Return to Level1Techs.com

The VEGA 56 / 64 Cards Thread! General Discussion

amd
vega

#461

If you buy in bulk microcenter is charging $800 per card


#462

And now that the numbers are public. I can say this for those who don’t know.

There’s an architectural limitation in how ethereum or most sha-256/ethash mining in general works that means that it’s not possible to optimize for FP16 ‘rapid packed math’ on Vega. And I’d be very surprised if someone managed to make that work at all.

From what I know about the algorithm, Coin mining is strictly FP32. So not really anything to ‘unlock’ there. You can read the full Vega ISA here to understand the new instructions and why the architecture wont really bring a magic bullet to mining http://developer.amd.com/wordpress/media/2013/12/Vega_Shader_ISA_28July2017.pdf

Keep CS Degree handy to understand that document.


#463

Really, really depends on the PoW algo. Cryptonight, for example, has vega blowing past Titans, and equihash is posting very competitive results. Dagger-Hashimoto (Ethereum’s PoW) is actually memory bandwidth-limited, not FP32. That’s not to say a lopsided setup will magically get better, but we’ve typically seen any small improvement in memory conveying a relatively proportional advantage on almost every card.

(The 1080 and other GDDR5X cards are outliers because none of the cuda mining authors and optimizers can figure out how the hell G5X works, this isn’t the case with HBM)

Also note that geth --mine is not what the majority of folks are hashing with, and isn’t representative of performance for optimized setups. Big miners don’t use publicly available software for the most part.

What’s also puzzling is the quoted performance by AiB partners on their internal tests. I don’t have direct lines to them, but Steve Burke and others have independently validated the high hash rates. Some of my contacts in the mining business have intimated that vega 56 should be about 40-55 MH/s on a large scale rig according to their FE hacking and testing, and 64 should be around 55-65 without going into it in more detail.

My guess would be that the differences mainly come down to how far they can push the HBM on each chip.

tl;dr is that we’re gonna need more time to see how it shakes out.


#464

Well, I’ll just borrow the forum grill and grab some steak from the Supermarket.


#465

The undervolting results from GN are particularly bizarre, especially if more generally applicable, and not wholly due to “luck” of the draw.


#466

I mean, we saw similar results with fiji and polaris. Newer AMD cards like lower voltage.


#467

Yeah sure, but then why not do that before shipping? Would make for much better review results, etc.,


#468

Because vega is near the reticle limit, so they have to set the spec very conservatively to ship every asic they can produce. If they can bin 10% more chips at the higher voltage, that’s a huge wafer savings.


#469

In that case, they would really do well to come up with an OS-based testing app that can test for stability at the lowest likely voltage & stock clocks, and then set that as the new normal. Run overnight or whatever, and done. Or would that be less non-trivial than writing this down makes it seem? :stuck_out_tongue:


#470

Because they’re working with a bunch of different companies that will be binning and modifying things within their spec. Those partners don’t want the hassle of complying with a software solution on top of a hardware spec. For factory overclocked cards and cheap, badly binned cards, that kind of software could easily cause issues. Even if an ‘auto voltage’ adjuster doesn’t work with 1-2% of cards, the customer backlash will be disproportionate to that.

Also, because it’s trivial for the end user to crank it down 100mv and AMD doesn’t have to stake their PR or any liability on it. The value add is tiny and the risk is considerable.


#471

It took more than a year of news releases, architectural sneak peeks, unboxings, and reveals, but Radeon RX Vega is finally here. AMD is back in the high-end of the gaming market. It’s not a home run, but after getting hands on with Vega Frontier Edition, I don’t think we had that mindset anymore, and as a result I view the performance increases from it to RX Vega a strong move in the right direction. The question that remains is how many people they have just convinced to buy in.


#472


http://www.3dmark.com/compare/fs/13363150/fs/13361891


#473

Anyone see the multiple articles on Facebook and elsewhere talking of AMD setting the pricing and then quickly moving to a higher price after launch? Initially reported as retailers doing it but it was AMD’s doing.

https://www.overclock3d.net/news/gpu_displays/the_prices_of_amd_s_rx_vega_64_has_increased_by_almost_100_within_two_hours_of_hitting_retail/1


#474

So a vega 64 in Canada is going for 789 CAD (619 USD) and with tax it’s over $900! I can’t believe it, that price may just be out of my grasp :disappointed_relieved:


#475

Seen one popup on my local Craigs list. Still had the microcenter price on it of 699. (vega 64) and he is trying to sell it for 900. As predicted the price gouging on these cards is crazy.


#476

In Germany it looks to be 649,- for Vega64 and 699,- for limited air cooled. I am actually hoping for massive price spikes now because I ordered the normal and the limited when they were a little cheaper. (609,- for standard, 679,- limited) Could not order two normal ones at caseking…

The plan is to sell the normal one for … millions or so and end up with one card for a somewhat ok price. I wasn’t going to do this at first but I am about to move my gaming rig into a rack case and if there is any use for a blower fan it is this. Otherwise I would stay with my fury.

I have two weeks to send it back if I can’t find a buyer and it cost me nothing so…


I still prefer to give money to AMD instead of Nvidia but … yeah.



#477

I will wait until it reaches sane prices before ordering.

My Sapphire 390x 8gb will do me for a little while longer.


#478

I shall wait until there’s a 1x8pin vega56 of some sort that’s decently priced (so not a “nano”). Would’ve bought a ref 56 if I had a PSU with a double 8-pin conn, but so happens that I don’t, and I don’t really want to replace it.


#479

It’s back.


#480

Interesting, he took it down already. Watched that when it went up and commented that right now there’s no real word from AMD about the pricing change, just potential FUD from an undisclosed source. Hopefully he puts up another one with an explanation.