AMD Computex Live Stream

[.vodka]

Am I the only one who feels stating 51% GPU Utilization using dual RX 480s isn't something to brag about?

Welps, I saw what I wanted to see. Time to go play some Overwatch.

It is something to brag about, since he stated there's some optimization that has to be done. That number and difference is going to grow. Of course the 1080 can be pushed to 2GHz to close the gap... but still, this is still GCN, there isn't a major break here like VLIW4 -> GCN was, Polaris/Vega will be at optimal driver state in little time compared to the 7970.

This is Pitcairn's replacement after all. If I can pocket $300 and have a similar experience, considering how DX12 multiadapter does away with the nightmare DX11 SLI and crossfire are where lately the second GPU is useless, then I'm all in for it.

Hell a single RX 480 could be a great replacement for my 290, that's a 150w up to date GPU kicking ass for only $200. This is a stock RX 480, we haven't even seen the effect of the 14nm node in GCN's overclocking ability, there is probably some more performance on the table that isn't being talked about.

It isn't as disruptive as the 4850 was back in its time, but it's certainly more exciting than whatever happened this year and the past one in the GPU landscape. The perf/price is just awesome.



This is good news for Vega and the 490/Fury tier. GCN4 is finally the update we didn't see in the past two versions of the architecture.

 

[IllogicalGlory]

I guess he just wanted to say that there's a lot of room for optimizations and performance may go up with time.

What I mean is, are both cards at 51%? Or is it something else, because that implies that it's barely faster than a single card. With AFR, that should be what it would mean, but I can't see why if they get with spitting distance of a 1080 with one card (probably even beating it, really as 51% on each card should be 102% of single card performance), they'd do this demo.

On a side note, nice to see 8GHz GDDR5 (not as nice as GDDR5X of course) and DP 1.4 on this thing. Hopefully it'll overclock well. Original rumors said 6GHz GDDR5.

Hopefully the arch improvements really are as great as they're saying.

 

[exar333]

somewhat disappointed. But $400 for 1080 performance in dx12. If you want to go multiGPU. DX12 explicit multiGPU is going to be in a few games.

There could still be a SKU above the RX 480. Have to learn if this is the full chip or not.

150W is in 1070 range so not impressed there. Have to see what typical consumption is. This card will do all its shining in dx12.

Yeah, they just focused on the 480. Not clear if that's the top or middle of the product stack. Hmm. I really think the top myself (8gb version) but who knows.

 

[2blzd]

the nvidia version of the game looked a lot prettier and had more units/troops on the screen. Im definitely getting a 1070 now for sure.

They're maybe room for a 480X at >$300?

 

[kondziowy]

Am I the only one who feels stating 51% GPU Utilization using dual RX 480s isn't something to brag about?

Welps, I saw what I wanted to see. Time to go play some Overwatch.

I guess he meant that Polaris will do to Pascal what GCN did to Kepler. It's not like it didn't happen before so I guess Polaris will be gaining power for up to 3-4 years( as that was also mentioned), and is much much better buy overall.

 

[Cookie Monster]

Been around the block long enough to know I would still rather have 90% performance from a single GPU rather than a 100% from 2. I liked Ryan's blog comment around pitching multi GPUs in light of how poor support has been recently from vendors. DX12 scares me a bit on that too, as recent dev work is so shoddy...

I agree. Plus two P10s are going to consume more power (2x150W as oppose to 180W of the 1080), generate more noise, take more space etc.

Hm, think ill wait for some real reviews before deciding on making the jump from my GTX780 (still plays Doom pretty well on the highest settings at 1920x1200).

 

[railven]

What I mean is, are both cards at 51%? Or is it something else, because that implies that it's barely faster than a single card. With AFR, that should be what it would mean, but I can't see why if they get with spitting distance of a 1080 with one card (probably even beating it, really), they'd do this demo.

This is exactly what I was thinking and posted in the other thread.

Feels like they missed a big "IN YO FACE" 1up. Our poster here who created the 90% of GTX 1080 in DX12 could have been vindicated.

 

[littleg]

What I mean is, are both cards at 51%? Or is it something else, because that implies that it's barely faster than a single card. With AFR, that should be what it would mean, but I can't see why if they get with spitting distance of a 1080 with one card (probably even beating it, really), they'd do this demo.

Yeah, doesn't make much sense. Possibly running on lower settings so it was CPU bound?

 

[swilli89]

The Return of the King: The new HD 4870 giant slayer.

2x Polaris beats the GTX 1080 in Ashes at 51% utilization. Not an impressive showing for Crossfire, but what does that suggest about a single card? Although what 51% utilization means, I'm not sure.

What I'm curious about is this number. Does that mean the two cards as a whole would be 100%? So at 51% utilization does that also mean basically 1 RX 480 is beating a 1080?

Or does it mean that the second card is at 51% which would put one $199 RX 480 75% as fast as a $700 1080...

 

[Sonikku]

As someone who missed out on the $200 290x's a couple years ago and as someone who's been kicking around a 7770 for perhaps too long I can honestly say a $200 480 looks VERY appealing.

 

[flash-gordon]

In the 1080 comparison he says the build cost <$500, not <$400, so it was the 8GB version and at $249 each.

 

[IllogicalGlory]

Yeah, doesn't make much sense. Possibly running on lower settings so it was CPU bound?

Hmm. I kinda doubt it. I see CPU bottlenecks as being an upper bound on the FPS numbers, and if that's true, the 1080 was GPU limited. It's probably 1080p Crazy according to AT.

I agree it makes little sense. Seems like they're hiding something.

 

[sm625]

That League of Legends video was a joke. "30 seconds until minions spawn"? I want to see a 5 man wombo with pentakill and 10 champs throwing out full endgame particle effects. It should still be able to handle that much, but still these "30 seconds until minions spawn" demos just make the game look bad.

 

[railven]

This Stream is terrible. I've had work orientations that were more engaging than this.

When they say AMD is run by engineers, they mean it. Woof. No charisma. I think I've stomached enough. Now to wait for Vega.

EDIT:

Haha
"How many people here use Windows 10?" <good show of hands>
"How many people like Windows 10?" <like 5 people raised their hands>

MSFT rep must be snapping a clipboard off camera. Okay, seriously closing the stream.

RX 480 can be a big game changer for AMD if they get it to market. GP106 can steal eat its lunch if they release around the same time. Good game AMD! Don't screw up Vega!

 

[SteveGrabowski]

Wow that was a terrible presentation. They only showed a bench for Ashes, the absolute strongest game for AMD vs Nvidia? I have a bad feeling this card is going to be a dog and that's why they're pricing it at $200. How can anyone get excited about benches for a single game? Raja didn't even look like he gave a shit.

 

[Head1985]

Am I the only one who feels stating 51% GPU Utilization using dual RX 480s isn't something to brag about?

Welps, I saw what I wanted to see. Time to go play some Overwatch.

It means single 480 is faster than 1080 or same fast as 1080 in ashes of singularity.

 

[exar333]

I thought the ambiguous '>5GFlops' was interesting. Makes me think the clocks are not yet finalized and they don't want to commit yet. Trying to squeeze a few more megahertz...

 

[deanx0r]

Not to egg on AMD, but watching their presentation makes me really appreciate the amount of showmanship Apple puts in their presentations.

 

[railven]

It means single 480 is faster than 1080 or same fast as 1080 in ashes of singularity.

Very probable, and that would had more effect in a demo/presentation versus what they showed.

But that boat is sailed.

 

[IllogicalGlory]

One thing that's MIA is talk about VR specific features. Looks like NV is going to be the leader in that department.

 

[exar333]

Very probable, and that would had more effect in a demo/presentation versus what they showed.

But that boat is sailed.

I think if that was true, they would have showed exactly that. Something odd with that...

 

[Sonikku]

AMD needs to hire people with charisma that can read off a card and deliver their lines in a "trendy" way. Sounds like a bunch of geeks just going over numerical details. They need to "sell" it if they want to be a successful business.

 

[Azix]

This Stream is terrible. I've had work orientations that were more engaging than this.

When they say AMD is run by engineers, they mean. Woof. No charisma. I think I've stomached enough. Now to wait for Vega.

was the nvidia one better? Though I guess this one is more professional. The issue is not enough information about the products and not enough hyping. More benchmarks would be nice. If they were running doom on 1440 res or something would be nice to mention.

e3 will be their more interesting one I hope. Computex is computex.

 

[JDG1980]

Haha
"How many people here use Windows 10?" <good show of hands>
"How many people like Windows 10?" <like 5 people raised their hands>

MSFT rep must be snapping a clipboard off camera. Okay, seriously closing the stream.

Why did AMD give Microsoft the stage to advertise Windows 10 anyway? This Bristol Ridge presentation is really horrible.

 

[swilli89]

One thing I noticed was the distinct lack of paid shills in the crowd due to the fact that this was hosted in a neutral setting. Remember all the tacky phrases and cheering coming from the front row at the nVidia event? Made all the more obvious they literally paid people to sound excited at their event.

Maybe that's what it takes to get that coveted brand image nectar.