Review AMD RX 480 By Javisoft

[Javisoft]

Update with total consumption, 6-pin connector and pciex slot.

[garfield]

I see that the 6-pin connector consumption is over 100w, when it's supposed to not pull more than 75, and that the consumption on the pci-e is 75 and several times during two seconds exceeds 82w. I know that the tolerance is up to 86w, and I suppose it's sustained, but the data I see on consumption has me confused. Not because of an error in your measurements, but because AMD has released that graph with those peaks. The VRM are going to go to hell...

[Javisoft]

I don't understand that such a beastly variation, but with a power meter in hand and duplicate power consumption I can't help but think that it is due to the slight throttling it suffers from the temps and that it tries not to collapse the connector since it abuses this much more than the pcie slot.

I will take data again with the fan at full to avoid the frequency dance and above all to lower the temperature ;D.

Goodbye!!!

Edit: I am uploading more data as I process it, so throughout the night I hope to give it life XD.

[Espinetenbolas]

@Javisoft:

I can't help but think that it's due to the slight throttling it suffers from the temps and that it tries not to collapse the connection

The throttling is not because of the temperature… it's because of the consumption.

You do more throttling against higher temperatures because the VRM conversion performance decreases with the increase in temperature... also at higher temperatures the GPU consumes more since it alters the conductivity of the silicon and the metals that make up the chip.

Consequently the TDP skyrockets above 150W and therefore lowers the Mhz.

That is to say it is not due to thermal causes but to the BIOS... if you modify the TDP in some way you should disappear that throttling either with a ModBios or with the MSI AfterBurner increasing the PowerLimit.

The old Hawaiians suffered from this ailment that depending on the application or game is noticed more or less... if you put the OCCT or the Furmark you'll see how it happens the second you run the test even though the graphics card is as cold as a pozero's ass.

When you put water you'll see that you have the same problem and you only solve it by modifying the TDP.

[Javisoft]

@Espinetenbolas:

Throttling is not due to temperature... it's due to consumption.

It makes you throttle more as the temperature rises because the VRM conversion efficiency decreases with increasing temperature... also, at higher temperatures, the GPU consumes more power since it alters the conductivity of silicon and the metals that make up the chip.

As a result, the TDP skyrockets above 150W and therefore lowers the MHz.

That is to say, it is not due to thermal causes but rather the BIOS... if you modify the TDP in some way, that throttling should disappear, either with a ModBios or with MSI AfterBurner by increasing the PowerLimit.

The old Hawaiis suffered from this problem, which, depending on the application or game, is noticed more or less... if you run OCCT or Furmark, you'll see it happen the second you start the test, even if the graphics card is as cold as a zero's ass.

When you put water on it, you'll see that you have the same problem and you only solve it by modifying the TDP.

The consumption peaks at 124w and averages 112w, which fluctuates a lot, yes, due to the voltage that also marks its bachata steps, but encouraged by the temperature that hovers around 90º XD, if you notice, just when the temperature exceeds 87º, the frequencies begin to drop but the voltage oscillates more or less within the same margins.

The curious thing is that what fluctuates the least, not to say that it is very, very stable, are the frequencies, it holds up like a champion.

Anyway, tomorrow I'm going to run the same tests with the fan at full blast and also measure the noise and fpm with the anemometer XD, surely we'll clear up many doubts.

Best regards!!!

[garfield]

Curious. With the fan at full blast, it draws less from the pcie and draws almost the same from the connector. In the end, it's going to be that those values are so disparate because of the poor cooling of the chip. I think the custom ones will be ahead in that matter, as has always happened to AMD with their heatsinks.
Every time I see more tests, it seems like a pretty decent graphics card for what it costs, and the time frame rate is sure to be improved with new drivers (or so I hope).

[Javisoft]

It is very curious, but the heatsink is a bit of a mess, they could have left an entrance from the back to get more air... The custom ones look good, but I fear that with the GPU at full capacity, they will see similar temperatures to what I have put with the fan at FULL (although with less noise of course), the OC will be another story XD. Best regards!!

[Javisoft]

Update with full oc data by air!!

[garfield]

I'm afraid of those pcie peaks, and the load on the 6-pin connector. I don't want it to load something from the source. But you don't even feel it.
The curious thing is that with that oc you haven't reached the consumptions of almost 200w that other pages put. It's to look at where they have failed, or in what equipment and environment it has been measured, because the consumptions that you have had are the ones I have seen in videos on youtube: between 150 and 165€ at full capacity.
The strange thing is that with those temperatures it hasn't made you throttling. Not everything is so bad in these AMD. If in the end they are going to be good and all.

[Javisoft]

I think the difference is in the ASIC, mine is almost 90% and according to what I have read, consumption is very closely linked to this in the 14 nm.

[garfield]

I've been looking, but I don't see any more asic data. What are the approximate normal asic values for these graphics?

[Espinetenbolas]

You don't have to suffer so much with the power consumption issue unless the card is pulling too much from the PCI-E because the cables can handle much more even if the power supply is a potato:

PCI Express Auxiliary Graphics Power Connectors - Power Supply 101: A Reference Of Specifications

If we talk about normal connectors, they can handle up to 8 amps and in the case of high-quality connectors, up to 12 amps.

Like I say, I don't understand why he distributed the load to 50% if, depending on the card, the consumption exceeds 150W.

[Javisoft]

I think I read about the asic topic on pcperspective, and as for the rest, the basis of GPU-z according to my asic is better than 96% of the cards that have used the program, or what is the same, there is only 4% with the same asic or better than mine.

So it would not be surprising that in some case they reach 200w without problems.

Regarding the rails and amps, logically it will depend on the source, but in my specific case each rail can hold up to 40A before the ocp trips.

If you want I can make a graph with the amps and volts per second XD.

Regards!!!

[Javisoft]

New update with photos, features and data!!!

[Fassou]

The first post was already a hit, so the next two are Summa Cum Laude

As a suggestion from the comfort of the couch :troll:, think if you might be more interested in the model of posting in parts, to improve the visibility of new contributions, and you save the trouble of reserving responses, being able to add the links to the continuations later.

Alright, alright I know... I'll shut up

[Javisoft]

Update with consumption and performance data of the Crimson 16.7.1!!!

[Javisoft]

New update with all the data under DX12 on the crimson 16.7.1, now I just have to put it through its paces and I'll finish the review!!!

[cobito]

What a beastly document you are creating :sisi:

[Javisoft]

It is the least that one should do when making a review, a complete analysis, what some websites do (that charge and dedicate themselves exclusively to it) is a bit laughable.

In my opinion, the fault lies with the reader for not demanding and swallowing such articles, but well, it doesn't rain to everyone's liking.

By the way, to say that the review has taken 81 hours of work.

Best regards!!!

[Sylver]

@Javisoft:

It's the least you have to do when you do a review, a complete analysis, what some websites do (that charge and dedicate themselves exclusively to it) is a bit laughable.

In my opinion, the fault lies with the reader for not demanding more and swallowing such articles, but well, it doesn't rain to everyone's liking.

By the way, to say that the review has taken 81 hours of work.

Best regards!!!

That's not paid... I would redefine the terms of economic supply of the forum and allocate a part as a salary for hardworking forum members of this level :troll:

Best regards!