ASUS Republic of Gamers Announces Strix RX 480

The elephant in the room is that we can talk all day about both cards but you pretty much cannot buy either. The buzz kill is the total and entire lack of availability.

 

For a lot people they don't have big income it can take up to 2 years to save up that kind cash and there pushing card to over $400+US not counting S/H so it not good option
Right now most people have 3 option, I just go for RX 480 4GB which I have enough save up right now to replace my HD 7870 or could wait in tell Dec time and I should enough for 8GB model or maybe even GTX 1060.

To me Gameworks and SGSSAA not a big deal any way

That true they Sold Out GTX 1060 really fast

 

For me FreeSync is additional $100 value, for AMD cards. I don't imagine buying next graphic card without adaptive sync technology.

 

That true when you factor in cost of new Adaptive Sync Technology Monitor then start seeing really big diff in cost with the G-SYNC $600+ where Free sync 27" 144Hz can be found for $400 and even far less then that with 1080p Resolution and 60Hz Refresh Rate.

 

You actually do get better FPS with a more efficient architecture. Let's say the 1060 gets 60fps at 70C while the 480 gets 60fps at 80C, the 1060 can be overclocked until it reaches the 80C threshold and produce a higher framerate than the original 60fps.

Alternatively, you can maintain the same fps while producing less heat. This is far more important for my next mini-atx/htpc build. I am planning to buy a mini 1060 card so that I'm exhausting less heat and have greater airflow than if I were using either a Polaris chip or a fullsize card. Lower TDP = less heat to push out of my case, simple as that.

While the electricity savings are going to be minimal and that benefit is overblown (unless you game 24x7), there is no doubt that a more efficient chip provides real benefits to the consumer.

 

Everyone is making comparissons with what they think is in their advantage. The Question is, why bother compare reference temperatures in an AIB card article? The same case was with the r9 290x with "omg 90 celsius, omg, silent mode, uber mode bla bla", nobody cared once AIB card released and the 290x was in the same boat with 780ti temperature wise.

Lets not forget, the 480x is power hungry. The reference design was there to somehow please the "power eficiency/usage" freaks. Once AIB partner add the 8 pin required for the 480x and throw a good cooler on it (like this Asus Direct CU3) i bet it already performs better then the reference design. Increase the power target in the drivers with 10% and you dont need to change the clocks and still get arround 10% performance increase.

 

https://www.computerbase.de/2016-06/radeon-rx-480-test/5/
Thats a Fractal Design Define R5 with an RX480 after 20 mins, clocks speeds are down to 1160mhz in the games they tested at 1440p.

Now i'm not going to get into if you think thats a great case or not, the lans i go to where people would buy $240 cards its would probably be mid-upper end for them.

So could Asus engineer that result , absolutely , would they have to try hard to pick the case to do it .. i don't think it would be hard looking at that.

Whats telling is Asus's same card for the 1060 (strix) they report about a 6-7% gain , so would they put themselves in a worse light there than with a rx480 strix ? i doubt it - they would use the same methods i would think.

Asus is telling us there strix model will gain 10%+ on the 480 vs their 1060 strix is the point.

 

Yes

The rx 480 has a 256 bit bus to the 192 bit bus of the 1060.
That with the extra memory pulls a extra 30 watts so the power usage of these 2 cards is much much closer than it seems right now.
Nvidia appears to be more efficient and it likely is but its not as much as it seems because of the memory bus different power requirements.
It would be nice to know what that actually is.
The vulkan doom benchmark results say to me all i need to here.
Amd new architecture is only going to get better and better over time as game devs use more and more of its abilities so its only going to get faster and faster.
Right now 8 gig on this rx480 is way to much it will never use it without it being unplayable.
I do think however AMD know something we dont hence the 8 gig.
Like the playstation 4 has.

 

I like way you introduce your System configuration, but to your statement.

I can make my Fury X eat good 370W and it will still be kept under 50°C while being quieter than your blower kind of FE 1060.
And the fun part, Fury X with some voltage tuning eats like 250W max.
- - - -
But to give you some credit...
You have it nearly right, but you think in wrong properties. Temperature is function of generated heat vs cooling capacity.
What matters with Power efficiency is maximum performance extracted at 300W standard point where top cards aim at desktop.
But in end of the day, many people will run their cards out of specification and not one of us will care.

And in past even with nVidia having much more power efficient GPUs. Once things hit top of this 300W limit, there were not so big difference.
- - - -
And side note, You do not overclock till you reach certain temperature, You overclock till you reach limits of stability. GPUs are not Vishera clocked to 8GHz under LN2, they have other electrical limitations.

 

Yes, they know that several manufacturers are discontinuing 0.5GB/chip memories
And that 1GB per chip memories cost nearly same.

At this time, nVidia should not be selling cards with "only" 3GB vram. Not because it is insufficient for 1080p for which GTX 1060 is targeted. But because there will be bunch of lazy developers who will see 6/8GB of vram on a lot of GPUs and will not care about better texture management.

It is not problem for older 3GB gpus, because they are bit weaker and people are used to reduce some details a bit.
But people who will have degraded performance of will have to sacrifice details on 3GB GTX 1060 just because each 0.5GB chip did cost $2 less?
(Good thing is that nVidia said that they'll not introduce 3GB version at launch time.)
And if they do so later, not many people will buy them as everyone will consider 4GB+ as new standard.

 

question

Can anyone say for sure what the power difference is between
8 gig 256 bus and 6 gig 192 bit bus.
It would be interesting to know this information if anyone has it.

 

And on the same page you can clearly see the difference of 4% between throttling card and RX 480locked at 1266MHz via increased PT/TT: AMD RX 480 MAX (8GB)
https://www.computerbase.de/2016-06/radeon-rx-480-test/5/

If you think that 1330 MHz overclock can gain more than 1330/1266 = 5% over such 1266MHz locked card... I won't try to dissuade you any further.
From the same reviewer:

 

When not throttling the 480 and 1060 are 1% from each other not that bad.

 

Nah...to get there you should compare "RX 480 Max" with "GTX 1060 Founders Edition Max" (2nd pic)

480 loses 4% at default PT/TT compared to Max
1060 loses 3%

GTX 1060 throttles only 1%... less than 480

 

OK 6%.
6% + 5% = 11%

11% faster than ref. 480

And only for those who don't know how to or refuse to increase PT/TT on their ref. 480.
Else - 1330/1266 = 5% increase

 

Don't forget they might have also increased memory speed. Or did they already announce mem speeds?

 

Some promising leaks from Saphire nitro models. Gives high hopes for this card aswell.

Take them with grain of salt. Found them from Anand forum, but don't know the source.

Spoiler

 

Efficiency matters at the top-end. Desktop Maxwell stopped at 250W and surpassed its competition by quite a decent amount; think about what that means, just for a second.

Nvidia's fault for what? For making an efficient architecture? Dammit Nvidia, how could you!
Until Maxwell the themal envelopes have been relatively similar (although the 290x was power hungry way beyond AMD's 250W envelope, but hey; let's ignore that small fact).
You can see Maxwell's fruit right now in Pascal, the 1070 needs about the same power as the 480 yet it kills it performance-wise in every possible way and by a very wide gap.

Now it's suddenly a problem how Nvidia decided to make an efficient architecture yet we can clearly see it was a smart move? Give me a break man. Maxwell is likely one of the reasons why Pascal was able to launch in such short order seeing as there are already 3 Pascal cards on the market compared to 1 Polaris card. Nvidia didn't have to do squat, they just solved 1 or 2 of Maxwell's deficiencies, shrunk the whole thing and clocked it like crazy.

Are you being serious right now? A fan. You're blaming the 480's power consumption on a fan? Toppest of keks this is the most ludicrous thing I've heard lately.

I also remember an article stating how Raja Koduri said that the VRAM on the 480 takes maybe 30W IIRC correctly? That sounds pretty ok to me, I can't find the article however but you're seriously overexaggerating how much VRAM + A FAN takes from the overall TDP.

HBM doesn't cut as much power as you think it does, in case that's where you were going with this.

On this I can agree, if ASUS manages to do something like that it will outright kill the 1060 once and for all. However keep in mind, marketing is the name of the game. ASUS might simply be cherry picking situations or outright spewing bull**** just to attract attention. Like AMD did with the 480, said 2 of them are better than a 1080 but it turned out it's only in AOTS.

I might not be a fan of AMD's vaporware tactics (not like I'm a fan of Nvidia's messed up business practices either) but AMD really needs to sell a lot of cards this generation.
I'm particularly scared for them because the 480 doesn't look like a spawn of Fiji to me considering the rather unimpressive TDP.

What improvement are we talking about? Performance? Because if so, you're saying nonsense. The 1070 is almost twice as fast as the 970, the 1080 is almost twice as fast as the 980 while the 480 is almost twice as fast as the 380. I see a similar increase.

Man I have no idea what's up with these posts of yours.

 

Maybe reading #30 will help a little.

Spoiler

Same improvement as I always talk about when I mention: "per transistor investment". It is improvement per building block of ROP, TMU, SP, ...

It is good to see you back, but you do not have to go into: "I do not understand place."
Because I am not always in mood to explain things I am sure people around should understand. And I am damn sure you would have no problem to extrapolate and understand all those things you did "not" understand while you made quoted post.