Radeon Pro Duo drops down to $800 MSRP

[Starelementpoke]

1 minute ago, goodtofufriday said:

1200w psu from when i ran 2 6990s. I fully intend to go 2x duo pros. 

Dope

[Energycore]

1 hour ago, goodtofufriday said:

1200w psu from when i ran 2 6990s. I fully intend to go 2x duo pros. 

They even cost less than 2 6990s did back in the day

[Prysin]

9 hours ago, Misanthrope said:

This can kinda be a Titan x pee alternative for those Korean 4k freesync screens. 

 

But I'd wait for even lower prices with Vega close to follow they could get even cheaper: the 295x2 did but this one might actually run on a sub 700watt psu.

The 295x2 + rest of the system can be ran on a good quality 650w psu,  like a EVGA G2 or RMx650

[goodtofufriday]

1 hour ago, Energycore said:

They even cost less than 2 6990s did back in the day

I try not to think about what i spent back then lol. 

[Citadelen]

14 hours ago, PCGuy_5960 said:

There are a couple of reasons why you don't wanna do that..... 

1. Size
2. Power consumption
3. Heat
4. Liquid cooled

For an HTPC the RX 480/GTX 1060 are perfect choices IMO....

1) Is it quite big but it looks amazing.

2)Power draw is lower than two 980 Ti's.

3)TDP is only 350W

4)Could be inconvenient but I think it's more of a plus than a negative.

[PCGuy_5960]

4 hours ago, Citadelen said:

1) Is it quite big but it looks amazing.

2)Power draw is lower than two 980 Ti's.

3)TDP is only 350W

4)Could be inconvenient but I think it's more of a plus than a negative.

Aren't HTPCs meant to be small and efficient? 

How can you fit it in something like a Node 202, let alone power it.....

For a normal desktop, it's a steal for the price!

[Agosto]

Still 4GB only for GPU... makes no sense

[dalekphalm]

1 hour ago, PCGuy_5960 said:

Aren't HTPCs meant to be small and efficient? 

How can you fit it in something like a Node 202, let alone power it.....

For a normal desktop, it's a steal for the price!

HTPC's are supposed to be small, yes, but efficient? Well in theory, everything should be efficient, even the largest full ATX gaming beast. But I would say that's not an inherent requirement of an HTPC, no.

 

Granted, fitting one or two of these into an HTPC sized case (ITX or smaller mATX) would be quite the challenge, no doubt.

[dalekphalm]

22 minutes ago, Agost said:

Still 4GB only for GPU... makes no sense

This isn't new...

 

Also, because of the increased VRAM bandwidth, some benchmarks have shown that it doesn't get held back as much by the 4GB HBM1, because it can load things on the fly much faster. Obviously you're still limited to a total dataset size, but that isn't as big of an issue as some may think.

 

Furthermore, these were never designed as a primary gaming card. They're a hybrid dev/gaming/workstation card.

 

And anyway, they can't add more RAM. It's a limitation of the Fiji XT HBM1 architecture. It'll be fixed with Vega, which is still potentially months away.

[PCGuy_5960]

1 minute ago, dalekphalm said:

HTPC's are supposed to be small, yes, but efficient?

Yes, in the sense that you can't *properly* cool a GPU that draws more than 200W in an SFF/ITX case.... 

[dalekphalm]

8 minutes ago, PCGuy_5960 said:

Yes, in the sense that you can't *properly* cool a GPU that draws more than 200W in an SFF/ITX case.... 

Sure you can. Depends on the setup. A GPU cooled with an AIO waterblock? Actually easier to cool in an ITX case, assuming there's room to mount the Radiator.

[PCGuy_5960]

1 minute ago, dalekphalm said:

Sure you can. Depends on the setup. A GPU cooled with an AIO waterblock? Actually easier to cool in an ITX case, assuming there's room to mount the Radiator.

Yes, but there is no room in something like the Node 202 or the A4 SFX for a radiator..... If there was, cooling wouldn't be a problem...

[dalekphalm]

7 minutes ago, PCGuy_5960 said:

Yes, but there is no room in something like the Node 202 or the A4 SFX for a radiator..... If there was, cooling wouldn't be a problem...

Sure but those aren't your only options. If you want something that compact and small, then yes, you're limited to ITX and a really slim/short GPU. But an HTPC can be larger. Hell the original "HTPC's" that came out were Mid-tower ATX computers lol. Obviously it's good that they've shrunk in size, but there's a lot of variation to choose from. I have an mATX HTPC case that can take full height GPU's (though only short-ish ones) and could definitely take at least one 120mm rad, possibly two.

[goodtofufriday]

14 minutes ago, PCGuy_5960 said:

Yes, but there is no room in something like the Node 202 or the A4 SFX for a radiator..... If there was, cooling wouldn't be a problem...

You can always mount it on the outside =P That was an extremely common practice until recently. 

[Misanthrope]

1 hour ago, dalekphalm said:

This isn't new...

 

Also, because of the increased VRAM bandwidth, some benchmarks have shown that it doesn't get held back as much by the 4GB HBM1, because it can load things on the fly much faster. Obviously you're still limited to a total dataset size, but that isn't as big of an issue as some may think.

 

Furthermore, these were never designed as a primary gaming card. They're a hybrid dev/gaming/workstation card.

 

And anyway, they can't add more RAM. It's a limitation of the Fiji XT HBM1 architecture. It'll be fixed with Vega, which is still potentially months away.

AFAIK that test was for a single Fiji chip, the use case for crossfire Fiji should be 4k thus, more vram switching cause of the added power could change things drastically. 

 

So I need to see Fiji crossfire doing 4k and pushing far above the 4gb vram limit. 

[dalekphalm]

10 minutes ago, Misanthrope said:

AFAIK that test was for a single Fiji chip, the use case for crossfire Fiji should be 4k thus, more vram switching cause of the added power could change things drastically. 

 

So I need to see Fiji crossfire doing 4k and pushing far above the 4gb vram limit. 

Agreed - but I would like to see benchmarks to confirm either way before just randomly decrying "4GB ISN'T ENOUGH ON FIJI!" with zero proof

[Doobeedoo]

While Vega is not here yet I wonder what a dual one would end up being.

[Agosto]

4 hours ago, dalekphalm said:

This isn't new...

 

Also, because of the increased VRAM bandwidth, some benchmarks have shown that it doesn't get held back as much by the 4GB HBM1, because it can load things on the fly much faster. Obviously you're still limited to a total dataset size, but that isn't as big of an issue as some may think.

 

Furthermore, these were never designed as a primary gaming card. They're a hybrid dev/gaming/workstation card.

 

And anyway, they can't add more RAM. It's a limitation of the Fiji XT HBM1 architecture. It'll be fixed with Vega, which is still potentially months away.

I know, I know. But 4GB is still not enough for UHD, even if it's fast RAM 

[AlTech]

8 hours ago, Agost said:

Still 4GB only for GPU... makes no sense

They technically couldn't make it any more.

 

4GB was the maximum limit for HBM1.

[Agosto]

43 minutes ago, AluminiumTech said:

They technically couldn't make it any more.

 

4GB was the maximum limit for HBM1.

Again

 

I KNOW

 

It still makes no sense for consumers which want to use it for gaming at high resolution, since it will be seriously vram limited.

[Bleedingyamato]

On January 19, 2017 at 0:52 PM, Mr.Meerkat said:

Me want...agh, can't decide whether I want a dual GPU beast again in my life or a efficient single GPU  (to be honest, my RX470 is serving me perfectly fine so...I'm actually tempted to not upgrade to Vega anymore and just get a 500 quid more expensive car )

Is 500 that much difference?

 

 

Is a dual GPU thing like this different from SLI/Crossfire?  

[Mr.Meerkat]

1 minute ago, Bleedingyamato said:

Is 500 that much difference?

 

 

Is a dual GPU thing like this different from SLI/Crossfire?  

When you have like a budget from anywhere of £1000-4500~, yes  

[Prysin]

5 minutes ago, Bleedingyamato said:

Is 500 that much difference?

 

 

Is a dual GPU thing like this different from SLI/Crossfire?  

its 2x Fury Nano + a AIO... a Nano still costs like 350-400$ each...

 

Technically not a difference between SLI or Crossfire. BUT, they have onboard PLX chips to have a physical PCIe interconnect between the PCIe connector on the board and the cores and between the cores themselves.

[leadeater]

1 hour ago, AluminiumTech said:

They technically couldn't make it any more.

 

4GB was the maximum limit for HBM1.

Well they could have made the memory interface on the GPU larger and doubled the number of HBM stacks to get 8GB of vRAM. It wasn't a technical limit it was a sanity limit as to accommodate the use of more HBM stacks at the time would have made the total card price so insanely high none would have been sold.

 

A technical limit is something like the maximum addressable memory space in a 32bit application. AMD imposed a design limit which is self chosen rather than a technical limit which is an impossibility to exceed due to factors outside of your control.

[Prysin]

8 hours ago, dalekphalm said:

Agreed - but I would like to see benchmarks to confirm either way before just randomly decrying "4GB ISN'T ENOUGH ON FIJI!" with zero proof

there was tests like that when Fiji came out... Fiji, due to the insane bandwidth of HBM and the extremely wide GPU pipeline, doesnt crap out before you hit 5K resolution....