Radeon RX 5700 (NAVI) AIB customized cards available in August

[CarlBar]

On 6/20/2019 at 2:23 PM, leadeater said:

 

It's probably something like a static pressure problem with the type of fans required for blower coolers. If you go longer or denser fins they might not be able to push the air through. Even the server GPUs that are passive large heat sinks that use the case airflow don't use all that high fin density.

 

From memory of stuff i've read i'd say it's a physics issue. Blower fans are effectively a form of centrifugal fan  and the problem there is that whilst they're great at getting static pressure they need fairlyu high RPM's to be efficient and their max intake velocity is actually limited. In addition they exhaust air equally in all directions so unless you've got a square heatsink or you use some really fancy, (and thus expensive), internal shaping a lot of the air is exhausting in the wrong direction leading to turbulent flow and different bits of air ejected by the fan getting in each others way. 

 

The dent to my knowledge should actually help in several ways, but blowers are really only remotely useful if you need high static pressure in a compact form factor, so outside of single exhaust outlet designs they don't make a lot of sense.

[maartendc]

RX 5700???

 

What the hell is wrong with the marketing people coming up with these names?!

 

I understand you don't want to call it RX680 or whatever, because it is a completely new architecture. You cant call it "Vega" something, since it is not Vega architecture.

 

But in the same logic as Vega, call it Navi 56, Navi 64, or whatever. Or just start a brand new number scheme, The "Radeon N 1000."  (N for Navi)

 

Why on earth is it the number 5700??? Why??? This is even worse than Nvidia with their GTX 1650, which at least was a number between 10 and 20. 

[Trixanity]

29 minutes ago, maartendc said:

RX 5700???

 

What the hell is wrong with the marketing people coming up with these names?!

 

I understand you don't want to call it RX680 or whatever, because it is a completely new architecture. You cant call it "Vega" something, since it is not Vega architecture.

 

But in the same logic as Vega, call it Navi 56, Navi 64, or whatever. Or just start a brand new number scheme, The "Radeon N 1000."  (N for Navi)

 

Why on earth is it the number 5700??? Why??? This is even worse than Nvidia with their GTX 1650, which at least was a number between 10 and 20. 

Because naming schemes based on the number of compute units are stupid when seen in relation to the chip and versions of it. It's better to name them after the performance level within the stack. 

 

AMD has done that and the naming scheme is pretty much the same as Nvidia's. So if this sucks then they all suck.

 

1650 is not a number between 10 and 20 by the way.

[DrMacintosh]

Can’t wait to see the Sapphire cards. 

[Froody129]

On 6/20/2019 at 3:23 PM, leadeater said:

 

It's probably something like a static pressure problem with the type of fans required for blower coolers. If you go longer or denser fins they might not be able to push the air through. Even the server GPUs that are passive large heat sinks that use the case airflow don't use all that high fin density.

I'd imagine the curve in the shroud must have something to do with it. It's ugly, so I don't see them including it otherwise

[Froody129]

1 hour ago, maartendc said:

RX 5700???

 

What the hell is wrong with the marketing people coming up with these names?!

 

I understand you don't want to call it RX680 or whatever, because it is a completely new architecture. You cant call it "Vega" something, since it is not Vega architecture.

 

But in the same logic as Vega, call it Navi 56, Navi 64, or whatever. Or just start a brand new number scheme, The "Radeon N 1000."  (N for Navi)

 

Why on earth is it the number 5700??? Why??? This is even worse than Nvidia with their GTX 1650, which at least was a number between 10 and 20. 

As I understand it it's reference to the naming scheme that ATI used before AMD bought them. 

[maartendc]

1 hour ago, Trixanity said:

 It's better to name them after the performance level within the stack.

Right, so 5700 because its a larger number than 590?

 

You're delusional.

 

[Trixanity]

17 minutes ago, maartendc said:

Right, so 5700 because its a larger number than 590?

 

You're delusional.

 

Not necessarily but it's an entirely new product family so it makes sense to distance itself from the old crap. 5700 is arbitrary but supposedly far enough away from Nvidia and is symbolically linked to the 50 year anniversary. Additionally, calling it 590 would be indicative of it being in the Polaris family and calling it 690 (or any variant of 90) would be indicative of it being at the top of the stack which this supposedly isn't.

 

Calling me delusional only highlights your poor grasp of the subject and goes to show you have nothing to offer the discussion. 

[CarlBar]

1 hour ago, Froody129 said:

I'd imagine the curve in the shroud must have something to do with it. It's ugly, so I don't see them including it otherwise

 

The shroud dent restricts flow through that part of the cross section, this causes the pressure to rise on one side, and then because the flow channel expands, fall on the other, that means the air moving into the neck moves slower whilst as it leaves the increase in volume means that the velocity must increase.

 

Given what i mentioned prior about centrifugal fan designs being good at static pressure this actually leverages the fan designs strengths as it has a relatively easy time generating the higher pressure on the high pressure side without losing as much flow rate as a traditional fan design. ironically if allowed to operate at higher RPM's it may well be the case that it would cool well for a blower. Albeit be loud.

[CarlBar]

Just dug up the wikipedia article on centrifugal fans to cross check my memory and i pretty much hit it on the head. Linky for those interested. https://en.wikipedia.org/wiki/Centrifugal_fan 

[leadeater]

8 hours ago, CarlBar said:

From memory of stuff i've read i'd say it's a physics issue. Blower fans are effectively a form of centrifugal fan  and the problem there is that whilst they're great at getting static pressure they need fairlyu high RPM's to be efficient and their max intake velocity is actually limited. In addition they exhaust air equally in all directions so unless you've got a square heatsink or you use some really fancy, (and thus expensive), internal shaping a lot of the air is exhausting in the wrong direction leading to turbulent flow and different bits of air ejected by the fan getting in each others way. 

 

The dent to my knowledge should actually help in several ways, but blowers are really only remotely useful if you need high static pressure in a compact form factor, so outside of single exhaust outlet designs they don't make a lot of sense.

They really aren't that good for static pressure, around 1-1.3 mmH₂O where as for a 120mm fan anything below 2.5 is considered rather bad for that usage (case fans don't need to be as high). They are even worse when you take in to account the RPM and RPM curve pressure performance, blower fans really are not good.

 

These are also high performance ones, don't think AMD is putting in 10W-17W fans.

Edit: God damn freedom units, more around 5-10 mmH₂O in SI (real units, fight me )

[CarlBar]

29 minutes ago, leadeater said:

They really aren't that good for static pressure, around 1-1.3 mmH₂O where as for a 120mm fan anything below 2.5 is considered rather bad for that usage (case fans don't need to be as high). They are even worse when you take in to account the RPM and RPM curve pressure performance, blower fans really are not good.

 

 

 

These are also high performance ones, don't think AMD is putting in 10W-17W fans.

 

How do you boop that up. Seriously that style of fan has all the advantages when it comes to pressure generation...

[leadeater]

3 minutes ago, CarlBar said:

 

How do you boop that up. Seriously that style of fan has all the advantages when it comes to pressure generation...

See the edit . Industry standard units exist for a reason, stupid Comair lol. Anyway you could get similar pressure gains by using a reduction baffle much like blower fans use. But you wouldn't be able to do that on a 80mm-120mm standard fan for a GPU so....

[leadeater]

4 minutes ago, CarlBar said:

How do you boop that up. Seriously that style of fan has all the advantages when it comes to pressure generation...

Either way the issue is still the same, blower fans don't have the pressure required to push the air though such a long air channel with such flow restrictions. That's a big limitation of them and why fin density is a bit lower and why they don't make the heatsink longer than they do even though there is technically enough physical space as it is.

 

Any longer or denser fins and you have to push harder which increases noise and we already complain about that as it is.

[CarlBar]

11 minutes ago, leadeater said:

See the edit . Industry standard units exist for a reason, stupid Comair lol. Anyway you could get similar pressure gains by using a reduction baffle much like blower fans use. But you wouldn't be able to do that on a 80mm-120mm standard fan for a GPU so....

 

No worries. I was busy going WTH in my head a lot though but was busy with somthing else so didn't look too hard at the specs, took your word on what they said :).

 

p.s Yes SI is best :D.

[leadeater]

2 minutes ago, CarlBar said:

No worries. I was busy going WTH in my head a lot though but was busy with somthing else so didn't look too hard at the specs, took your word on what they said :).

Yea I was looking at it going "How TF does a 10W fan still only get 1mmH2O?", went over to Delta which has the units in both SI and Imperial and yea.... fml lol.

[GOTSpectrum]

On 6/20/2019 at 2:23 PM, leadeater said:

 

It's probably something like a static pressure problem with the type of fans required for blower coolers. If you go longer or denser fins they might not be able to push the air through. Even the server GPUs that are passive large heat sinks that use the case airflow don't use all that high fin density.

Ita not only this, but then you have airflow that has space between the fan and the fin array you get back pressure and currents/vortexes in the space between. This gets worse as fun density increases. This causes both lower performance and higher noise. The way around this is to increase pressure. But the higher pressure can only really be made by increasing airflow with radial fans and this increases the vortex problem even further. 

 

A good example of how this turbulence and vortexes affect noise is a standard hairdryer Vs the laminar flow 'Dyson supersonic' hair dryer.