Vega 56 undervolting can beat GTX 1080 in benchmarks.

[ravenshrike]

Just now, LAwLz said:

 

Even if we say that Raja is a super genius, designing a GPU is a massive undertaking which requires many people. One person getting replace won't make a company go from making bad cards to making great cards.

*looks at Apple* One person in the right position can make all the difference.

[Misanthrope]

27 minutes ago, LAwLz said:

Please... No more hyping up future AMD products.

How many times do people need to get burned before they stop?

Unfortunately since Zen ended up delivering this will continue for years.

[MyName13]

4 minutes ago, ravenshrike said:

*looks at Apple* One person in the right position can make all the difference.

*cough cough* the right man in the wrong place can make all the difference in the world.

 

29 minutes ago, LAwLz said:

Please... No more hyping up future AMD products.

How many times do people need to get burned before they stop?

 

Even if we say that Raja is a super genius, designing a GPU is a massive undertaking which requires many people. One person getting replace won't make a company go from making bad cards to making great cards.

But hype is the only thing RTG has left

 

2 minutes ago, Misanthrope said:

Unfortunately since Zen ended up delivering this will continue for years.

Even Zen is overblown, year sure hexa core and octa core CPUs are great because intel has nothing to offer for that price, but quad cores?Nahh, they are nothing new in the best case, they perform worse than intel's equivalents in certain workloads.

[Misanthrope]

4 minutes ago, MyName13 said:

Even Zen is overblown, year sure hexa core and octa core CPUs are great because intel has nothing to offer for that price, but quad cores?Nahh, they are nothing new in the best case, they perform worse than intel's equivalents in certain workloads.

Well that hype was a bit weird: we knew for months that this would pretty much match Sandy to Haswell and that's pretty much what we've got. The thing is that people overestimated how important was to have 4/4 cores/threads for the lower end chips since they really don't outshine the i3s by that much: almost always worst maximums and better minimums but averages are pretty consistent with Kaby 2/4 chips.

 

However 6/12 actually does makes a difference for the midrange: a 1600 shouldn't be able to compete well with i5s at substantially lower cores but it still does since things like BF1 or Wildlands actually do like more than 4/4 threads. A 6/12 chip at 4/4 prices from intel is very disruptive imo.

[MyName13]

3 minutes ago, Misanthrope said:

Well that hype was a bit weird: we knew for months that this would pretty much match Sandy to Haswell and that's pretty much what we've got. The thing is that people overestimated how important was to have 4/4 cores/threads for the lower end chips since they really don't outshine the i3s by that much: almost always worst maximums and better minimums but averages are pretty consistent with Kaby 2/4 chips.

 

However 6/12 actually does makes a difference for the midrange: a 1600 shouldn't be able to compete well with i5s at substantially lower cores but it still does since things like BF1 or Wildlands actually do like more than 4/4 threads. A 6/12 chip at 4/4 prices from intel is very disruptive imo.

Actually r3 doesn't even have better minimums than i3, it's even worse than the Pentium according to some benchmarks (in gaming).Why shouldn't a 6c12t CPU be able to compete with a 4c4t CPU?Even if it performs like Sandy bridge (and in some workloads it does, at least the r3) it still has more cores.1600 and 1700 are disruptive, r3, 1400 and 1500x are not, they actually offer less in certain workloads and a little bit more in others, maybe that's the reason why nobody mentions them, a quad core that costs more than 160$ simply isn't a good investment today especially with coffee lake around the corner.

[Tenelia]

10 hours ago, mr moose said:

EDIT: talking purely from a gaming perspective here.   Compute is a different kettle of fish.

 

Indeed. I quite enjoy the compute/price. 

[Misanthrope]

1 hour ago, MyName13 said:

Actually r3 doesn't even have better minimums than i3, it's even worse than the Pentium according to some benchmarks (in gaming).Why shouldn't a 6c12t CPU be able to compete with a 4c4t CPU?Even if it performs like Sandy bridge (and in some workloads it does, at least the r3) it still has more cores.1600 and 1700 are disruptive, r3, 1400 and 1500x are not, they actually offer less in certain workloads and a little bit more in others, maybe that's the reason why nobody mentions them, a quad core that costs more than 160$ simply isn't a good investment today especially with coffee lake around the corner.

Well you kinda make my argument for me at the very end: Coffee lake wouldn't be about to offer anything beyond 4 cores if it wasn't for Ryzen, full stop.

[mynameisjuan]

1 minute ago, Misanthrope said:

Coffee lake wouldn't be about to offer anything beyond 4 cores if it wasn't for Ryzen, full stop.

Or you know they were planning on releasing a 6 core anyway.....Not everything intel does is because of Ryzen. Intel was working on coffee lake long before Ryzen release.

[MyName13]

4 minutes ago, Misanthrope said:

Well you kinda make my argument for me at the very end: Coffee lake wouldn't be about to offer anything beyond 4 cores if it wasn't for Ryzen, full stop.

Maybe, although we have known for coffee lake for quite some time.

[Misanthrope]

26 minutes ago, mynameisjuan said:

Or you know they were planning on releasing a 6 core anyway.....Not everything intel does is because of Ryzen. Intel was working on coffee lake long before Ryzen release.

Intel also knew AMD would come up with more cores so it actually doesn't proves anything. At the end of the day intel spend 5 years with nothing above 4 cores and the very same year AMD comes up with Ryzen they're ready with a 6 core: If they were ready for a while it also means they held it back for a while since well, nobody was competing with them anyways.

 

Look I'll go back to shit on Vega but I am not gonna take credit away from Ryzen: if anything most of us are capable of calling out both the good and the bad.

[Jito463]

4 hours ago, LAwLz said:

Please... No more hyping up future AMD products.

How many times do people need to get burned before they stop?

 

Even if we say that Raja is a super genius, designing a GPU is a massive undertaking which requires many people. One person getting replace won't make a company go from making bad cards to making great cards.

The thing is, we know for a fact that AMD cut R&D money for RTG.  We know for a fact that they focused their priority on mobile/APU graphics.  This is well known, and the reason I wasn't surprised by the lack of performance in Vega (I was hopeful they would pull it off, and disappointed that they didn't, but not surprised).

 

It's also a fact that they're investing more money and resources into RTG now, which means there is hope for Navi.  You may have abandoned any hope that they'll turn things around, but some of us still are willing to give them at least the benefit of the doubt, given the circumstances.

2 hours ago, Misanthrope said:

Well you kinda make my argument for me at the very end: Coffee lake wouldn't be about to offer anything beyond 4 cores if it wasn't for Ryzen, full stop.

2 hours ago, mynameisjuan said:

Or you know they were planning on releasing a 6 core anyway.....Not everything intel does is because of Ryzen. Intel was working on coffee lake long before Ryzen release.

2 hours ago, MyName13 said:

Maybe, although we have known for coffee lake for quite some time.

 

And as Misanthrope has pointed out already, Intel has known about Ryzen for some time now.  So either they've finally reached a tipping point where they can't eke out any more performance while still keeping it a quad core, or they realized that they needed to increase their core count to remain ahead of AMD (or some combination of the two).

 

Even if it's strictly the former, you can't discount AMD has done their part to force Intel to reduce prices, at the very least in the HEDT segment.

 

[Misanthrope]

Similar case: Nvidia immediately saying "Yeah we don't need Volta anytime soon" after looking at Vega. You know they could release Volta early if needed since it's ready for Quadros already so I have to assume that intel could have said something very similar and said "We don't need a 6 core chip for consumers Ryzen can't compete with Cannon lake" if it turned out Ryzen was quite shit by comparison. 

[Guest]

So, its wasted fame ?

[ravenshrike]

7 hours ago, MyName13 said:

But hype is the only thing RTG has left

The Vega Nano will be by far the smallest card with the highest capability. So for small form factor builds AMD will win the race. Especially once properly undervolted and tweaked.

[MyName13]

15 minutes ago, ravenshrike said:

The Vega Nano will be by far the smallest card with the highest capability. So for small form factor builds AMD will win the race. Especially once properly undervolted and tweaked.

Aren't there sff GTX 1070/1080/1080 tis?I remember some small zotac and single fan MSI models, a hot power hungry GPU and sff don't go together.

[ravenshrike]

48 minutes ago, MyName13 said:

Aren't there sff GTX 1070/1080/1080 tis?I remember some small zotac and single fan MSI models, a hot power hungry GPU and sff don't go together.

The Vega Nano will be smaller as the HBM2 will allow it to be the length of the socket. Moreover, it's unlikely the 64 will be what its based on, and properly powered Vega 56 is significantly less power hungry.

[Nowak]

19 hours ago, Trik'Stari said:

I do not understand how they did so well with Ryzen and Threadripper, then completely drop the fucking ball and start eating glue with Vega. Seriously, someone needs to get fired from that division because he's a fucking moron.

Really, the Answer is Just so dAmn obvious.

[Trik'Stari]

11 hours ago, Taja said:

Yeah, thats what I tought. There is no way that is consistent. Its like getting a monster 980ti with 1450Mhz and compairing to a 1080 FE that is not boosting properly because temps. Nowhere near a real usecase comparison.

 

Vega was bad, so bad. Man, I was hoping it would bring 1070 performance at lower prices, so I could MAYBE buy it in my country

And then you look at a 1080 ti FE in a ridiculous water cooling loop like mine, and laugh at everything else.

 

My GPU runs cooler playing games at 5760x1080p than the CPU does whilst playing said games.

[wrath]

huh interesting

 

[cj09beira]

23 hours ago, ravenshrike said:

The Vega Nano will be smaller as the HBM2 will allow it to be the length of the socket. Moreover, it's unlikely the 64 will be what its based on, and properly powered Vega 56 is significantly less power hungry.

it will be the full 64 Cu chip, but probably running lower voltages and binned chips, lower voltages can do wonders to these gpu's power consumption 

[roax206]

well if its anything like their CPUs(and yes I know that is practically a different company) reducing the voltage is practically a given. AMD has a bad habit of setting really high voltages on their processors as a blanket stability measure. It is impossible to actually damage a processor(CPU or GPU) by reducing the voltage, in fact a large amount of wear is caused by excess voltage. NOTE: What usually scares people is when they set the voltage too low for the processor to run in which case ensure you have a way to reset the given voltage if the processor doesn't start.

[Humbug]

On 8/23/2017 at 3:21 PM, Misanthrope said:

The card was not ready, not even with the delays. It was probably a matter of "just fucking release as is now or you might end up having to scrap out the entire Vega lineup"

yes. They even did a paper launch of Vega FE non gaming card at the end of June just so that they could technically keep their promise of a 1st half 2017 launch of Vega.

 

On 8/23/2017 at 6:05 PM, ravenshrike said:

The guy who dreamt up Vega has been removed/fired

LOL, who might that be?

 

 

 

[Humbug]

On 8/23/2017 at 6:22 PM, LAwLz said:

Please... No more hyping up future AMD products.

How many times do people need to get burned before they stop?

Apart from the obvious supply issues I think AMD should get their product positioning right.

 

If they know that they have to push their architecture too far in order to reach the performance of a given Nvidia product then I think they should not try... Rather than doing it at all costs and then having memes made about them being power hungry.

In this case I really don't think AMD should have launched both the Vega56 and the Vega64.

 

They should have had one gaming Vega product; possibly with the full 64 CUs and 256 TUs, but it should have been running at Vega56 clockspeeds (stock). Now you have a card with not unreasonable power consumption, faster than a GTX 1070. If users want to raise the power limit and overclock to match the GTX 1080 then they can do so... But at stock it comes at GTX 1070 prices and beats it.

Yes we would still be lamenting the fact that AMD so late still does not have a stock 1080 and 1080ti beater. But at least then they are giving us the best most optimal version of Vega which dominates in it's segment.

 

There is no point trying to over reach and hit every market segment of your competitor. Give us the best product/s of your uarch.. When it's time to launch Vega20, and later Navi again make a reassessment of which markets you can win at.

[TheOriginalHero]

On 23/08/2017 at 4:00 AM, Trik'Stari said:

I agree.

 

I do not understand how they did so well with Ryzen and Threadripper, then completely drop the fucking ball and start eating glue with Vega. Seriously, someone needs to get fired from that division because he's a fucking moron.

 

Hell, wasn't "make a card that can beat the 1080 when crossfired, so that two of that card costs less than a 1080" an option? I'd prefer that option than "not really any better performance reliably, for a similar price point"

 

Find a way to price it so three of that card beat a 1080ti but all together costs less than the 1080ti. That's a better option than this.

Vega isnt even a bad architecture? You've to keep in mind that amd barely has a fraction of Nvidia's RnD capital. They have to design an architecture capable of competing in multiple applications. Vega is actually very good on the raw compute side of things. The professional vega based cards all seem pretty good for the price. Sure gaming doesnt benefit from the raw compute power of the architecture but it still performs decently. 

[Delicieuxz]

On 23/08/2017 at 5:35 AM, ravenshrike said:

The guy who dreamt up Vega has been removed/fired. However by the time they brought Koduri back in, it was too late to do a complete design so he salvaged it as best he could. Navi will be Koduri's brainchild alone so it should be interesting to see what happens.

Can you share a link with some info about the circumstances under which the original Vega designer was moved from their position?