MSI RX 480 Gaming X 8 GB

[Recus]

Funny, I can see plenty of 480s on overclockers.co.uk...

No AIB models. Also people still waiting for Sapphire Toxic Unicorn edition which will oc till 1.6 GHz and will be faster than GTX 1070.

 

[GhostRyder]

Well its the best overclocker out of the lot so far....But still completely disappointing overall. The only thing this card really has going for it is the cheap price for the reference models as the others seem to eclipse the 1060. Unfortunate...Hopefully some cheap after market 4gb cards will hit the shelves or the price will come a tad bit down on the 8gb variant.

 

[okidna]

I'm waiting to buy an AIB RX480 here in the UK, there are tons of GTX 1060s to choose from in stock, it's impossible to buy an AIB 480 right now. Constantly out of stock.

It's not out of stock, it's not on sale yet, most of the AIB cards (Sapphire, XFX, ASUS, MSI, Gigabyte, PowerColor) will be on sale in mid-August.

 

[daimonass]

Where test with 12DX ?

 

[rtwjunkie]

Where test with 12DX ?

He stated in Post#5 that DX12 just got fixed in Hitman,and will rebench later in DX12.
https://www.techpowerup.com/forums/threads/msi-rx-480-gaming-x-8-gb.224494/#post-3498559

"DX12 in Hitman has been fixed since, on next rebench it will use DX12"

 

[efikkan]

GTX 1060 remains the superior choice.

 

[bug]

GTX 1060 remains the superior choice.

It's slightly more power for slightly more $$$ (i.e. not superior if you don't have the extra $20-30). I'll probably go for the 1060 because of the Linux drivers (yet again). It seems AMD hasn't fixed any of their OpenGL implementation issues in the new driver (which seems to be actually the old driver, but with some parts open sourced).

 

[LooKMaYnE]

Custom GTX1060s are actually cheaper in EU than custom RX480s.

 

[wiak]

hey @W1zzard can you change out the 900 full size resolution to 1920 or 2560 full size resolution images in the reviews?, the current ones look pretty bad if you look at them on a 1200p monitor and horrible on 1440p/4k one
great review as always

 

[Air]

It is when one card is silent and the other sounds like a vacuum cleaner.
I find it a little confusing how a 150Watt card (even if its a little more than 150Watt) is having problems with heat. Pretty much all the AIB partner coolers have run virtually silent for a few generations now, even on highend 250watt cards. Even with current gen other, higher end cards draw more power than the 480. Most of these coolers are designed to handle up to 250-275 watts. I dont understand it.

1. Because its not really 150 W. In "typical gaming" reference uses 163 W, this one from MSI uses 196 W, close to ref. Fury (200 W) and ref. GTX 980 ti (211 W).
2. Die area is much smaller, 232 mm², while GTX 980 Ti and Fury are both 600 mm². Its like having to remove almost the same amount of heat through a much smaller "pipe".

That said, results for this card are pretty good.

 

[Fluffmeister]

No AIB models. Also people still waiting for Sapphire Toxic Unicorn edition which will oc till 1.6 GHz and will be faster than GTX 1070.

Man, that thing would suck power like it's going out of fashion!

 

[Captain_Tom]

I see. So the 480 sells above MSRP because of high demand and the 1060 because of different reasons.

Yes, so you can read.

 

[Captain_Tom]

1. Because its not really 150 W. In "typical gaming" reference uses 163 W, this one from MSI uses 196 W, close to ref. Fury (200 W) and ref. GTX 980 ti (211 W).
2. Die area is much smaller, 232 mm², while GTX 980 Ti and Fury are both 600 mm². Its like having to remove almost the same amount of heat through a much smaller "pipe".

That said, results for this card are pretty good.

AMD chose the 14nm process for 2 big reasons:

1) It is MUCH cheaper than TSMC's 16nm, and that allows them to be ultra competitive in the low - high end brackets. People just never seemed to notice when AMD took the performance crown. As such, AMD finally understands that their best days were the 4000 - 6000 series where they dominated price/perf and rarely worried about Halo products.

2) GloFlo's 14nm process is 9% denser than TSMC's 16nm. As it matures its clock-speeds should mostly catch up to Nvidia's, and when it does they WILL be able to scale higher than Nvidia can in the top end. Expect nvidia to hit a wall at about 16 billion transistors while AMD manages to scale to 18-20 billion. Sure it will probably use 320w, but who cares if it manages to win by 10% performance and costs less to make?

 

[efikkan]

GloFlo's 14nm process is 9% denser than TSMC's 16nm. As it matures its clock-speeds should mostly catch up to Nvidia's, and when it does they WILL be able to scale higher than Nvidia can in the top end. Expect nvidia to hit a wall at about 16 billion transistors while AMD manages to scale to 18-20 billion. Sure it will probably use 320w, but who cares if it manages to win by 10% performance and costs less to make?

What Samsung/GloFo calls "14 nm FinFET" is roughly equivalent to what TSMC calls "16 nm FinFET", and would be called "20 nm FinFET" if it were made by Intel. Both processes are basically a 20 nm node with FinFET, they have chosen a different measure just for marketing.

 

[xkm1948]

AMD card won't get better in old DX11/OpenGL games. The design is not geared towards DX11 at all. However DX11 still represents the majority of games on market right now. So Wizzard review is spot on to reflect how RX480 is doing for current and old games. With newer games utilizing new API we may see improvement of RX480 over time. DOOM is a glimpse into the future of how RX480 will do.

 

[dj-electric]

I read this review and i loudly said to myself:

"This RX 480 card consumes 207 watts at gaming"

I was not happy after that. This is a ridiculously high amount of power compered to cards like 3rd party GTX 1060s.
It is alarmingly and unresponsively high for a mid-range 14nm card.

 

[the54thvoid]

I read this review and i loudly said to myself:

"This RX 480 card consumes 207 watts at gaming"

I was not happy after that. This is a ridiculously high amount of power compered to cards like 3rd party GTX 1060s.
It is alarmingly and unresponsively high for a mid-range 14nm card.

Power consumption only mattered in the Fermi days. Then when AMD started to use more power and Nvidia started to use less power (relatively speaking), that power efficiency no longer mattered. Funny that.

Perf/Wat you're better off with a Fury or a Fury X (for AMD). In the UK you can buy a Fury (non X) for £300 which is a damn good deal and has that fangly HBM that everyone bitches about.

 

[dj-electric]

Power consumption ALWAYS matters.
Its what drives the mobile computer market, and gives headroom in GPUs to squeeze more performance out of.

If you would ask me if i would like a card that performs like X and consumes 120W in gaming, compered to one that performs like X+5% and consumes 200W, i would take the first one in a heartbeat.
Here, we have a dystopian situation where the X+5% is at the 120W mark and the X is at 200W. This is "back to the future of biff" bad.

It raises the question how will a RX 480 + 50% performance card will look like. Red hardware will cross the 300W mark at 14nm at this rate (we know how it looks on the Nvidia side). That means more heat, more noise, and gigantic chunks of metal to cool it all.

You need efficiency to move forward.

 

[geon2k2]

What Samsung/GloFo calls "14 nm FinFET" is roughly equivalent to what TSMC calls "16 nm FinFET", and would be called "20 nm FinFET" if it were made by Intel. Both processes are basically a 20 nm node with FinFET, they have chosen a different measure just for marketing.

How did you reach such conclusion? There must be some difference between them not just marketing.
Maybe their process does not allow them to pack the transistors so much as intel and by leaving more space between the transistors they are wasting wafer space, however it would be nice if you could provide more details about this.

 

[geon2k2]

Power consumption ALWAYS matters.
Its what drives the mobile computer market, and gives headroom in GPUs to squeeze more performance out of.

If you would ask me if i would like a card that performs like X and consumes 120W in gaming, compered to one that performs like X+5% and consumes 200W, i would take the first one in a heartbeat.
Here, we have a dystopian situation where the X+5% is at the 120W mark and the X is at 200W. This is "back to the future of biff" bad.

It raises the question how will a RX 480 + 50% performance card will look like. Red hardware will cross the 300W mark at 14nm at this rate (we know how it looks on the Nvidia side). That means more heat, more noise, and gigantic chunks of metal to cool it all.

You need efficiency to move forward.

Efficiency has many variables.
See this link: http://www.legitreviews.com/amd-radeon-rx-480-undervolting-performance_183699

With 0.05 voltage decrease they reduced power consumption with 10-30W making this an ~130W card instead of 150, while also getting more performance due to lower throttling. From what I see MSI has done the opposite just to squeeze a little bit more performance. You need to strike the right balance between clocks/performance and you will get great efficiency. Maybe with 0.1 less and 100 mhz less they can get fantastic off the chart efficiency, but then they will not meet the desired performance target, which was VR Ready.

I used to have phenom II in the past with stock voltage of 1.35 and running at 3.0Ghz. I ran it for most of its life in my PC at 3.6 Ghz and 1.25 V. Considering that power consumption is directly proportional with voltage square 2, I would say I improved the efficiency of phenom II with 30-40%, with very little tweaking.(Didn't do exact calculation, there is a formula somewhere on the net if you want to see, but it is more than the 20% which comes directly from overclock). I think AMD should do a bit more checks from the factory and be a bit more aggressive, but then there might be a risk that in some scenario (maybe furmark, with renamed exe) it will hang so they don't want to take this risk.

 

[dj-electric]

What i wanted to see is companies doing their adjustments.

I am also looking forward for the Samsung made GPUs it is clear as day to me that they will be of a much higher quality than GloFo's.
I'm still hoping for an increase of 30%+ in perf\watt ratio. At its maxwell-like state, even stronger GPUs than the polaris 10 will just not a appeal to me.

My plan is to get my 290Xs out of there for 180Wish GPUs

 

[dyonoctis]

Power consumption ALWAYS matters.
Its what drives the mobile computer market, and gives headroom in GPUs to squeeze more performance out of.

If you would ask me if i would like a card that performs like X and consumes 120W in gaming, compered to one that performs like X+5% and consumes 200W, i would take the first one in a heartbeat.
Here, we have a dystopian situation where the X+5% is at the 120W mark and the X is at 200W. This is "back to the future of biff" bad.

It raises the question how will a RX 480 + 50% performance card will look like. Red hardware will cross the 300W mark at 14nm at this rate (we know how it looks on the Nvidia side). That means more heat, more noise, and gigantic chunks of metal to cool it all.

You need efficiency to move forward.

Amd has definitely lost the mobile market. Unless they are going to do a custom chip with HBM there is no point in making a RX480m, when the 1060m will be faster and way more power efficient. Assuming that Vega will be the RX 490, I don't think that Red team will hit the 300w mark. I rather think that it will be like Fury/390: way moar core for equal/lower power. Unless Amd choose to go nuts and release out of the blue a 3072 core gcn with GDDR5.
Then there is Navi that we don't know **** about but is planning to be more power efficient .

(However if Nvidia is planning to go full HBM for it's next gen, then it's gonna be trouble again. The huge lead they got in power efficiecy will get bigger and they will be able to cram a shitload of shaders units on their gpu.)

 

[Jism]

AMD did not give any much more headroom on OC's, so every card (reference or AIB) is looking at a already maxed out chip where OC's are really marginal. Yes it could reach 1500MHz but proberly not for 24/7 usage and just good enough for one run and nothing more.

This is the proces limitation. It suits best at 1200 up to 1400Mhz in very rare cases, but anything above is a litterally no go on Air.

It's not the VRM, it's overdesigned up to 250W, it's not the 6 Pins vs 8 Pins header either, 6 pins is enough for a rough 300 up to 400Watts far above spec, it's not the cooling as well, still 130W TDP / chip (there is some TDP loss at VRM & memory chips as well) is very excellent to cool.

Nvidia's approach is different, they render frames based upon tile's rather then 1 frame at a time. This puts all the 'cores' at work and with the right programming utilizing all available cores.

The moment AMD applies the very same rendering technique, nvidia can pack up their cards and brainstorm for something new. Because AMD cards have a much better raw performance then Nvidia in general, they tend to utilizing things in a more efficient matter.

The 480X is not a bad card or chip, when was the last time we had a chip for 200 up to 270$ with this type of performance? It's bin a while, and we all should praise AMD anyway for giving us that chip that does very very well up to WHQD resolutions. Compare that with Vulkan API and you have a very decent and beefy card. Wait 1 or 2 years, buy yourself a second 480x and double the performance in many games. What more do you want?

 

[deu]

So 480 non ref max OC = 1060 ref stock.

In that game and that resolution = yes; Im not sure if you're aiming for a specific point. You could also say that 480 = 980Ti if you take hitman or DOOM with vulkan (almost), but id say its a pretty narrow comparison. The reason why battlefield 3 is used (as far as I understand), in the OC test, is to best show the represented percentage advancement in performance. You could argue that BF3 maybe a little outdated and that newer games will benifit differently from the average OC but I assume Wizzard knows when its time to change this

 

[deu]

Amd has definitely lost the mobile market. Unless they are going to do a custom chip with HBM there is no point in making a RX480m, when the 1060m will be faster and way more power efficient. Assuming that Vega will be the RX 490, I don't think that Red team will hit the 300w mark. I rather think that it will be like Fury/390: way moar core for equal/lower power. Unless Amd choose to go nuts and release out of the blue a 3072 core gcn with GDDR5.
Then there is Navi that we don't know **** about but is planning to be more power efficient .

(However if Nvidia is planning to go full HBM for it's next gen, then it's gonna be trouble again. The huge lead they got in power efficiecy will get bigger and they will be able to cram a shitload of shaders units on their gpu.)

AMD actually have quiet alot of the mobile market with APU's but yes they have NO gaming solutions whatsoever