Are shims still a thing?

I agree they just need to cut a recess into the gpu die mating area so the hbm2 then makes contact. The problem is there are 2 different versions of gpu package (height) so they'd have to manufacture two different versions of heatsink face types. That's slot of money and ballache on a low turnover product.

 

[ubersonic]

Why didn't they just machine a recess into the heatsink/add a shim so they both will be flush? Are shims still a thing?

Well I can't speak for Powercolor but I would assume that they didn't do the former because it would have added to production costs (and thus retail costs) without actually benefiting performance.

Also that they didn't do the latter because in order to shim a 0.04mm gap with space for TIM underneath the shim would need to be so thin that it would cost more than the graphics cards value to produce (you'd be talking about a piece of copper less than 1/20th the thickness of a human hair).

 

[KillBoY_UK]

excuses, excuses

Why didn't they just machine a recess into the heatsink/add a shim so they both will be flush? Are shims still a thing?

Extra cost to manage the GPU's (as there seems to be more than one type that could affect heatsinks

Extra cost to design make and install a shim on the GPU heat sink, extra chance os something going wrong (what of the shim is to thick or thin, the shim could clip something onthe gpu package etc)

Extra cost mill the heatsink in terms of tooling and and time

Basically it seems it seems they would need to whole sale change there production lines or run two depending on the batch of GPU's they where sent from AMD, or just run one and risk stuff like this.

It be like if a Car factory was spec-ted up to fit a particular engines dimensions into there car's production line only they are never sure what dimension would be on a batch by batch basis.

for such a high cost, low margins low sales SKU I doubt its worth them running two production lines

 

[hyperseven]

Also that they didn't do the latter because in order to shim a 0.04mm gap with space for TIM underneath the shim would need to be so thin that it would cost more than the graphics cards value to produce (you'd be talking about a piece of copper less than 1/20th the thickness of a human hair).

no it wouldn't you can buy the stuff dirt cheap from alibaba, should I send Powercolor a quick email for this top tip ?

 

[Razor Time]

no it wouldn't you can buy the stuff dirt cheap from alibaba, should I send Powercolor a quick email for this top tip ?

Damn look at that power draw, its more than 2 x 1080s in SLI.

https://overclock3d.net/reviews/gpu_displays/powercolor_vega_64_red_devil_review/14