Square Enix shares 'AMD Vega 20' Final Fantasy XV benchmarks

[HEXUS]

Mysterious card with 66AF:C1 device ID is a whisker away from GeForce GTX 1080 perf.

Read more.

[plexabit]

1080 performance at $250 would absolutely kill Nvidia. All their cards are so above that price point this will be the easy choice for most people.

[terrence81]

250$ for that performance would be a sweat spot

[aidanjt]

As if any more evidence were needed that FFXV is an absolute dumpster fire as a benchmark. Vega 10 already bats back and forth with the 1080, Vega 20 should be pressuring the 1080 Ti. Now if this were Navi, on the other hand, that'd be a different matter. But it's not like AMD's going to flippantly push out fraudulent code to the Linux kernel that'll need to be fixed in version following the card's launch. Linus, the display drivers devs, and distro kernel maintainers would get quite upset with them over that.

[nobodyspecial]

1080 performance at $250 would absolutely kill Nvidia. All their cards are so above that price point this will be the easy choice for most people.

If they charge less than 400 for 1080 perf, they deserve yet another bad Q report and management should be fired for trying to win an IMPOSSIBLE price war. You can't win a price war vs a company with no debt and billions in the bank when you are IN DEBT with less than a billion in cash. You will die long before they run out of money unless they put out an absolutely crap product for 3-4yrs. IE, Intel can bleed making ~20B a year now for ages vs. AMD's chips. The mistake AMD will make is charging too little which will just force Intel to respond by price cutting you to DEATH (as they can afford it), or at least until they're back on top perf wise. Meaning expect price cuts for a few years (2020 it stops? ONLY if 10nm is good) so you gain zero profits (share means nothing) so you can't do R&D for the next round. HOWEVER, if AMD actually pulls their heads out of their collective butts and charges prices that are WORTHY of Intel performance (while winning benchmarks), they will make money that will pay for R&D instead of just enough to cover interest on their debt as they've done for ages (idiots.)!

I'll tell everyone I know to sell their stock (myself included) if they charge less than $399 vs. 1080 if it performs like a 1080 as shown. I could understand a MINOR price disparity IF you're lacking in watts/heat dept, but if everything is equal anything less than $399 is moronic and begging for more analysts to down your stock! YOU WOULD DESERVE IT. I don't know who is picking pricing at AMD but they have been needing some firing for a decade. Price right, or sell the company to someone with an IQ so we can get back to real competition with AMD having the funds to do REAL R&D. They fund a gpu/cpu development then blow it on pricing front and have nothing to continue. I fear we'll do it again. Quit choosing HBM/HBM2 (soon HBM3?) and claiming 4k for your cards when NOBODY is using it and thus chooses the winner at their res of 1080p or 1440p (and not many at the higher res, never mind 4k...LOL). Why do you think NV owns the market? They make cards that aim at what we actually do. AMD keeps trying to chase things that don't exist yet or never will like console margins..LOL...or poor people APUs... LOL. Neither will make the company money to do anything other than cover your interest on your debt. You want real money, win in what we actually do. No not mining crap that is temp, but rather GAMING or APPS people will use for decades. Nvidia spends money on R&D that benefits it's users, and drops everything that doesn't make money (modems, mobile, quit if your bleeding massively before you die). Intel should have bought NV instead of spending 16B in losses on mobile...ROFL. They probably could have had them for less than that before NV took off. Intel's pride runs about as deep as AMD management stupidity.

[kalniel]

Given the current price of vega cards I don't think anyone is seriously suggesting they'd charge $250 for Vega 20, that's more likely a target price for a Navi card, the Polaris replacement.

[aidanjt]

I'll tell everyone I know to sell their stock (myself included) if they charge less than $399 vs. 1080 if it performs like a 1080 as shown.

That's daft, they can't command nVidia prices without nVidia's mind and market share. They have to compete on price to pull in customers or they wont sell. 7nm-sized Polaris chips will be cheap enough to give them a healthy profit margin even at $250.

[Corky34]

Best to ignore nobodyspecial IMHO as (s)he takes any and every opportunity to slate AMD and doesn't bother responding when challenged, (s)he'll crawl under that bridge any moment now.

[Lanky123]

I imagine AMD are worried though. It's well known in investment circles that liberal use of CAPS LOCK indicates stock market analysis of real gravitas and authority.

[tweake628]

interesting to see the 590 on that list . hummm

[plexabit]

If they charge less than 400 for 1080 perf, they deserve yet another bad Q report and management should be fired for trying to win an IMPOSSIBLE price war..

Holy long comment batman, and you might be right.

Doesn't mean the consumer side of me is complaining. I'd buy 2. I just bought my 1080 for $400 new though, so I'd say 350 would be the max considering it's still slightly less performance.

[watercooled]

It's not an impossible price war if, for example AMD ended up competing at that performance level with a traditionally small die, maybe on 7nm, vs a huge 16nm die from Nvidia with all RTX stuff (which we're yet to see a good performance demonstration of). There would come a point where Nvidia would have to be selling at a loss to compete i.e. AMD would be turning a profit and eating Nvidia's lunch.

$250 might be low-balling in the current market but it wouldn't be far off typical market expectations of actually getting more for your money with each generation, something Nvidia has utterly ignored for the 2000 series. Maybe $350 would be more realistic?

Not saying any of that will happen, but saying AMD undercutting Nvidia for 1080 performance is an outright bad idea is pretty short-sighted. You also have to consider the marketing side as aidanjt says.

As for only slightly undercutting the 1080 for 1080 performance? Now that would be daft - anyone who wanted a 1080 at 1080 pricing or close to it would have already bought one at some point over the last few years? If they're not offering more performance than the competition can offer, they quite simply need to offer substantially better value if they actually want to sell any, e.g. by attracting buyers who would see a performance increase over what was already available, for that price bracket. Offering similar performance and pricing makes sense when you're competing for sales on launch; it makes pretty much none years later!

But, having said all that, this is about Vega 20 which is AFAIK not even targeted at the gaming market, and I agree with others is unlikely to be that cheap because of its target market. The rest of my post above this line is general rather than specifically about Vega20, e.g. a hypothetical gaming card to compete at the 1080 performance level.

[DanceswithUnix]

My gut feeling is that that isn't the $250 gpu, that's the huge Instinct 7nm chip that AMD sell for AI training which they said wasn't cost effective for gaming. It would make sense, 40% more floating point throughput with no gaming optimisation could easily give the 11% more performance than the Vega 64 that we are seeing here, in which case if that is an expensive card then people would point and laugh if AMD tried to sell it to gamers for £1000.

[kalniel]

In which case, who's the cretin running it on gaming benchmarks that the public can see?

[DanceswithUnix]

In which case, who's the cretin running it on gaming benchmarks that the public can see?

There's always someone who's going to mess up. If I look around the room and can't tell who it is, I have to worry it might be me

[watercooled]

My gut feeling is that that isn't the $250 gpu, that's the huge Instinct 7nm chip that AMD sell for AI training which they said wasn't cost effective for gaming. It would make sense, 40% more floating point throughput with no gaming optimisation could easily give the 11% more performance than the Vega 64 that we are seeing here, in which case if that is an expensive card then people would point and laugh if AMD tried to sell it to gamers for £1000.

Likewise; while there's a possibility of any GPU capable of gaming making it to the retail gaming market, I don't think that's where 7nm Vega is targeted at all. And nor do I think it would be the $250 one.

But that won't stop people buying it for say $1000 and slating its gaming performance. But to avoid that, AMD could consider separating their pro card branding from Radeon...