Untitled

techpowerup.com/242216/nvidias-new-gpp-program-reportedly-engages-in-monopolistic-practices

Attached: Buttmad Jensen.jpg (401x237, 15K)

Sup Forums doesn't care because it is a good little goy.

>Sup Forums is a Sup Forumsood little Sup Forumsoy

While it has lessened on the cpu side due to ryzen, Sup Forums and PCMR (same thing) want AMD to compete so they can buy Intel and Nvidia parts for cheaper. As such user actively encourages monopolies and practices associated with it because marketing is more important than any other factor.

I care it's why I always tried and bought amd

Vega was so bad I just couldn't

Fuck paying 1k+ for a high end GPU

I hope rtg does a ryzen and shakes up the GPU market but we all know we are gonna have to wait until 202x

INTEL 2.0

IT BEGINS

It was bound to happen Vega was confirmed dead as soon as it missed the original launch window

Vega certainly didn't deliver for gamers but it - like all previous GCN revisions - is a compute monster. We still can't claim AMD is innocent into getting into the position it is now, the GPU sideo f the business has some monumental hurdles to overcome that basically need them to redesign their hardware to common coding trends because as it stands people simply aren't bothering to code to GCN's advantages which is why GCN in general (and vega is particularly bad for it) suffer from acheiving throughput. A good one is draw calls - a common trend to to hit the gpu with one massive amount of drawcalls at once which works great for Nvidia but it fucking kills GCN which prefers a constant stream over lower quantity drawclls.

What if amd doesn't get rid of gcn or fix the front end?

I don't know Raja, care to tell us why?

So this mainly effects Asus, Msi, and Gigabyte. Those are 3 pretty big fucking brands. If this is true, we could see all Radeon cards vanish from their next line up. I don't see how Nvidia can keep this hidden. Most people I know with Radeon have Asus, Msi, or Sapphire
Have they realized that antitrust fines usually outweigh the dosh you make in the process of crippling your competitor?It worked for Intel, Oil companies, and others.

I suspect Vega was a in between step. Designed to remoce all legacy GCN and get a new base architecture design going forward. Then you do MCM GPU's (what NAVI will be) where they take smaller dies and stitch them together to make a larger GPU like what they did with Ryzen cpus

Infinity fabric is heavily dependent on the latency and bandwidth of memory hence the wild variation in performance on ryzen when you switch to much higher speed RAM with low CAS

That problem is nonexistent in a Mesh GPU design beause the memory especialy HBM is so much faster than DDR RAM and the bandwidth is so much higher with such low latecny that no such performance loss between GPU dies should occur.

The issue they will have is saturating the GPU. Like you said the large batches work better on nvidia and this is because of two things: Nvidia handle scheduling on the Software side and AMD uses a hardware scheduler. So nvidia can optimize better. Second the much higher raw clockspeed helps alleviate bottlenecks in the pipeline.

Solution to both issues is for AMD driver department to get more staff and a larger budget for game ready drivers and for partnering directly with game studies for optimization of games. Second solution is to modify the architecture to get the raw clockspeed higher even if you lose shaders in the process.

The new small die components for MCM gpu's should be pushing a higher clockspeed over number of shaders design then you just mesh together as many of the dies as you need to get the number of shaders for any given performance envelope.

Who was his replacement?

Never really kept up with it after he was fired
Amd fucked themselves though the 390 fury and 580 where all awful cards just like vega
I was hoping MCM was this gen but NVIDIA are stalling again with ampere

We need only to look at the past to know the future.

Nvidia will drag their heels like the money grubbing jews they have always been. AMD will be First to have MCM products on the market as a result with their NAVI release in either 2019 or 2020. These products will make them competitive again for a short time then Nvidia will finally release their MCM design which will blow AMD out of the water but at the cost of it being horribly overpriced.

People will still buy Nvidia anyway due to shitcoin miners buying up all the NAVI cards. AMD will give zero fucks because their product is selling and they are making money and since all their customers are miners and not gamers they will slash the budget for their drivers department and gaming performance on their cards will stagnate further. Then because they are so much further along in the benchmarks for games Nvidia will price jack even more.

Hawaii and polaris are the last balanced versions of GCN AMD made which is why they both performed (and continue to do so) exactly where they were meant to. Fiji has fallen off a cliff due to the framebuffer and vega is a massive die struggling to achieve parity with am id-size die.

390 and 580 are fucking great. What are you on about? I got my 480 at $199 and it kicks ass for 1080p gaming and 144hz Esports.
The memory / infinity fabric bit I suspected too. I think they wanted to leverage economy of scale to get HBM production rolling and hopefully get it cheaper for the McM gpu's. The thing is, the global memory shortage will fuck them in that regard.
I also think AMD was teetering on the edges of bankruptcy before zen came out, so they probably couldn't put much of a budget into the gaming software guys. I may be talking out of my ass here, but I remember Raja said he was frustrated with lack of budget.

I'm just waiting for Microsoft anti-trust. The store is so obviously problematic.

Im amazed they fucked themselves out of a niche product monoply with this Intel partnership.

They could have 100% in-house build an UltraThin laptop with an AMD Ryzen CPU, Vega GPU and integrated HBM that would have been equivalent to a nvidia gaming laptop with a gtx 1060 maybe even halfway to a 1070. A laptop that small and light which those specs? they could have charged a kings ransom for that shit and no other company would be able to offer a response for years.

RX580 is a solid 1080p product. Just which they did a better job with vega and came out with it sooner.

Normies ate this shit up with the app store and play store even Linux has accustomed Sup Forums types to walled gardens.

The days where you will have to root your motherboard/soc to put on 3rd party non store apps is very close anons

Why don't you just buy AMD? oh wait it's fucking shit. So shut up and enjoy

>I didn't read the thread, or even the OP.
Idk why I'm even replying, it's probably just a driveby post.

Not surprising with how people ate garbage like Steam up in 2004.

*2003

Besides nix has had GUI repos/app stores since the 90s

>It's that Sup Forumsack NEET who keeps making shitposts about nvidia
Dude you're not even being paid to shill for AMD go back to your subreddit so you can cope about vegas being great in your fantasy world.

>"""""news"""""

this, pretty solid prediction.

>Asus, Msi, and Gigabyte
They are all fucking chink quality, reddit just intentionally forgets how awful they are because they have tunnel vision about AMD.

Asus releases coolers that don't even sit properly on the GPU die.
MSI is awful chinese quality and good luck with your fucking RMA, 7 months later and you get your faulty GPU sent back with a "tested and found no problems", but hey at least their fans are quiet.
Gigabyte it's ASUS+MSI put together, bad quality, fucking fans hitting the plastic shroud(really how do you fucking miss this) awful warranty, AMD GPU are specially awful, shitty coolers that can't even cool the GPU at stock settings and of course gotta lock that voltage since your cooler can barely keep up.

Pascal has been out for what two years and it took them like one year and half to release a GPU as good as Nvidia reference GPU.
If they expect people to pay $600 for a fucking GTX1060 they better up their quality because Nvidia sells their reference GTX1060 for $299.

Attached: 1497060378194.jpg (600x450, 40K)

A series of missteps at the peak.
I warned of this coming. Wait until they launch/dont launch their new line of GPUs. That's where the fun is going to begin.

Attached: indigo.jpg (1024x683, 388K)

>Im amazed they fucked themselves out of a niche product monoply with this Intel partnership.
That speaks volumes about their confidence in Zen going forward though. They are going to get their GPUs into a key market with that move and Zen will increase their CPU sales by itself, particularly in low-power environments. It's also likely that future generations of consoles will go with AMD and semi-custom.

>RX580 is a solid 1080p product. Just which they did a better job with vega and came out with it sooner.
I think, as someone in said, that Vega is just their next step up. It's a good improvement on GCN, but it doesn't push pixels as well as Nvidia GPUs do despite having a lot more horsepower.

>AMD will give zero fucks because their product is selling and they are making money and since all their customers are miners and not gamers they will slash the budget for their drivers department and gaming performance on their cards will stagnate further. Then because they are so much further along in the benchmarks for games Nvidia will price jack even more.
I agree with most of what you said, but I doubt this. Miners aren't something that either one of them can rely on as a stable income. Once PoS hits Ethereum, I expect a lot of miners will sell their cards, while the other big pile of miners will go and mine Ethereum Classic or something.

Neither AMD or Nvidia want to rely on miners, so they will continue to chase gaming improvements for their consumer parts.

>JENSEN
>NO COOLGUY BIKER JACKET
FULL SUIT AND TIE CORPOSHILL

ITS OVER NVIDIA IS FINISHED
PS BUY NVDA, PERMANENT MOONBASE, AAPL WATCHOUT

Attached: 1.7%woodscrewhousefire.jpg (320x200, 9K)

intel never lowers prices

Precisely because they have been in such a commanding position they have not had to. Its why Nvidia has been jacking up prices generation after generation and Sup Forums applauds them for it.

I said 2004 because that is when they released the over rated garbage known as Half Life 2 to be bait to get people to install Steam.

I expected nothing more from that shit company.

Amd?

The stakes have increasing along with prices, do you think they could create a comparable architecture by today's standards with budget from 2008?

Vega isn't dead, not by a long-shot. AMD took a whopping 10% market share from NVidia the last quarter (yes, really). That's miners buying every last one, though - I haven't even seen one at any local retailer in months. I know it may seem like a failure because of the gaming performance and this and that. It's not. They are selling as fast as the board partners can make them with months of production already pre-ordered.

?
Vega M GH is the most balanced gaming GPU AMD has ever made.

I really tried I did.

I got a R9 390x and it died. Then Vega was like 6 months away. So I went for the 1070.

I even got a FreeSync monitor.

Attached: c6d.jpg (374x363, 44K)

Rajs wasn't fired, he was poached by Intel

Vega M GH has the performance of an RX470 from a roughly ~165mm^2 die compared to the RX470's 212mm^2 die. Or 200mm^2 of the 1060. That's almost 1/4th the size of a GTX 1080 die for ~half its performance.
I'm comparing to the cut down versions of those 1060 and 480 dies, but still, Vega is smaller and more efficient. It has the largest performance-to-die-size ratio of any GPU.
And it's very power efficient as long as it's not clocked too high.

Past AMD GPUs have been held back by per perf per clock per ROP, and not enough ROPs.
Vega M GH both has more ROPs, but it also has a good ratio of ROPs to CUs that Vega64 and still the 56 last.
Vega 64 should have had 80-86 ROPs, but it was made as an enterprise-first card with gamers getting the bad yields.

If AMD made a 2x size Vega M GH at desktop clocks(~1500MHz), it'd barely be behind Vega56 performance in gaming. Similar to 1070 perf, on average. 64ROPs is just not enough to feed 56CUs in almost every game. But 48 Vega CUs to 64 Vega ROPs is good and would be much more balanced.

Source for Vega M GH die size is my own extrapolation from comparing both Hades Canyon with the known ~122mm die size of the 7700k and the Hades Cayon die being slightly smaller, and the known size of the HBM package compared to the die and to Vega64's die.