Who bought/is buying a Radeon RX 480?

[Arachnotronic]

OK so this is definitely not a 2560x1440 card; just tried Doom with the RX 480 and it dipped too much into the 40fps range for my taste. Turning down the resolution to 1920x1080 kept it at >60fps for the majority of the time.

If you are gaming at 2560x1440, or are gaming at 1920x1080 144Hz get a GTX 1070/1080 or a Radeon R9 Fury X if you prefer AMD. RX 480 is a solid 1080p/60Hz card.

 

[Silverforce11]

OK so this is definitely not a 2560x1440 card; just tried Doom with the RX 480 and it dipped too much into the 40fps range for my taste. Turning down the resolution to 1920x1080 kept it at >60fps for the majority of the time.

If you are gaming at 2560x1440, or are gaming at 1920x1080 144Hz get a GTX 1070/1080 or a Radeon R9 Fury X if you prefer AMD. RX 480 is a solid 1080p/60Hz card.

Ultra?

 

[Arachnotronic]

Ultra?

Ultra + Nightmare settings on the few settings that go that high. No AA. I have definitely seen FPS dip below the minimum shown there, although this game has some scenes that are performance hogs (i.e. My GTX 1080 will go to ~65fps).

 

[Silverforce11]

Diminishing returns on the graphics settings in play here. Just like Mirror's Edge Ultra vs Hyper. Enjoy your 40% frame rate loss for almost no gain in image quality. Wow.

Great video that highlights this diminishing returns: https://www.youtube.com/watch?v=HmsoIIMGZ00 <- Around 9 min mark is the side by side comparison.

 

[tential]

Waiting on an AIB variant.

I am borderline in terms of this being an upgrade, I have a reference 290 back from 2013, so a stock rx 480 should still be a slight boost compared to that, but I'd rather get a bit more with a higher clocking variant. I chose not to wait when I got the reference 290, not doing that again.


I consider this a tide me over card until vega arrives. I'm still using one of the original korean 1440p monitors, no freesync. My next upgrades will be:


480 AIB

new core system:
zen 8 core / 16 thread cpu + new MB + 16GB ram + 1TB SSD

Vega gpu
40-43" 4k monitor/tv - 120Hz with dp 1.4 and freesync, ideally with HDR



The monitor is the biggest unicorn item though, NO one seems to be focusing on 40-43" monitors aside from the freaking korean off brand panel makers. It's kind of ridiculous. I don't even want to bother trying to wait for an oled to get sub 1k for the display, a VA will have to do, with some local dimming. Not ideal, but lg has not even bothered to stick a freaking dp 1.2 on their 5000 dollar oled tvs. Again, everyone assumes the biggest monitor any pc gamer wants is 27 inches, or 34 if its a widescreen.

What about the 50+ inch monitors? You only have Wasabi mango with free sync there I was hoping for more competition but it feels like there won't be much at all which is a shame as I was hoping for a 65 inch free sync monitor

 

[catboy]

I was on the fence about buying an RX 480, on the condition that someone makes one without the crappy blower cooler.

But its failure to meet PCI-E spec is now pushing me off of the fence. I'd rather not take any risk of having my hardware fried and/or degraded over time due to that issue, no matter how much AMD and/or its fans downplay that risk.

I'm also pushed off of the fence because the market in Canada is being price gouged on the RX 480. It costs way more than it should based on the fair exchange rate of USD to CAD. I'm so fed up with being ripped off in such a way that if I do buy an RX 480, I'll import it from the USA rather than give my business to price-gouging Canadian sellers (which at the moment seems to be all of them). I don't even care if doing that will cost me more money after shipping and currency conversion costs are factored in. Taking those hits will still be worth it in order to keep my money out of any scammy price gouger's hands.

The only way I'll buy an RX 480 now is if:

1. A board partner makes an RX 480 that is guaranteed never to violate the PCI-E spec and has a high quality cooling solution.

2. The NVIDIA 1060 GTX ends up being a worse value for the dollar than is the RX 480.

Otherwise, if both of those conditions are not met, then I'll buy an NVIDIA GTX 1060 from a board partner that uses a high quality cooling solution instead of the RX 480.

 

[tential]

Why would you ever settle for a low quality cooler. The cooler is literally one of my most important purchasing points

 

[3DVagabond]

OK so this is definitely not a 2560x1440 card; just tried Doom with the RX 480 and it dipped too much into the 40fps range for my taste. Turning down the resolution to 1920x1080 kept it at >60fps for the majority of the time.

If you are gaming at 2560x1440, or are gaming at 1920x1080 144Hz get a GTX 1070/1080 or a Radeon R9 Fury X if you prefer AMD. RX 480 is a solid 1080p/60Hz card.

Actually I'd rather reduce AO or GI, or shadow quality, than resolution. Far less loss of IQ.

 

[R0H1T]

The only way I'll buy an RX 480 now is if:

1. A board partner makes an RX 480 that is guaranteed never to violate the PCI-E spec and has a high quality cooling solution.

2. The NVIDIA 1060 GTX ends up being a worse value for the dollar than is the RX 480.

Otherwise, if both of those conditions are not met, then I'll buy an NVIDIA GTX 1060 from a board partner that uses a high quality cooling solution instead of the RX 480.

Don't see that being guaranteed, even certain Nvidia cards have gone over spec on Toms, however the 8pin will ensure that the PCIe slot doesn't get hammered, with current & heat, over time. I'll bet some of the AIBPC will be more efficient, faster & cooler than the 480, much like the 290/x :sneaky:

That's (virtually) guaranteed but even if they are priced closer to the MSRP of 480 they'll be worse value IMO because of 1/2 GB less VRAM.

 

[Bacon1]

Diminishing returns on the graphics settings in play here. Just like Mirror's Edge Ultra vs Hyper. Enjoy your 40% frame rate loss for almost no gain in image quality. Wow.

Great video that highlights this diminishing returns: https://www.youtube.com/watch?v=HmsoIIMGZ00 <- Around 9 min mark is the side by side comparison.

Excellent video, really sad how Ultra vs High on GTA V cuts framerates in HALF on his 980 TI yet there is no visual IQ differences. Wish he showed the FPS difference for all of the games.

I was on the fence about buying an RX 480, on the condition that someone makes one without the crappy blower cooler.

But its failure to meet PCI-E spec is now pushing me off of the fence. I'd rather not take any risk of having my hardware fried and/or degraded over time due to that issue, no matter how much AMD and/or its fans downplay that risk.

It did not fail to meet PCI-E spec when they sent it in for testing. It is just auto adjusting the voltages too high and you can either manually fix it now, or they will be fixing it next week with the next driver update.

http://i.imgur.com/CFJYSWK.png

I don't see any price gouging at newegg.ca, it is the same price as US equiv.

$309 CAD vs $239 USD

And according to google: 309 Canadian Dollar equals 239.35 US Dollar

4GB is $259 CAD which is $201 US, same as US pricing.

 

[Qwertilot]

Actually, the real problem isn't the overvolting - its being careless enough to build a card that can draw much more power than allowed from a PCIE slot. There's simply no reason at all to do that.

Very definitely shouldn't have put out a whole batch of review/sale cards which did.

Honestly, it doesn't say remotely good things about the Q&A they're doing on these reference cards. Only relevant for the reference models of course, because the AIBs actually have brand names they care about and so will actually do a decent job of this sort of thing.

Why on earth AMD don't seem to is frankly baffling given these are what gets reviewed first. Presume its some sort of self defeating cost saving.

 

[ultima_trev]

Since one or more of the fans died on my 390, I plan on getting an MSI Twin Frozr or Gigabyte Windforce depending on which models tends to OC better. At 1.4 GHz it will probably match the GTX 980 in most scenarios and for $100 less.

 

[RussianSensation]

Since one or more of the fans died on my 390, I plan on getting an MSI Twin Frozr or Gigabyte Windforce depending on which models tends to OC better. At 1.4 GHz it will probably match the GTX 980 in most scenarios and for $100 less.

Sapphire Nitro will have removable fans for an easy replacement from the factory. Check to see if the card has dual ball bearing fans. GTX 1070 Armor doesn't but GTX 1070 Gaming does. For Gigabyte, their Xtreme series do but all other series don't. Even the popular G1 1070 still has the same fans that have been failing for years.

 

[sze5003]

The stock/availability issues of all the new generation cards has left me soured. I was really looking forward to buying the RX 480 on launch day, but alas that day has come and passed and I haven't been able to snag one yet. I did see a few marked up to $360CAD but I refuse to pay those prices. Not for a reference edition anyways.

Last night I ended up picking up a MSI Gaming GTX 1070 for $590CAD from Newegg. I wasn't looking at this class of GPU since I only game at 1080p for the moment, but a 4K HDTV/monitor is in my horizon in the near future and the RX480 was intended to be a stopgap. However, at the going "gouged" prices for the RX480, picking up the GTX1070 instead was a no brainer for me.

Newegg.ca showed the cards at normal prices as in the US. If anything the 1070 never launched at the stated "msrp" of $379.

It was just a matter of some more waiting until stock comes in. The only reason I would consider a 1070 would be if the AIB 480 models were priced too high or if I currently had a 1440p monitor.

But at the same time I get it, I've been without a GPU before and about to be soon now too since I am giving this card to my brother. Hopefully I'll be able to get a custom 480 soon.

 

[Gikaseixas]

Getting AIB version for my HTPC. Great card for 1080p gaming. Power consumption is great vs performance

Sent from my HUAWEI MT7-L09 using Tapatalk

 

[Sonikku]

Actually, the real problem isn't the overvolting - its being careless enough to build a card that can draw much more power than allowed from a PCIE slot. There's simply no reason at all to do that.

Very definitely shouldn't have put out a whole batch of review/sale cards which did.

Honestly, it doesn't say remotely good things about the Q&A they're doing on these reference cards. Only relevant for the reference models of course, because the AIBs actually have brand names they care about and so will actually do a decent job of this sort of thing.

Why on earth AMD don't seem to is frankly baffling given these are what gets reviewed first. Presume its some sort of self defeating cost saving.

Why wouldn't they limit PCIe slot voltage and just let any excess power come from the 6 pin connector?

 

[Thala]

Actually I'd rather reduce AO or GI, or shadow quality, than resolution. Far less loss of IQ.

Right, increasing the resolution is the worst thing you do in order to increase image quality. Its particularly bad, if you are forced to save elsewhere in order to maintain framerate when increasing resolution.

 

[Silverforce11]

My 2x RX 480s came in today. Some preliminary testing and brief conclusion.

Stock voltage definitely too high for stock clocks.

It had 1150mV @ stock clocks. At everything auto, it was boosting to around 1190mhz during benchmarks.

Not even 1266mhz boost clocks.

I manage to get it stable at 1100mV, but actually OC to 1.3ghz.

It was stable at this clockspeed at +5% power target. At default power target, it was hovering around 1288mhz.

Both card behave similarly, the 2nd one would tolerate lower at 1085mV @ 1.3ghz, but I saw artifacts in the benches, so I bumped it up to 1100mV.

My conclusion is AMD's power management is fail because it's defaulting to too high voltages for stock clocks, at each power state.

Other things I've noticed, fanspeed on auto is worst than on manual. It heatsoaks due to slow ramping, and it gets to around 83C with 2200 RPM and stays there. On manual setting, I just set the temp target max 80 and target 80, and it ramps up a lot quicker, reaching around 1800 RPM to get ~78C. Basically, it's quieter & cooler with your own temp target of 80C.

And this card is absolutely not noisy at all. It was barely noticable above my other case fans even at 2200 RPM.

 

[guachi]

I appreciate the information.

I'm waiting on (probably) getting a 480 until non reference versions show up. But it's nice to know that reference is fine for you (and others)

Just don't know if I'll upgrade my CPU or not yet.

 

[Silverforce11]

Reference isn't good, it has no OC headroom due to the power limit and voltage max. But at $379 AUD, it's the best in that price range for me by a longshot. I should have waited for Sapphire RX 480s... but I ran out of patience.

Definitely wait for custom RX 480s though if you're in the market for something of this price range.

 

[coercitiv]

It was stable at this clockspeed at +5% power target. At default power target, it was hovering around 1288mhz.

Both card behave similarly, the 2nd one would tolerate lower at 1085mV @ 1.3ghz, but I saw artifacts in the benches, so I bumped it up to 1100mV.

My conclusion is AMD's power management is fail because it's defaulting to too high voltages for stock clocks, at each power state.

If you told us it takes 100mV to see artifacts from either card, I would agree this was a power management fail, but with 65mV being all it takes to get artifacts, there isn't that much of a voltage guard band, is it?

Moreover, going from 1150mV to 1100mV equates to lowering power usage by just under 10%, which puts worst power usage example just under 150W. So the really "bad" cards need this 50mV cut to get within spec even though your "better" card showed artifacts at 65mV.

Would you feel comfortable sending thousands of cards on the market with as little as 15mV separating them from screens with artifacts? Even the slightest deviation beyond measured ASIC quality might generate a RMA.

Then again, let's see what the fix does.

 

[Silverforce11]

If you told us it takes 100mV to see artifacts from either card, I would agree this was a power management fail, but with 65mV being all it takes to get artifacts, there isn't that much of a voltage guard band, is it?

Moreover, going from 1150mV to 1100mV equates to lowering power usage by just under 10%, which puts worst power usage example just under 150W. So the really "bad" cards need this 50mV cut to get within spec even though your "better" card showed artifacts at 65mV.

Would you feel comfortable sending thousands of cards on the market with as little as 15mV separating them from screens with artifacts? Even the slightest deviation beyond measured ASIC quality might generate a RMA.

Then again, let's see what the fix does.

I am running 1.3ghz at an under-volt, not 1266mhz at 1150mV. I haven't tested stock clocks and seeing how far I can under-volt yet.

By default, it was running 1266 @ 1150mV.

Let me test that and see if it can handle 1050mV at stock clocks.

 

[Silverforce11]

Okay, here's the stock clocks @ 1050mV.

Note a drop of around 25W, avg power of the GPU goes from ~125W @ stock volts -> 100W via GPU-Z.

AMD's power management is not doing a good job at all if it's defaulting to 1150mV for these cards.

 

[JDG1980]

If you told us it takes 100mV to see artifacts from either card, I would agree this was a power management fail, but with 65mV being all it takes to get artifacts, there isn't that much of a voltage guard band, is it?

Moreover, going from 1150mV to 1100mV equates to lowering power usage by just under 10%, which puts worst power usage example just under 150W. So the really "bad" cards need this 50mV cut to get within spec even though your "better" card showed artifacts at 65mV.

Would you feel comfortable sending thousands of cards on the market with as little as 15mV separating them from screens with artifacts? Even the slightest deviation beyond measured ASIC quality might generate a RMA.

Then again, let's see what the fix does.

But the new features of Polaris should allow all this to happen automatically. The better cards should get lower voltages and the worse ones higher voltages. Reduction (albeit not elimination) of the necessary voltage guard band was explicitly shown on the release-day slides as one of the features of P10. So was boot-time power calibration. Based on AMD's own claims, we should be seeing the cards auto-adjust the voltage table on each boot to reduce it to the minimum necessary for stable operation. As of now, that's not happening.

 

[StrangerGuy]

My 2x RX 480s came in today. Some preliminary testing and brief conclusion.

Stock voltage definitely too high for stock clocks.

It had 1150mV @ stock clocks. At everything auto, it was boosting to around 1190mhz during benchmarks.

Not even 1266mhz boost clocks.

I manage to get it stable at 1100mV, but actually OC to 1.3ghz.

It was stable at this clockspeed at +5% power target. At default power target, it was hovering around 1288mhz.

Both card behave similarly, the 2nd one would tolerate lower at 1085mV @ 1.3ghz, but I saw artifacts in the benches, so I bumped it up to 1100mV.

My conclusion is AMD's power management is fail because it's defaulting to too high voltages for stock clocks, at each power state.

Other things I've noticed, fanspeed on auto is worst than on manual. It heatsoaks due to slow ramping, and it gets to around 83C with 2200 RPM and stays there. On manual setting, I just set the temp target max 80 and target 80, and it ramps up a lot quicker, reaching around 1800 RPM to get ~78C. Basically, it's quieter & cooler with your own temp target of 80C.

And this card is absolutely not noisy at all. It was barely noticable above my other case fans even at 2200 RPM.

Mind if you can try how much volts can you reduce for a stable 1190MHz?