Rumor Has It Vega Mines Ether Like A Beast, But It Could Be Fud

Status
Not open for further replies.
Holy sweet Jesus... since Etherum is linked to video memory, this could be true with HBM 2.

AMD even claim that HBM 2 is having huge impact on minimum FPS. Well, that's a good selling point if this is true.
 

bit_user

Polypheme
Ambassador

They're really vague about exactly what they sped up:
Our latest driver — available today — delivers 3x more performance in applications like Maya to help you create and design faster than ever.

So, I'm not sure if they flipped the switch to enable the same OpenGL rendering performance on it as their professional cards (which would make sense, if you see how well Vega Frontier does against their $5000 Quadro P6000) or what.
 
This article is both plausible and angering. Because what Hannibal said may very well be true, & because it merely reminds one the GPU industry has more or less become FUBAR and full of FUD, thanks to CC mining, hyping, & whatever else I'm neglecting to mention. However, we won't know until we know, so... Until its release, really no other choice but to sit tight & hope for the best.
 
Isn't HBM2 is supposed to reduce memory latency, which is the main issue for the gtx 1080 series cards when it comes to cryptomining? And in terms of raw horsepower, much like the 480/580 Vega has a significant edge over the 1080 series cards, even the Titan XP. The problem has always been efficiently utilizing it in actual games, whereas cryptomining more closely resembles a synthetic benchmark. If that holds true, you'd expect it to perform at about twice the rate of a 1070 (~13.7 Tflops vs 6.5) and then however much the RAM helps, assuming it helps at all.
 

mapesdhs

Distinguished
"The idea that a driver could double the performance of a GPU isn’t unprecedented--Nvidia just released a driver this week that tripled the compute performance of Titan Xp cards."

How does this not warrant an entire investigation in its own right? If a car company did some tweak to an engine management system which tripled the fuel economy, there'd be a giant WTF, have they been crippling the tech in the past? Why is in the computer tech industry an article can say something like that and the response is, meh, whatever, expect that sort of thing these days.

Come on tech journalists, dig into this stuff! What's going on here? Is the tech being deliberately held back? Is it legal? (baring in mind btw that such a change means the original stated performance claims must be misleading)
 

hannibal

Distinguished
You can make a program that is really slow and another that is much faster. In Computer technology there is always room for better effiency. It is just matter of skill and time.
Another thing is when certain features Are locked and you can pay more to get new features to your equipment. For example Intel cpus where you can pay xxx$ and get raid feature etc.
That is harder to judge. Nowadays you can go to car shop. Pay xxx dollars and They blug a Computer to you car and vola you have 50 horse powers in your engine. And that is completely legal.
 


Nobody needs to dig into it, Nvidia were crippling the cards. The professional Quadro cards had the drivers and the performance, the Titan XP didn't. Nvidia also has the drivers on lock down for anything below the Titan XP say a GTX 1080 for example. This is all known and doesn't warrant any investigation.
 

bigdragon

Distinguished
Oct 19, 2011
1,111
553
20,160
Vega is now dead for gamers. You won't be able to buy one especially at MSRP. Get ready for these cards to constantly be out of stock and marked up 50 to 200% in price. Miners are going to buy Vega like crazy.
 


To be fair, it's probably worth at least looking into whether the "3x performance" claims are true, and not just writing that they are in an article and linking to Nvidia's press release as a source. And more specifically, Nvidia only claimed 3X performance "in applications like Maya" so those numbers might only apply in very specific workloads in a limited selection of software going by what they said, so it might be inaccurate to claim that they "tripled the compute performance" in general. I'd at least like to see this article link to a source that tested these performance improvements.
 




Yes because it makes perfect sense to purposely hold back a GPU when competition is nipping at one's heels. Well let's just get an international government consortium involved and fine Nvidia into bankruptcy because they released GPUs to the world that had governors on them until owners paid a driver ransom. That'll learn 'em!

Oh wait, what's that? Nobody is being forced to pay for driver improvements? Absurdity. Everyone sees the architecture breakdown of every new GPU release and what a GPU is capable of in a review. If you like what you see, you buy it. If you do not like what you see, you do not buy it. It's as simple as that.

I wonder why these accusations only apply to Nvidia GPUs and never AMD GPUs when driver or OS updates mysteriously show gains? I submit to you Fury X and a driver update showing gains over the 980Ti in certain apps when previously it was behind that GPU it targeted:

https://www.eteknix.com/windows-10-drivers-gives-amd-r9-fury-x-advantage-over-nvidia-980-ti/

The last time that I'm aware of a GPU maker ever "lying" about its GPU capabilities was Nvidia and the GTX 970. Nobody can prove if that was intentional or a massive breakdown in communication between the marketing department that puts the words out there or engineering that built the GPU. What we do know is that they paid a heavy fine for that mistake. Unless the outside of the box says things like "Guaranteed to hit 60FPS in Dues Ex at 4K" and it doesn't, what difference does it make what performance improvement driver updates mean to the GPU?
 

bit_user

Polypheme
Ambassador

It's long been known that both companies crippled OpenGL performance of the drivers for their non-workstation GPUs.

AMD's decision to offer their professional GPU driver for their Vega FE cards was the game-changer, here. That allowed a $1000 AMD card to compete with Nvidia's $5000 Quadro P6000. So, that forced Nvidia's hand in unlocking the OpenGL performance of their Titan Xp driver.

The winner is anyone needing to run these professional applications. They can now use a GPU costing 1/4th or 1/5th of what they previously had to pay, and they now still have a choice of AMD vs. Nvidia.

 
Come on tech journalists, dig into this stuff! What's going on here? Is the tech being deliberately held back? Is it legal? (baring in mind btw that such a change means the original stated performance claims must be misleading)

yes it is. they are called as product segmentation. and both AMD and nvidia did this. but in this case it is not being held back. the capabilities have always been there. only the specific optimization for specific application not included in initial driver. just that this time AMD decided to provide pro drivers on their semi pro card. nvidia is merely responded to it by giving similar treatment to their semi pro card. (though the target market for AMD Vega FE and Nvidia Titan is quite different to begin with).
 

Josh_killaknott27

Prominent
Mar 22, 2017
18
0
510
This is going to screw gamers over....bottom line. Better bottom line for AMD but im pretty sure these things will exist in the wild like a white Rhino. Which is really sad.
 

carmzatha

Commendable
Jul 25, 2016
1
0
1,510
Flood the market with Vegas 64 and 56 plus the AIB and have stores keep record of no more than 2 cards per month with one Credit card only. Let the Miners borrow credit cards from friends,mom's dads,and sisters and brothers. Sometimes it is hard to get the card's to work without the owner of such cards plus you cannot stop it but slow it down,
 

bit_user

Polypheme
Ambassador

Supply will eventually catch up (and possibly even surpass) demand.
 

bit_user

Polypheme
Ambassador
Scott Herkelman, AMD's gaming business unit VP & GM had this to say more specifically about mining optimizations during a presentation last weekend:
These guys who do this professionally, let's just say they are hackers more than you'd think . . . Since ours [GPUs] are based on OpenCL, they can go in there and manipulate and do things that are outside of our bounds. . . . That can really increase the hash rate. There are public numbers [hash rate]. . . And there are private hash rate numbers that we see and just kind of go whoa. . . It's kind of crazy what you would see from crypto-currency guys that do this for a living. They hire tons of software folks . . . And they hack into the firmware.
He makes it sound more esoteric than it even is.

AMD opensourced most of their driver & software stack. Their Pro driver contains a few proprietary bits, but largely they're moving to an opensource model, where Clang is used as the OpenCL compiler. This means anyone can tune: the driver, the compiler, and the mining software specifically for one hardware platform.

You still have to know what you're doing, but gone are the days of actually having to reverse engineer anything. This is not hardware hacking, like in days past, so much as an exercise in optimizing and tuning various parts of the software stack for a specific workload.
 
Status
Not open for further replies.