AMD Radeon Vega in the League of GTX 1080 Ti and TITAN Xp

[oxidized]

Your reading too much into it.

Maybe it was a moment of honesty and your going to need good voltage to run it.

Yeah, sure

 

[renz496]

This definitely rings true, and a big win for AMD. Nvidia will have to adapt to FreeSync or be left behind. Right now they are probably thinking hard on how to make that move and still maintain face in the process.

there is no holding monitor maker to make much cheaper Gsync monitors but "nvidia" itself was associated with "premium" so many monitor maker just take advantage of that. i still remember when nvidia mention that they want the fist gsync based monitors to cost no more than $400 but in the end it end up being nvidia own pipe dream when partners still charge way more than that. and that licensing fee also makes sure all gsync panel to work the same regardless of the maker. even AMD themselves thinking about charging "premium" with freesync 2. because they are very aware to make sure all freesync 2 monitors out there work as they intended to be they need to dedicated more resource into it and they are thinking to past the cost to consumer.

On the topic of whether Vega will compete with 1080ti, I'm pretty sure it will, at least in Vulkan or DirectX 12 games. Might even be faster, honestly. And for much cheaper? If they can do this, people will flock to the AMD platform in masses. They know this, so I hope they are not planning to disappoint. This is the biggest test for AMD after they proved themselves in the CPU market against Intel, and might be the biggest test for them yet, whether they can also beat Nvidia? I really hope they can make Vega what people want and show it to Nvidia after they (kinda) beat Intel.

they might be able to compete with nvidia on performance but to be cheaper at the same time is definitely not. HBM2 is expensive than GDDR5 module. if the performance is very competitive the price will will directly reflect that. just look at the recently released RX580 and RX570. some people expect AMD will set RX580 MSRP at $200 and RX570 at $150. but in the end RX580 MSRP is only $10 lower than RX480 and RX570 MSRP is the same as current RX470 MSRP. many people forgot that while it is important for AMD to recapture their market share it is more important for them to get more profit or else they can't further fund their R&D. to me this refresh is more like giving the chance for board partner to "re-inflate" the price back up. remember some of new custom RX580 cost around $260-$280 right now. but if you're lucky you can will be able to get great deal on existing RX400 series. last time i saw some custom RX480 8GB can be had much less than $200 on newegg.

The good part is that Intel is now announcing consumer 6 cores for their next gen CPUs and even going to release them sooner than they "planned". But in reality, would they have done it if AMD didn't kick their asses in multi-core? Same for Nvidia, would they be making HBM2 and releasing it sooner now if AMD wasn't able to kick their asses in Graphics? I bet Nvidia already smelled or even knows for sure that Vega will be a threat. Otherwise, they would never announce Volta for so much sooner, wouldn't they?

nvidia not using HBM2 on consumer grade cards is not because AMD not competing with performance. it is because of economic reason. also their architecture is much more power efficient than AMD that they still able to get away using more power hungry GDDR5/GDDR5X. and nvidia for their part never underestimate AMD regardless AMD being competitive or not. after that HD4800 moment nvidia never really let their guard down when it comes to AMD.

 

[Gasaraki]

I often become frustrated when there is a cpu or gpu comparison. It's always seems to be i7 6700k or i7 7700k compared to every AMD Ryzen cpu that has been released. Doesn't Intel manufacture anything else or must these comparisons always rely on Intel's best clocking cpu against AMD's least powerful to their best. Is there pressure from Intel for analysts to do this? We all know that these two Intel CPU's clock to almost 5.0 ghz so it's a win straight away on single core fps comparisons. Then there is the Nvidia comparisons. We all know that Radeon GPU's don't clock as high as Nvidia's 10 series of cards. But when I see a GTX1070 compared to the RX480 or RX580 clocked up to 2,000 ghz to prove that their card can achieve 300 fps compared to 200 fps for the Radeon. Who the fuck cares. Sure, Nvidia wins, but do they really. The maximum card speed should be no more than the monitor clock speed. The maximum these days is 144 mhz and both AMD and Nvidia shoe it in. Then there is Radeon Chill for the AMD card which allows users to set maximum and minimum fps speeds for the game. What a fantastic addition to the card with such finesse and control. The GPU is slowed down, power and heat are both reduced. Amazing tech.

Then there was the Nvidia fanboy who bragged about how his GTX 1080 which he overclocked to 2,000+ mhz was better than the AMD Radeon RX Vega. What a fuckwit. Who in their right mind would run their GPU clock so fast on a continuous basis. The Vega card was an engineering sample most likely running with a conservative clock playing Doom. I saw 70-80 fps continuous with an occasional dip to 60 fps running 4k on ultra settings. Not shabby at all.

I can go on and on particularly on how DX12 and Vulkan favour AMD and how DX11 favours Nvidia. Just more absolute rubbish.

It's by price and stock speeds. If I spend $230 on an Intel processor what can I get for $230 on the AMD camp. I could care less about the raw MHz rating.

 

[HD64G]

BTW, to anyone who doesn't remember well, Volta should be on sale since 2016 (info taken from nVidia future release graphs from 2014)...

And you complain about Vega being late before the Q2/17 ends, which is where it was to launch from start...

And finally, imho, being late at a party isn't bad if your entrance makes everyone look upon you...

 

[m1dg3t]

The rough 3.9-4.1Ghz limits of Ryzen (whether it's 8 cores or the 4 core versions) is a limiting factor for some. I game at 1440p so it doesn't matter to me. I also dont need 144fps so again - no biggy. But the fact is pretty blatant and refusing to see it is obvious contrary. Ryzen clocks slower, therefore, in situations where raw clocksped is king, Ryzen loses some of those fights. It's simple. So for Vega, again, clockspeeds will be the determining factor, not so much the hardware. AMD have put more hardware into their chips for a while now but the clocks have kept them back (or is the chatter about RX580 clocking higher meaningless fluff for it's performance? No, of course not - it's fundamental to it's increase over the RX480).

I believe that is a misconception by the community, AMD isn't "suffering" from a clock speed defecit, they are suffering from a lack of efficient software. They cram so much hardware into their chips that a great portion sits idle, consuming power...

Just look at their GPU history since hd 7xxx, did the hardware magically 're-fab' itself for more performance? No. It just took the industry that long to make use of what they were given, the 7950 I have was the best PC purchase I ever made! $330 and it keeps getting better, although it is loooooooooong in the tooth now

Polaris is competitive with Pascal, it just lacks the streaming processors to compete with high end NVIDIA cards. AMD's focus lately has clearly been selling cards in volume rather than trying to be king. I suspect that's because of GloFo more than anything.

Navi will follow Volta.

I doubt GloFo was the sole reason, they went after the best part of the pie. With limited resources, they have to be extremely efficient at maximizing ROI.

Unfortunately, AMD do not have the same billions of dollars for R&D and staffing numbers. They have been extremely inventive currently and in the past and have put to notice both Intel and Nvidia.

Considering their competition, it's a miracle they're still producing new tech and trying to spark/innovate software. All these people talking on forums about how/what AMD could/should do are a joke! They wouldn't last 5min in that hot seat.

Intel/nVidia are happy to milk the industry with status quo.

Unfortunately this isn't Nvidia or Intel's fault. It's their business decisions that led to this. The ideas they have are indeed something that they should focus on and hope that this time around they would actually work (HBM1 was a flop, their venture into ARM was a flop as well, Mantle was a flop and so forth)

Unfortunatley it was a perfect storm, anti compete tactics from BOTH Intel & nVidia coupled with poor management.

HBM1 is/was not a flop, neither is/was Mantle. How's nVidia doing with DX12 & Async shaders?

 

[renz496]

And when has AMD failed to deliver? Their last gen R9 290X was so fast they could just rebrand it to R9 390X and still compete with brand new GTX 980. That's more a definition of pwnage than fail. The RX480 was a mid end economy targeted product. They literally targeted segment where most profit is made. They literally didn't even bother to compete with top end products. And the card is far from bad. FAR FROM IT. Even the R9 Fury X which many consider as fail is not one at all. Sure, in some games it's beaten by GTX 980Ti by a lot, but in others, it almost matches GTX 1080 and certainly goes past GTX 1070. For a card that targeted GTX 900 series as competition, that's again hardly a fail. But whatever...

the mainstream hardware might have the most volume. but did the segment also have the most profit? no. most profit coming from high end segment where profit margin obviously bigger.

https://www.jonpeddie.com/press-releases/details/pc-gaming-hardware-market-minting-billions

those hawaii is really nice (and lets's face it AMD hardware in both major console also help more games to be tweaked more for their hardware) but fury not so much. the card suffer from imbalance design. to be honest if the person only play on 1080p i have much harder time recommending the card. it is no joke when 1060 capable of beating Fury X at some titles at 1080p:

https://www.techpowerup.com/reviews/MSI/GTX_1060_Armor/6.html
https://www.techpowerup.com/reviews/MSI/GTX_1060_Armor/7.html
https://www.techpowerup.com/reviews/MSI/GTX_1060_Armor/10.html
https://www.techpowerup.com/reviews/MSI/GTX_1060_Armor/13.html
https://www.techpowerup.com/reviews/MSI/GTX_1060_Armor/16.html

 

[oxidized]

...

Same old story, which sounds so politically correct from a couple of years, nVidia and intel are to blame for AMD or ATi failures. Yeah, keep telling that to yourself

 

[m1dg3t]

Same old story, which sounds so politically correct from a couple of years, nVidia and intel are to blame for AMD or ATi failures. Yeah, keep telling that to yourself

Clearly you fail at reading comprehension. Thanks for coming out!

 

[oxidized]

Clearly you fail at reading comprehension. Thanks for coming out!

Maybe, my understanding isn't that good, but the last part of your post sounded pretty clear to me, correct me if i'm wrong!

 

[m1dg3t]

Maybe, my understanding isn't that good, but the last part of your post sounded pretty clear to me, correct me if I'm wrong!

??? I did. LoLoLoL

 

[the54thvoid]

I believe that is a misconception by the community, AMD isn't "suffering" from from a clock speed defecit, they are suffering from a lack of efficient software. They cram so much hardware into their chips that a great portion sits idle, consuming power...

Just look at their GPU history since hd 7xxx, did the hardware magically 're-fab' itself for more performance? No. It just took the industry that long to make use of what they were given, the 7950 I have was the best PC purchase I ever made! $330 and it keeps getting better, although it is loooooooooong in the tooth now



I doubt GloFo was the sole reason, they went after the best part of the pie. With limited resources, they have to be extremely efficient at maximizing ROI.



Considering their competition, it's a miracle they're still producing new tech and trying to spark/innovate software. All these people talking on forums about how/what AMD could/should do are a joke! They wouldn't last 5min in that hot seat.

Intel/nVidia are happy to milk the industry with status quo.



Unfortunatley it was a perfect storm, anti compete tactics from BOTH Intel & nVidia coupled with poor management.

HBM1 is/was not a flop, neither is/was Mantle. How's nVidia doing with DX12 & Async shaders?

Ryzen does clock slower. I don't mean IPC, I actually think it's better than Skylake at the same clocks. It's not a community thing, it's a real, evidentially based thing. Skylake and Kabylake clock higher than Ryzen. Even the 4 core Ryzens seem to be topping out at 4.1Ghz.
But I firmly believe a refresh and process maturity will allow higher clocks #thats why i bought the platform. I'd like to think in 2 years time i can upgrade to an 8 core Ryzen at 4.2-4.4Ghz. That would be pretty cool.
I am a Ryzen fan, it's working great for me.

 

[oxidized]

??? I did. LoLoLoL

Alright now i get you...

 

[m1dg3t]

Ryzen does clock slower. I don't mean IPC, I actually think it's better than Skylake at the same clocks. It's not a community thing, it's a real, evidentially based thing. Skylake and Kabylake clock higher than Ryzen. Even the 4 core Ryzens seem to be topping out at 4.1Ghz.
But I firmly believe a refresh and process maturity will allow higher clocks #thats why i bought the platform. I'd like to think in 2 years time i can upgrade to an 8 core Ryzen at 4.2-4.4Ghz. That would be pretty cool.
I am a Ryzen fan, it's working great for me.

Yes it's absolute max frequency is less than Intel, but like you yourself said their IPC is equal to or greater than Intel - at the same freq. How does Intels 8/16 compare to Ryzen 8/16???

I'm not interested in high Mhz ST perf on a 4c CPU. I've had that since 2012, thank you very much.

Now imagine, for a second, an ecosystem designed to make best use of AMDs designs, where would that leave Intel/nVidia???

I see it coming, and I'm embracing it...

 

[bug]

BTW, to anyone who doesn't remember well, Volta should be on sale since 2016 (info taken from nVidia future release graphs from 2014)...

And you complain about Vega being late before the Q2/17 ends, which is where it was to launch from start...

And finally, imho, being late at a party isn't bad if your entrance makes everyone look upon you...

Well, AMD planned to have HBM2 parts in 2016: http://www.anandtech.com/show/9233/amds-2016-gpu-roadmap-finfet-high-bandwidth-memory
If that wasn't Vega, I don't know what is. Like you, I don't really care when it's released. But if it only offers last year's performance, I'm not impressed. Last year's performance at half the price would be an entirely different story

 

[m1dg3t]

Alright now i get you...

Nobody "Gets me". I'm retarded, and proud of it

 

[m1dg3t]

Well, AMD planned to have HBM2 parts in 2016: http://www.anandtech.com/show/9233/amds-2016-gpu-roadmap-finfet-high-bandwidth-memory
If that wasn't Vega, I don't know what is. Like you, I don't really care when it's released. But if it only offers last year's performance, I'm not impressed. Last year's performance at half the price would be an entirely different story

Lack of HBM supply has nothing to do with AMD, you need to put SK Hynix in the deep fryer for that 1. Where are all of nVidias HBM equipped cards?

 

[renz496]

Well Hype is what users create. AMD or any other company is just presenting a product. So please stop being a user and put yourself together. I'm tired of this Hype this Hype that. It is so boring. Every time somebody has a different point of view or disagrees with something there must be a Hype hypothesis in the content of people who's wishful thinking that what they already bought is better no matter what. Disappointing really.


you think that's lovable? Maybe same thing that we should love intel for introduction of new gen CPU's with 2% IPC gains over previous with a tremendously high price point. I can see similarities with those 2.
That's what I see in NV strategy, which last year shows flawlessly.
1070, 1080, Titan X(pascal) 1080TI, Titan Xp new. Think of the price tags on those when released.

then what about RX480? the card is even much slower than 390X that it should replace. and if you look at it closely there isn't really much difference between nvidia and AMD when it comes to performance improvement over the year. nvidia still coming up with 10%-15% performance increase each year and for AMD while we got much bigger performance jump from their previous flagship to new flagship you also need to wait longer for it because of their 2 year cadence in upgrading their flagship. also pricing wise it is not much different since 2013. you mentioned 1080ti for example but 780ti also cost $700 back in 2013. the only mistake that nvidia did with 1080/1070 was the FE pricing. because it encourage the board partner to price their card near or even exceed the FE base price instead of the actual MSRP.

 

[bug]

Lack of HBM supply has nothing to do with AMD, you need to put SK Hynix in the deep fryer for that 1. Where are all of nVidias HBM equipped cards?

Short attention span? I was responding to the guy's claim that Vega was always meant for a 2017 launch.

Nvidia's HBM2 cards are Teslas, available for almost a year now.

 

[renz496]

Yeah I'm confident it won't be just gtx 1080 levels, heck that is _old_ gpu released 27.05.2016, so almost year old gpu(=~ year old if vega really be released 25.05.2017).

Rumored die size makes it bigger than gp102 and it's on a little bit denser manufacturing process of GF, so there's more transistors/mm² too. But that does not tell much, vega might just have fatter shaders for fp64 and compute(If amd have not take nvidia way to make two chips one vega for compute and other vega for gaming).

Well we can only wait and see(and try to filter hype out of the rumors).

it seems Vega 10 will be configured the same way as fiji when it comes to DP. i heard it will be vega 20 that will be AMD next true DP card replacing their current Hawaii.

 

[cdawall]

Yes it's absolute max frequency is less than Intel, but like you yourself said their IPC is equal to or greater than Intel - at the same freq. How does Intels 8/16 compare to Ryzen 8/16???

It completely depends on application. There is still a large number of workloads that the 5960x wins not only clock for clock, but in maximum clockspeed.

 

[oxidized]

Nobody "Gets me". I'm retarded, and proud of it

Watch out, someone could take you seriously

 

[m1dg3t]

Short attention span? I was responding to the guy's claim that Vega was always meant for a 2017 launch.

Nvidia's HBM2 cards are Teslas, available for almost a year now.

Yeah, and? I got ADHD so what LoLoLoL

Maybe you should have messaged him/her in PM if you wanted a private convo, this is a public forum after all

The question was rhetorical, but please tell me, how is availability of those cards?

AMD in all it's sadness has been giving us HBM for how long now?

 

[xkm1948]

I do hope AMD can compete in high end. But seriously I am not getting my hope high this time. NCU sounds great on paper. However the 64CU with 4096SP sounds awesfully close to a tweaked FijiXT core instead of a built-from-ground-up design. With so many resources poured into RyZen I don't believe RTG have enough brain power to get something new from the ground up. Best case scenario it will be an optimized FijiXT with a lot higher clock speed(1500MHz maybe for the AIO version)

 

[Dippyskoodlez]

Right, and when NVIDIA releases same info, everyone jumps on it as absolute fact...

The difference here is that nvidia pulls through reliably - and then tells you it's available monday.

Vega is already climbing a steep hill if competing with 1080ti is it's target, because 1080 ti's are already in gamers hands.

Vegas target market is shrinking every time a 1080 ti is sold.

 

[cdawall]

I do hope AMD can compete in high end. But seriously I am not getting my hope high this time. NCU sounds great on paper. However the 64CU with 4096SP sounds awesfully close to a tweaked FijiXT core instead of a built-from-ground-up design. With so many resources poured into RyZen I don't believe RTG have enough brain power to get something new from the ground up. Best case scenario it will be an optimized FijiXT with a lot higher clock speed(1500MHz maybe for the AIO version)

AMD's platform iteself didn't change, everything they build is an expandable design. It also sounds like two RX480 cores with HBM added.