AMD RX 400 series (Polaris) MegaThread! FAQ & Resources

[Sakkura]

To be fair Nvidia has a fair amount of memory compression, and I have always found myself gaining more from a memory overclock on AMD cards. Just because the 1080 does not use more bandwidth, does not mean more is not needed or more could be taken advantage of.

AMD has memory compression too, including the delta color compression that Nvidia introduced with Maxwell.

 

[Robert Cook]

To be fair Nvidia has a fair amount of memory compression, and I have always found myself gaining more from a memory overclock on AMD cards. Just because the 1080 does not use more bandwidth, does not mean more is not needed or more could be taken advantage of.

AMD has memory compression too, including the delta color compression that Nvidia introduced with Maxwell.

yes, although from what I remember not as much?

 

[Sakkura]

I think so, but I haven't seen any in-depth tests.

 

[-Fran-]

Anandtech had a very good article when they first introduced those features with Tonga. From then on, they've been updating that information with each card release.

That being said, I have no idea what the specific differences between AMD's and nVidia's compression shenanigans are.

Cheers!

 

[TehPenguin]

You showed me a bench of 3 games before and it didn't bother you then ^^
Yes, 10% increase in memory speed does give as little as 3% in some games, but if you'd read the article you'd see that 12% increase in GPU clock only gives 4% more performance in some games which is also 1fps at 4k so what is your point?

i'm not really talking about performance increase with overclocking. my point is more about if current GDDR5/GDDR5X holding back our gpu performance or not. if the gpu are severely limited by it's bandwidth then overclocking the memory should uplift it's performance quite noticeably.

I understand your argument. From what I've understood you're saying that if the memory overclock yields only a couple % of extra performance then it means the chip is not being held back by the memory i.e. wouldn't see any meaningful performance bumps from faster memory.

Am I right on that? Is that where you stand?

Going forward, with the same logic, the core clock frequency increase also gives a couple of extra % of performance, which should mean that a faster clock speed would not be meaningful for the overall performance of the chip.
This is where I disagree. I believe it is safe to say that, looking at the performance difference between Pacal and Maxwell, Pascal has the edge mostly because of the so much higher clock speed. We're talking about solid 600MHz++, which was achievable due to the much higher efficiency of the Pascal architecture.

Where am I going with all this? I try to back my theory that a much faster memory will indeed bring significant performance increase to the table, just as a much faster chip does. It should not be written off just because a slight increase in bandwidth doesn't make a big difference.

I think we'd be surprised if we could test the performance of the Titan XP with HBM2.

 

[renz496]

then maybe we will have to see how Quadro GP100 vs Quadro P6000 in rendering performance. Quadro GP100 was expected to have 720GB/s of bandwidth. Quadro P6000 will only have 432GB/s of bandwidth with it's GDDR5X was running at 9Gbps. and while i was looking at this two i was reminded how well Quadro P6000 vs against Titan XP in games:

http://hothardware.com/reviews/nvidia-quadro-p6000-and-p5000-workstation-gpu-reviews?page=6

despite being slower in games Titan XP actually have more bandwidth than Quadro P6000. Titan XP bandwidth is rated at 480GB/s

 

[Sakkura]

It's a balancing act. As the rest of the GPU gets more powerful, you need more memory bandwidth to keep up. Clocks on GDDR5 ran out of scaling headroom, GDDR5X can only extend that so much. Then you need to move to wider memory interfaces to keep increasing the bandwidth, but that's expensive. That's when HBM starts to make sense.

As such, we'll never see HBM at the low end*, and perhaps not the midrange either. But I would expect it to gradually take over the high end, pushing GDDR5X and GDDR5 further down the totem pole. Only question to me is whether HMC or other new memory types show up and steal HBM's spot.

*Also, iGPUs are likely going to continue to eat up the low-end market. So in that sense the low end is moving towards DDR4 memory. With affordable platforms being limited to dual channel memory though, iGPUs will be held back and not threaten the midrange dedicated GPU market.

 

[TechyInAZ]

It's a balancing act. As the rest of the GPU gets more powerful, you need more memory bandwidth to keep up. Clocks on GDDR5 ran out of scaling headroom, GDDR5X can only extend that so much. Then you need to move to wider memory interfaces to keep increasing the bandwidth, but that's expensive. That's when HBM starts to make sense.

As such, we'll never see HBM at the low end*, and perhaps not the midrange either. But I would expect it to gradually take over the high end, pushing GDDR5X and GDDR5 further down the totem pole. Only question to me is whether HMC or other new memory types show up and steal HBM's spot.

*Also, iGPUs are likely going to continue to eat up the low-end market. So in that sense the low end is moving towards DDR4 memory. With affordable platforms being limited to dual channel memory though, iGPUs will be held back and not threaten the midrange dedicated GPU market.

I wonder if IGPUs are actually going to get hindered. Cause Ryzen CPUs don't have any onboard graphics that we know of (i'm NOT talking about the Zen APUs), so that will help out the low end graphics card market stay live.

 

[TehPenguin]

The quadro p6000 has 24GB vRam tho, compared to Titan's 12. Wouldn't that be a reason for lower bandwidth being sufficient enough?

 

[Sakkura]

I wonder if IGPUs are actually going to get hindered. Cause Ryzen CPUs don't have any onboard graphics that we know of (i'm NOT talking about the Zen APUs), so that will help out the low end graphics card market stay live.

But how many people will be buying a high-end CPU and a low-end GPU, intending to game on it? I mean, outside of the usual misconfigured prebuilt PCs...

 

[Robert Cook]

i7 7700K and a GTX 1050ti I am set for all the n00b Pwning rite?

 

[RCFProd]

Atleast let me know when a post I made is removed lol. Not sure what I did wrong otherwise or if it was a mistake.

 

[ex_bubblehead]

Atleast let me know when a post I made is removed lol. Not sure what I did wrong otherwise or if it was a mistake.

Zen/Ryzen thread is here: http://www.tomshardware.com/forum/id-3327589/amd-ryzen-megathread-faq-resources/page-2.html#19286866

This thread is for AMD RX 400 series (Polaris) discussion. Anything else is off topic.

 

[AndrewJacksonZA]

Hmmm... On Monday I just bought a MSI RX470 Gaming X 4GB for +-USD165, which is an AMAZING deal in my country. I had _no idea_ that Nvidia would be launching OC editions. I wonder if they will force retailers to lower the prices of the 1060 6GB or just price the OC cards higher than the normal 1060s.

*sigh*

(However, I bought the RX470 for a specific purpose: to play BF1 on ultra at 1080p at 60+ FPS, so in that respect it fulfills it purpose, and I'm happy.

 

[Martell1977]

Hmmm... On Monday I just bought a MSI RX470 Gaming X 4GB for +-USD165, which is an AMAZING deal in my country. I had _no idea_ that Nvidia would be launching OC editions. I wonder if they will force retailers to lower the prices of the 1060 6GB or just price the OC cards higher than the normal 1060s.

*sigh*

(However, I bought the RX470 for a specific purpose: to play BF1 on ultra at 1080p at 60+ FPS, so in that respect it fulfills it purpose, and I'm happy.

They will likely just price the OC cards higher, as they always tend to do.

I have to admit, with the going price of many of the 480 8gb cards, it's difficult to recommend a 1060 6gb from a price/performance standpoint. They perform so similar and yet the 480 is much cheaper....for some models.

 

[Math Geek]

on average the 480's are cheaper than the 1060 cards. since they are pretty much neck and neck in performance, the lower price gets the nod for sure. i like the 480 over a 1060 everytime at their current prices.

 

[TechyInAZ]

on average the 480's are cheaper than the 1060 cards. since they are pretty much neck and neck in performance, the lower price gets the nod for sure. i like the 480 over a 1060 everytime at their current prices.

I wonder why the RX 480s got so cheap?

 

[-Fran-]

on average the 480's are cheaper than the 1060 cards. since they are pretty much neck and neck in performance, the lower price gets the nod for sure. i like the 480 over a 1060 everytime at their current prices.

I wonder why the RX 480s got so cheap?

Well, nVidia never lower its prices unless its forced to; sort of speaking.

I do remember, very well, the 600 series never went down in price in the mid-range when AMD went down in price with time.

The other option is nVidia has a fixed wafer price of something, so when production ramps up they don't get cheaper or something?

Cheers!

 

[Sakkura]

on average the 480's are cheaper than the 1060 cards. since they are pretty much neck and neck in performance, the lower price gets the nod for sure. i like the 480 over a 1060 everytime at their current prices.

I wonder why the RX 480s got so cheap?

Nvidia's GTX 1060 has been outselling the RX 480 by a wider and wider margin. The aggregate is about 4 GTX 1060s for each RX 480, but in recent months the ongoing sales ratio has been more like 5:1.

It's sad because the RX 480 has been if not a better buy, then at least on par.

 

[Math Geek]

if you read any kind of forum, folks still bash AMD cards universally. dos not even matter if what they are saying is true or not, they still say it. so the overall impression novices have from reading such drivvle is that nvidia is somehow vastly superior. which we know is not true, but novices don't know this.

for instance the 1080ti is gonna be a 300w card for custom high end models. is this gonna start the "space heater" comments like AMD got for their 300w cards? of course not, won't even be mentioned.

sadly this is what AMD is fighting. at least on this forum most of us know the truth and correct the fanboys when they pop their heads up to spout nonsense.

 

[cdrkf]

if you read any kind of forum, folks still bash AMD cards universally. dos not even matter if what they are saying is true or not, they still say it. so the overall impression novices have from reading such drivvle is that nvidia is somehow vastly superior. which we know is not true, but novices don't know this.

for instance the 1080ti is gonna be a 300w card for custom high end models. is this gonna start the "space heater" comments like AMD got for their 300w cards? of course not, won't even be mentioned.

sadly this is what AMD is fighting. at least on this forum most of us know the truth and correct the fanboys when they pop their heads up to spout nonsense.

This is the thing- nVidia's greatest strength is marketing imo. There was a youtube vid the other day with a guy who has been critical of Ryzen going on about how 'fair' he was to AMD giving them lots of videos but how it's universally known that 'AMD have always been the value proposition, in cpu and graphics' and 'They have *never* been ahead of nVidia'...

I can't get my head around that statement from someone who's supposed to be an older tech enthusiast. I remember AMD not just being better value but being the outright uncontested leaders in both the CPU space and on the GPU side of things over the years. The Athlon, Athlon T-Bird, Athlon XP and Athlon 64 all matched or outright bested Intel in their respective generations (heck going back further the AMD 286, 386 and 486 parts were superior to Intels). On the graphics side they've been ahead with the 5870 and the 7970 and were on performance parity with the 290X. It's only really been a matter of a few years where nVidia have had an outright technical advantage, before then they traded blows and leapfrogged each other on a regular basis. I wouldn't be shocked if Vega took the lead from the 1080ti this generation, even if it doesn't I doubt it will be far off. The way the tech sites talk about them though you'd think Vega will struggle with the 1070

 

[Math Geek]

i agree completely. all i can do is correct the fanboys when i see them on this site. it's the only forum i lurk in. only one worth my time really. not seen any other tech sites like this where intelligence is the norm rather than hopes dreams and wishful thinking.

at least folks who really want to know the truth can come here to get it. the rest can keep the nvidia execs in new cars every year i guess. i'm not swayed by marketing and hype but rather facts and figures. so long as i can know what the ideal part is for my needs, that's what i'll go with. no matter who it is. at the top end right now it's nvidia of course but when Vega comes out, we'll see how that goes.

i'd love to see them beat or match the new 1080ti but somehow i doubt that's gonna happen. but 1080 performance at a lower price point is a total win to me. hell even same price performance as 1080, and i'd favor AMD just to help them along a bit.

 

[renz496]

if you read any kind of forum, folks still bash AMD cards universally. dos not even matter if what they are saying is true or not, they still say it. so the overall impression novices have from reading such drivvle is that nvidia is somehow vastly superior. which we know is not true, but novices don't know this.

for instance the 1080ti is gonna be a 300w card for custom high end models. is this gonna start the "space heater" comments like AMD got for their 300w cards? of course not, won't even be mentioned.

sadly this is what AMD is fighting. at least on this forum most of us know the truth and correct the fanboys when they pop their heads up to spout nonsense.

lol nvidia have that moment with "thermi". AMD even make official video about that just to poke fun at nvidia:

https://www.youtube.com/watch?v=wCi5hiIO4fs

when we talk about being "bash" i don't think only AMD gets them while people turn blind eye to what nvidia did. just look how bad it is with nvidia "gimp works" for example. but AMD also try to do something similar before but we never really see the internet rage about it. all this aside nvidia did have excellent marketing and devrel. and they willing to spend big money on this. i think this is the difference between nvidia and AMD in the past. just look what nvidia did with PhysX and what AMD did with Bullet physic. we see nvidia sponsor some games to use GPU PhysX. AMD on the other hand despite their official partnership with Bullet forget about sponsoring games to use Bullet (for gpu accelerated feature) they don't even help promoting bullet. they announced the partnership in late 2009 and then nothing. of course AMD in the end understand to see stuff being adopted you need to spend resource to make it happen. but when they finally aware about it the company already in bad position financially.

 

[CapsLuke]

is rx480 8gb still better than gtx 1060 6gb (ON DX12) after nvidia dx12 driver release?

 

[renz496]

is rx480 8gb still better than gtx 1060 6gb (ON DX12) after nvidia dx12 driver release?

i haven't seen direct comparison for 1060 vs RX480 yet. but the jump actually not coming from single driver updates like from last WHQL to current latest one. but more like gradual performance upgrade from the release driver.