AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[maddie]

Also, the fins are perpendicular to the air flow, that makes a lot of incongruities, really, and the perspective is not at play if we take the closer parralel holes...

Anyway thank you for getting out of a long sleep to provide us thoses precisions, really the effort is much appreciated..

How do you get that?

If you rotate the heatsink on the y = -x axis, it lines up with the fins lined up with the exhaust vents

 

[Abwx]

How do you get that?

If you rotate the heatsink on the y = -x axis, it lines up with the fins lined up with the exhaust vents

The fins are visible on the bottom of the pic, they are perpendicular to the direction of the air flow while they should be parrallel, there s no way to correct the thing than by rotating 90°, that it is on the wrong side on the pic change nothing since a 180° rotation on the Z axys will keep the other oientations unchanged.

Btw i measured the holes distance on my screen, the difference is too big to be explained by either perspective or any other optical effect.

 

[MrTeal]

Also, the fins are perpendicular to the air flow, that makes a lot of incongruities, really, and the perspective is not at play if we take the closer parralel holes...

Anyway thank you for getting out of a long sleep to provide us thoses precisions, really the effort is much appreciated..




As said the orientation of the fins is not logical at all, if the cooler is rotated 90° to make the fins parralel to the air flow then the holes distance is even more unrealistic.

It must be rotated 90 degrees. I believe it is a perspective issue; it's also the reason the hole spacing looks non-symmetrical when other images show a square hole pattern.

 

[oussama-tn]

Do you guys think we will be seeing water cooled version from aibs ? mehh that would be awesome for quite/cool crossfire.

 

[DeeJayBump]

Btw i measured the holes distance on my screen, the difference is too big to be explained by either perspective or any other optical effect.

Unfortunately (for your "measurements") the items in question (heatsink/mounting holes/etc) exist in real life rather than in a picture on a screen.

So, in spite of your "measurements" there may be no difference at all other than the one you are "measuring" and imagining.

 

[Armsdealer]

Because 11% overclock done right increase power by 1.11^2 = 1.23x, when i say done right it means that voltage margin is preserved..

I think the formula is P = c*V^2*f

ie assuming constant voltage power usage scales linearly with frequency.

However, if you are voltage limited, ie you need more volts to increase the frequency, Voltage and frequency also scale linearly

V= b*f

Thus as you wish to increase the clock above its voltage constraints, power scales as the cube of frequency.

From this we can conclude that 150w will likely be insufficient above 1500mhz (1500/1266)^3 * 100w. I personally will not buy a 6 pin card.

 

[Abwx]

Unfortunately (for your "measurements") the items in question (heatsink/mounting holes/etc) exist in real life rather than in a picture on a screen.

So, in spite of your "measurements" there may be no difference at all other than the one you are "measuring" and imagining.

Estimation of die sizes were made using such pic end ended being quite
accurate.

So unless you have an actual sample at hand my measuements
have at least the merit to be based on available pic rather than general statements that one can be wrong, granted i could happen, but then provide you own estimations and why they are such..

 

[3DVagabond]

Looks to have a copper core. Aluminum dissipates heat faster, but copper conducts it faster, so using both together usually yields better performance than just one or the other.

A vapor chamber may be present, but would not be needed to dissipate 150W of heat and maintain <80C with decent airflow and ambient temperatures.

I'm just thinking that they may have finally put a quiet blower on a reference card. I think that would be fantastic.

 

[Abwx]

I think the formula is P = c*V^2*f

ie assuming constant voltage power usage scales linearly with frequency.

However, if you are voltage limited, ie you need more volts to increase the frequency, Voltage and frequency also scale linearly

V= b*f

Thus as you wish to increase the clock above its voltage constraints, power scales as the cube of frequency.

From this we can conclude that 150w will likely be insufficient above 1500mhz (1500/1266)^3 * 100w. I personally will not buy a 6 pin card.

V = b.sqrt(f)

Voltage is proportional to the square root of frequency, so power increase as a square of fequency.

 

[DeeJayBump]

Die size measurements are made from comparison of objects on the same PCB area, NOT from objects of two different thicknesses and different distances from each other.

If the PCB and the heatsink/mounting holes were the same relative distance from view, then your comparison would have merit.

For example, look at the GPU die on the right, then look at the area of the heatsink where the thermal paste contacted/covered the GPU die on the left. On the heatsink, that area appears distinctly larger on the heatsink/thermal paste area versus the ACTUAL SIZE of the GPU die on the right.

Why? Because the heatsink is closer to view, making it APPEAR LARGER, when it is not.

 

[Abwx]

Die size measurements are made from comparison of objects on the same PCB area, NOT from objects of two different thicknesses and different distances from each other.

If the PCB and the heatsink/mounting holes were the same relative distance from view, then your comparison would have merit.

The distance between the heatsink and the PCB is much smaller than the distance between either these points and the camera sensor, so the magnitude of the error is as small as the ratio of thoses respective distances...

 

[3DVagabond]

Also, the fins are perpendicular to the air flow, that makes a lot of incongruities, really, and the perspective is not at play if we take the closer parralel holes...

Anyway thank you for getting out of a long sleep to provide us thoses precisions, really the effort is much appreciated..




As said the orientation of the fins is not logical at all, if the cooler is rotated 90° to make the fins parralel to the air flow then the holes distance is even more unrealistic.

The cooler is simply rotated 90°. Obviously not a case for Sherlock Holmes.

 

[3DVagabond]

Do you guys think we will be seeing water cooled version from aibs ? mehh that would be awesome for quite/cool crossfire.

I think it was on [H] forum (maybe?) the XFX rep stuck his head in and asked what people though of an AIO 480 for <$300. So, maybe they are planning one? Or at least thinking about it?

Why would they put a cooler in the photo that wasn't for the card? People need reasons to do things. If they simply didn't want to show us the cards heat sink, don't. No need to put some fake cooler in the photo.

 

[Abwx]

The cooler is simply rotated 90°. Obviously not a case for Sherlock Holmes.

That s possible, anyway we ll soon know exactly what it is...

 

[Armsdealer]

V = b.sqrt(f)

Voltage is proportional to the square root of frequency, so power increase as a square of fequency.

I am no expert on this, and you very well might be, so I'll go with your word, but I think there's some information here:

http://physics.stackexchange.com/questions/34766/how-does-power-consumption-vary-with-the-processor-frequency-in-a-typical-comput

In particular the answer at the bottom is quite detailed and claims that for older processes, the scaling was as you described, but in newer processes it comes closer to f^3.

Edit: I'm wrong - my answer is the outdated one.

The correct answer seems quite nuanced though, and not nearly as simple as before. Would love to get more info on this if possible

 

[sirmo]

I think it was on [H] forum (maybe?) the XFX rep stuck his head in and asked what people though of an AIO 480 for <$300. So, maybe they are planning one? Or at least thinking about it?

It was on reddit. And yeah it was an XFX guy: https://www.reddit.com/r/pcmasterra...orking_on_a_gpu_design_and_i_want_your_input/

 

[3DVagabond]

It was on reddit. And yeah it was an XFX guy: https://www.reddit.com/r/pcmasterra...orking_on_a_gpu_design_and_i_want_your_input/

That could be it and I was wrong on the price. Or he could have made another post elsewhere? Thanks though for finding it.

 

[Abwx]

I am no expert on this, and you very well might be, so I'll go with your word, but I think there's some information here:

http://physics.stackexchange.com/qu...h-the-processor-frequency-in-a-typical-comput

In particular the answer at the bottom is quite detailed and claims that for older processes, the scaling was as you described, but in newer processes it comes closer to f^3.

Edit: I'm wrong - my answer is the outdated one

Of course i m talking of the most favourable range wich is the one used by all designs, eventualy parts of the curve wich are at a 2.5 power are used to boost one core in CPUs, but so far GPUs seems to be exploited in the efficient range only.

The article you linked provide some good explanation but it miss the fact that current within a fet increase as a square of voltage while he seems to assume that it s proportional, he confuse the square law of the mosfets with the parameters weights in the equation of the power in function of several variables that are frequency, capacitance and voltage.

 

[desprado]

Another troll and irresponsible hype from Red team. AMD is comparing their card with R9 390 and yet some people are dreaming and hyping it to on par with GTX 1070.

 

[VulgarDisplay]

Another troll and irresponsible hype from Red team. AMD is comparing their card with R9 390 and yet some people are dreaming and hyping it to on par with GTX 1070.

I often think English isn't your first language but I think it's against the rules to call out your own post as "another troll".

Sent from my XT1575 using Tapatalk

 

[desprado]

I often think English isn't your first language but I think it's against the rules to call out your own post as "another troll".

Sent from my XT1575 using Tapatalk

480 is for VR entry with a good performance. It is on par with R9 390 on stock and even if it OC then still it is no way near GTX 1070. It cannot be OC like people are hyping it and whole purpose of the card is demolished by it. Its main goal is performance per watt.

 

[guskline]

are you going with crossfire again, 2x rx480?

NOPE! Went to a single GTX 1080FE in my 5960x rig and moved my GTX 980TI SC to the 4790k rig.

 

[ozzy702]

People who are comparing 480 with GTX 1070 are insulting tech and human intelligence.

480 is for VR entry with a good performance. It is on par with R9 390 on stock and even if it OC then still it is no way near GTX 1070. It cannot be OC like people are hyping it and whole purpose of the card is demolished by it. Its main goal is performance per watt.

We really don't know how much it can be overclocked and how much performance is left on the table before power consumption skyrockets. Of course performance/watt will suffer with overclocking but it's not always by a hideous amount. 7850 overclocked like a boss and didn't suck down a lot of juice.

I don't see 1500mhz as being unbelievable given FinFet and 14nm. Higher than that, yeah, that's probably a stretch but I could see 1500mhz being the upper ceiling.

Really, I don't care one iota if power consumption goes up a few watts if I can get 1070 performance for $200-$300. If that's the case I'll buy Polaris to replace my shitty GTX 780 which I'll turn around and sell for $150.

 

[oussama-tn]

Need some advice. I found a deal for my 9months old r9 390 nitro (265euros), although they sell for 280euros in europe, in my country they are still expensive. Do you guys think i should sell it and get an rx 480 ? if the leaks are true do you think the polaris 10 will be close the the 980ti in 1440p ? Mehh if i wait longer i will not find a better deal in the futur but at the same time i don't want to get f*** up if the rx 480 isnt as good as people are predicting...

 

[Maverick177]

It is on par with R9 390 on stock

Where do you get this kind of information? Overhyping is one thing but to deliberately put down the card to this level is disgusting.