Radeon RX 480 "Polaris" Launched at $199

[malventano]

Main issue seems to be the standard test for pci-e then and the certification.
if you send cards, they pass as they are tested then someone at the pci-e certification protocol responsability messed up.

This is definitely a valid question. It's possible that their test sample might have been right on the line or within some additional tolerance allowed by their test. Also consider that their test likely does not include overclocking.

 

[Yrkoon]

I'm looking to upgrade from an Powercolor Radeon R9 270X.

2 Things.

Firstly, is it a viable upgrade that will give a nice performance boost
Secondly, will I need to wait for DVI-DL (120hz monitor) and if so, how long.

If anyone could answer these couple of question, i owuld appreciate it .

thanks.

i can answer, i just went from the 2 gig HIS ICE Q 270x,

the sapphire 480x has about double the performance,


witcher 3 on ultra at 1200p is mostly between 50 and 60 fps and i just noticed a setting that was limiting fps so it went up to 70 in some places lol

i only have a dvi port on my monitor so i ordered a dvi to hdmi cable from a competitor and my monitor is old.

fire strike graphics card test one had about 89% over my 270x
graphics card test 2 had over 90%

same cpu 6600k skylake on both tests i ran the 270x ones about a week ago, it is a massive upgrade

my friend has a nitro 390 oc and he says the performance is similar, he lives in the spare room, eg i can walk in his room and see a 390 nitro and the 480 is on par and can only get better

http://www.hisdigital.com/un/product2-774.shtml is my old card

 

[D.P.]

Actually, that's the issue. We asked motherboard makers prior to posting our updated story on this, and they stated that sustained (average) load exceeding 75W can lead to damage. This was further confirmed when we saw the voltage droop to 11.5V from the motherboard slot while the 6-pin connector (carrying the same current) was able to maintain 12.0V.

The other examples folks are citing are the Toms reviews which are measuring in a way that can see the instantaneous current spikes resulting from the DC-DC switching power supply on (all) GPUs. A switching supply, when measured in that way, is going to potentially spike above the limit. Motherboard makers have told us that those instantaneous spikes are not the issue. The issue is sustained load from the slot, which the 480 is exceeding.

What is causing more confusion is that plenty of cards in the past have exceeded the 6/8-pin spec, but we / others don't make such a huge deal about that since that is a direct link to the +12V rail of the PSU, which can usually handle way more than the stated limits for those connectors. Motherboard traces / PCIe pins are way more limiting.

Thanks for clarifying this. That is quite parrying then.

 

[Yrkoon]

Thanks for clarifying this. That is quite parrying then.

so do i have to worry about https://www.overclockers.co.uk/giga...cket-1151-ddr4-atx-motherboard-mb-537-gi.html with a single rx480 8 gig on stock?

im guessing not?

 

[Alexrose1uk]

The original 380X was also reviewed as a 1440P when it came out. Kinda bizarre I think.

The answer is in the slide. They are using 'popular' games like League of Legends and 60FPS as the goal post.
Obviously these games arent that demanding. You have to remember this marketting is for the more average joe not the enthusiast!

 

[flopper]

The answer is in the slide. They are using 'popular' games like League of Legends and 60FPS as the goal post.
Obviously these games arent that demanding. You have to remember this marketting is for the more average joe not the enthusiast!

in any forum, you have maybe max 10 users often less than that has 90%+ of forum posts about a subject normally. so are those 10 people representative for the forum gpu or are they trolls, entusiasts or plain ignorant along the way?

If you check this thread you find that very few of the major posters have an amd card or are buying a 480 so why do they even post here or are allowed to?

 

[james.miller]

in any forum, you have maybe max 10 users often less than that has 90%+ of forum posts about a subject normally. so are those 10 people representative for the forum gpu or are they trolls, entusiasts or plain ignorant along the way?

If you check this thread you find that very few of the major posters have an amd card or are buying a 480 so why do they even post here or are allowed to?

Because they have an interest in technology all the same. They have just as much right to post and debate in any thread as you do. Owning an AMD card is not a prerequisite for posting in AMD threads. Get over yourself.

 

[Gregster]

in any forum, you have maybe max 10 users often less than that has 90%+ of forum posts about a subject normally. so are those 10 people representative for the forum gpu or are they trolls, entusiasts or plain ignorant along the way?

If you check this thread you find that very few of the major posters have an amd card or are buying a 480 so why do they even post here or are allowed to?

Some like to discuss new technology, regardless of vendor. I have seen some very silly comments which have been removed by the mods but most of what I have seen has been fair talk. It would be boring if only people buying a 480 were allowed to post here. Did you ever post in Pascal/Maxwell threads?

Edit:

A quick look in the Pascal thread and seems you posted several times.

https://forums.overclockers.co.uk/search.php?searchid=23989115

 

[0h_no_not_again]

Hi All,
First post in these here forums, but a loonnggg time lurker. Quite a speculative first post as well, but here goes...

Does anyone else get the sense AMD were expecting significantly lower leakage from the GloFo Samsung 14nm node?

The cards are quite severely limited by the power input, with a secondary thermal limitation. My understanding is that a lot of recent cards (AMD/NV) apply a power target and will throttle to meet it, but the RX 480 is so constrained that under clocking it has no impact to bench performance (as per OC3D tests).

It's as if their PCB/cooling design was based on a 125W max design target and the silicon has fallen way short. What made matters worse, was the requirement to be "VR Ready", so they have thrown power at it and presumably hope that future silicon will sort things out.

There interesting thing is the ~33W reduction in power drawn through under-volting (German reviewer). It bodes well for the RX 470, as I suspect it will be running at lower clocks/voltage. It also bodes well for when they get the 14nm node sorted out, we should see big drops in power and gains in perf.

One question, maybe for DP. Why is the card drawing so much from the PCIe +12V rail? Are the half the VRMs tied to PCIe +12V and the others to the 6-pin +12V? I guess it makes sense to do this, so you have complete control of your current draws, rather than letting I=V/R sort out where the current comes from. Anyway, who has a RX 480 that we can ghetto rig another 6 pin to the PCIe +12V power plane, bypass the shunt resistors (for measuring current draw) and see where it goes?

 

[Meaker]

AMD and Intel create CPUs with 125W limits yet are fine pushing 250W through the motherboard, vrms and pins let alone the CPU itself. Rated at, and can run at, are very different.

Likewise plenty of people overclocked the 750ti without pci-e connector and used well beyond 75W... it's a non issue.

Also a droop of 11.5v on the slot makes no difference, vrms aren't only capable of dropping 12v precisely down to the required voltage, on cpu, gpu or anything else, vrms work in a range, as do psus, as does everything electrical. Incredibly standard on most electrical items is input being +/- 10% of the rated amount.


Saying sustained load exceeding 75W can cause damage is confirmed by seeing a voltage droop is insanely bad writing. In what way did that confirm damage... it didn't. Vdroop is not a sign of damage nor an indication that damage will occur.

11.5V is within ATX specifications exactly

 

[flopper]

Did you ever post in Pascal/Maxwell threads?

If memory serve, one post a long time ago
Normally dont read nvidia threads at all on any forum.
Not interested in a lying ceo and company that try to limit the pc gaming field selling bad cards and knowing they did that burned a load of computers and along the way sells 3.5gb cards as 4gb cards.
such behaviour shouldnt be supported anywhere.

If everyone accept bad stuff its not bad stuff at some point and you become corrupted thinking its normal. Hatecrimes for example after brexit raised by 57% in the UK. You guys have issues in your country.

I enjoy my amd cards I had and have, I find their way a better one as I look at the overall usage I have for gaming and amd fullfills what I do with a computer and I can be ensured I at least can live with my decision choosing a sensible manufacturer. (you did ask)

I am happy for ocuk selling 1000+ cards in a couple of days.
as far I can tell a really good seller of cards here whatever brand.

 

[james.miller]

11.5V is within ATX specifications exactly

only just, 11.4v is the bottom limit.

 

[Gregster]

If memory serve, one post a long time ago
Normally dont read nvidia threads at all on any forum.
Not interested in a lying ceo and company that try to limit the pc gaming field selling bad cards and knowing they did that burned a load of computers and along the way sells 3.5gb cards as 4gb cards.
such behaviour shouldnt be supported anywhere.

If everyone accept bad stuff its not bad stuff at some point and you become corrupted thinking its normal. Hatecrimes for example after brexit raised by 57% in the UK. You guys have issues in your country.

I enjoy my amd cards I had and have, I find their way a better one as I look at the overall usage I have for gaming and amd fullfills what I do with a computer and I can be ensured I at least can live with my decision choosing a sensible manufacturer. (you did ask)

I am happy for ocuk selling 1000+ cards in a couple of days.
as far I can tell a really good seller of cards here whatever brand.

You should check my edit lol - Seems you posted a lot more than 1 time but that's fine with me, as we live in a democratic society and are allowed an opinion. Quite surprised you said what you did, when you clearly have no intention of buying NVidia but posted in the Pascal thread many times but here you say that anyone not buying AMD shouldn't be allowed to post in this thread. That is known as a dictatorship and thankfully we don't live by those rules.

Enjoy your RX 480 and interested in what you think of it and how it compares to your Fury.

 

[Orangey]

Remember this?



Watt went wrong?

 

[Gregster]

Remember this?

[/img]

Watt went wrong?

It appears AMD lied from that slide. The 480 uses more power than the 1070 but more concerning is the ability to control watts on bigger GPUs. If they are struggling this much with a small chip, I hope they can get it down with a bigger chip.

 

[Meaker]

only just, 11.4v is the bottom limit.

Yeah, looks like motherboard makers could do better on this front.

 

[Sacrifice]

It appears AMD lied from that slide. The 480 uses more power than the 1070 but more concerning is the ability to control watts on bigger GPUs. If they are struggling this much with a small chip, I hope they can get it down with a bigger chip.

Could be a 470. (It was on medium preset). Remember efficiency for larger chips should get better. Relatively speaking, components such as audio/memory controller are the same, and take up relatively more consumption in lower power rated cards

 

[0h_no_not_again]

Yeah, looks like motherboard makers could do better on this front.

The motherboard isn't where the droop is coming from though. It would take hundreds of amps to get a +12V power plane to considerably droop under DC loads.

It's the PCIe connector to edge connector interface that is the issue. The contacts are tiny, and while gold plated, tend to get dirty after a few insertions. One of the writers confirmed the 12+ supply was not drooping, and the +12V plane is good for 0.1v droop at most.

Just for reference, the PCIe spec (as of 2010 - sorry don't have the most recent) stipulates +12V ±8%, which is 11.04V to 12.96V @ the card.

Interestingly, @ 75W and 11.4V, you're burning 0.5V @ ~6.5A, giving 3W of heat in the connector! That's a lot! Look at the size of a 3W SMT resistor - they are huuggeee

EDIT: Used a PCB calc for a +12V "trace" to drop 0.1V under 6.5A of load over 100mm length (assuming normal copper weights, +15C temp rise). 4mm wide. A power plane would be 50mm average area after it's been cut by vias as a minimum

 

[GizF1]

Hurry up AIBs!

 

[Gregster]

Could be a 470. (It was on medium preset). Remember efficiency for larger chips should get better. Relatively speaking, components such as audio/memory controller are the same, and take up relatively more consumption in lower power rated cards

Yer fair point.