AMD Capsaicin live stream

JediFragger · 15 Mar 2016 at 13:25

muziqaz said:
Yeah, AMD took 980ti, clocked it at 1500Mhz and ran Hitman at 1440p@60fps minimum and called it a day

Fury X sure ain't gonna do it with a 20Mhz increase

muziqaz · 15 Mar 2016 at 13:29

JediFragger said:
Fury X sure ain't gonna do it with a 20Mhz increase

Exactly

So either AMD used nvidia card, or Polaris

D.P. · 15 Mar 2016 at 14:20

Interesting enough i though. It does seem the rumours that Nvidia will partner with Samsung for HBM2 and AMD will stick with Hynix could be true, thus NVidia could have a decent head start getting HBM2 supply. However, i think Nvidia will be using HBM2 for their Tesla parts first so consumers might not get access much earlier than big Polaris.

Nutella33 · 15 Mar 2016 at 14:24

CAT-THE-FIFTH said:
Next Total War is meant to be DX12 according to the dev,so looks like another AMD sponsored game.

has it? The specs just indicate 64bit and Dx11.

D.P. · 15 Mar 2016 at 14:25

muziqaz said:
Yeah, AMD took 980ti, clocked it at 1500Mhz and ran Hitman at 1440p@60fps minimum and called it a day

I wouldn't read to much in to an IHV performance statement. Remember AD's slides showed the FuryX trouncing the 980Ti by 20-30% but if if you looked in the fine print the setting were all weird, like no AF filtering, medium quality settings, some standard things turned off. Thus they are liekly legitimate results based on a load of people testing dozens of different combinations of settings to find the exact set that maximizes the lead over competition without technically lying. And yes, Nvidia are just as guilty.

My point is simply that they may have got 60FPS with a load of settings turned down, something should take a FuryX and start selecting some medium quality settings and see what they get

JediFragger · 15 Mar 2016 at 18:33

D.P. said:
Interesting enough i though. It does seem the rumours that Nvidia will partner with Samsung for HBM2 and AMD will stick with Hynix could be true, thus NVidia could have a decent head start getting HBM2 supply. However, i think Nvidia will be using HBM2 for their Tesla parts first so consumers might not get access much earlier than big Polaris.

I can certainly imagine that nVidia would have such need for HBM2 that Hynix had no chance of satiating their demand! Hence them hunting out the Samsung partnership...

D.P. · 15 Mar 2016 at 18:44

JediFragger said:
I can certainly imagine that nVidia would have such need for HBM2 that Hynix had no chance of satiating their demand! Hence them hunting out the Samsung partnership...

For the tesla parts Nvidia can pay Samsung a huge mark-up which will make a Samsung partnership more likely.

LoadsaMoney · 15 Mar 2016 at 19:06

I just don't see the logic in them doing HBM1 and GDDR5 together again, not when HBM1 is limited to 4GB, as everyones just going to buy the GDDR5 cards, as no doubt they'll be 8GB like the 390s.

muziqaz · 15 Mar 2016 at 19:09

LoadsaMoney said:
I just don't see the logic in them doing HBM1 and GDDR5, when HBM1 is limited to only 4GB, as everyones just going to buy the GDDR5 cards, as no doubt they'll be 8GB like the 390s.

Why? If polaris 10 kicks some serious arse, it doesn't matter how much vram it has, as long as it is still kicking arse. AMD proved they can optimize memory usage.

LoadsaMoney · 15 Mar 2016 at 19:32

muziqaz said:
Why? If polaris 10 kicks some serious arse, it doesn't matter how much vram it has, as long as it is still kicking arse. AMD proved they can optimize memory usage.

It does though doesn't it, as look at the 390s, they are slower than the Furys, but they get recommended over them now for games, due to having more, and also, games coming now where you have to turn settings down on the Furys, which you don't have to do on the 390s, because they have more.

muziqaz · 15 Mar 2016 at 19:33

LoadsaMoney said:
It does though doesn't it, as look at the 390s, they are slower than the Furys, but they get recommended over them now for games, due to having more, and also, games coming now where you have to turn settings down on the Furys, which you don't have to do on the 390s, because they have more.

They get recommended due to them being great performers for their price. Simple as that. Not everyone can afford fury/x.

LoadsaMoney · 15 Mar 2016 at 19:38

muziqaz said:
They get recommended due to them being great performers for their price. Simple as that. Not everyone can afford fury/x.

Its games specs where they are getting recommended over the Furys, look at GOW ultimate, the makers are saying, ideally, you need a 390 or a 980Ti, now the Ti smashes the 390 on performance, so its obviously not because of that.

r7slayer · 15 Mar 2016 at 19:41

LoadsaMoney said:
I just don't see the logic in them doing HBM1 and GDDR5 together again, not when HBM1 is limited to 4GB, as everyones just going to buy the GDDR5 cards, as no doubt they'll be 8GB like the 390s.

But are they limited to 4GB with HMB1?

muziqaz · 15 Mar 2016 at 19:42

LoadsaMoney said:
Its games specs where they are getting recommended over the Furys, look at GOW ultimate, the makers are saying you need a 390 or a 980 Ti ideally, as they both have more vram.

yeah, all we need is a devs who have no clue what they are actually talking about to conclude that 8GB of VRAM is much better

LoadsaMoney · 15 Mar 2016 at 19:42

r7slayer said:
But are they limited to 4GB with HMB1?

Yes.

Id rather buy an arse kicking Polaris with at least 8GB vram, over an arse kicking Polaris with only 4GB, even if it is HBM1, for the reason i said above, game makers are now recommending the slower GDDR5 390s with 8GB ,over the faster Furys, with their 4GB HBM1, and are also now saying, if you don't want to turn settings down on your Fury, you should have got the slower card with more vram, as you don't have to on them.

r7slayer · 15 Mar 2016 at 19:45

LoadsaMoney said:
Yes.

I remember Joe Macri said the following

“You’re not limited in this world [die stacking world] to any number of stacks, but from a capacity point of view, this generation-one HBM, each DRAM is a two-gigabit DRAM, so yeah, if you have four stacks you’re limited to four gigabytes. You could build things with more stacks, you could build things with less stacks. Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better. So if you have four stacks you’re limited to four [gigabytes], but we don’t really view that as a performance limitation from an AMD perspective.”

who's to say they don't manage to use more stacks of HBM? And afaik they are using HBM2 on their top end cards.

Also the other cards will be using GDDR5X which has lower power usage and better memory bandwidth than standard GDDR5 so it's not the same.

Legend · 15 Mar 2016 at 19:49

r7slayer said:
I remember Joe Macri said the following

“You’re not limited in this world [die stacking world] to any number of stacks, but from a capacity point of view, this generation-one HBM, each DRAM is a two-gigabit DRAM, so yeah, if you have four stacks you’re limited to four gigabytes. You could build things with more stacks, you could build things with less stacks. Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better. So if you have four stacks you’re limited to four [gigabytes], but we don’t really view that as a performance limitation from an AMD perspective.”

who's to say they don't manage to use more stacks of HBM? And afaik they are using HBM2 on their top end cards.

And AMD also said that the Fury X would be an "overclocker's dream"... I don't put any stock in anything AMD or Nvidia say, or even the press... I will believe it when I see it. It's all marketing hype and nonsense, and invariably believing a shred of it only leads to disappointment.

LoadsaMoney · 15 Mar 2016 at 19:49

r7slayer said:
I remember Joe Macri said the following

“You’re not limited in this world [die stacking world] to any number of stacks, but from a capacity point of view, this generation-one HBM, each DRAM is a two-gigabit DRAM, so yeah, if you have four stacks you’re limited to four gigabytes. You could build things with more stacks, you could build things with less stacks. Capacity of the frame buffer is just one of our concerns. There are many things you can do to utilise that capacity better. So if you have four stacks you’re limited to four [gigabytes], but we don’t really view that as a performance limitation from an AMD perspective.”

who's to say they don't manage to use more stacks of HBM? And afaik they are using HBM2 on their top end cards.

Oh, i never knew that, interesting, just have to wait and see then i guess.

r7slayer · 15 Mar 2016 at 19:54

Legend said:
And AMD also said that the Fury X would be an "overclocker's dream"... I don't put any stock in anything AMD or Nvidia say, or even the press... I will believe it when I see it. It's all marketing hype and nonsense, and invariably believing a shred of it only leads to disappointment.

Yea I know they said that and it was total garbage but honestly do you beleive they can't use more stacks of HBM? For all we know they have created more room on the interposer to allow for more stacks of HBM (because of a smaller die?). Anyones guess atm. Don't think a previous cards limitation will mean future cards will suffer from the same limitation. Kinda naive.

LoadsaMoney · 15 Mar 2016 at 19:55

r7slayer said:
Yea I know they said that and it was total garbage but honestly do you beleive they can't use more stacks of HBM? For all we know they have created more room on the interposer to allow for more stacks of HBM (because of a smaller die?). Anyones guess atm. Don't think a previous cards limitation will mean future cards will suffer from the same limitation. Kinda naive.

It is AMD though remember.