• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Radeon R9 390X Arrives In 1H 2015 – May Feature “Hydra” Liquid Cooling

BAO isn't demanding(in comparison with the rest of the titles used) they are all high.

MLL isn't vram intensive?

It's very lean on vram, iirc Kaap posted the amount@4K but it isn't high.
 
"The GeForce GTX 980 is the world's most advanced graphics card, powered by next-gen NVIDIA Maxwell architecture."

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980

Watch out for this lad, he's more informed than the manufacturer from which he buys cards. He's the Van Halen groupie who knows more about a guitar solo than Eddie Van Halen. :rolleyes:

Lad, if you mean kaap then he know's his onions, you on the other hand clearly lack technical comprehension, hence the 980 appears high end to you.
 
"The GeForce GTX 980 is the world's most advanced graphics card, powered by next-gen NVIDIA Maxwell architecture."

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980

Watch out for this lad, he's more informed than the manufacturer from which he buys cards. He's the Van Halen groupie who knows more about a guitar solo than Eddie Van Halen. :rolleyes:

Welcome to OcUK :P

That is just PR headline fluff doesn't mean anything at all.
 
Can only be answered when it arrives really.

The bus is a genuine argument, not an issue unless running 4K imo:

d56owW3.png



Batman/Metro aren't vram intensive even at high res=Maxwell stretches it's legs, intensify the vram and it starts to trip up.


I would probably pay the premium and get a 970 over a 290 if I was upgrading today, very good card matching last years AMD and go 120Hz 1440p as 4K holds very little interest, 120/144Hz is more important personally.:)



:D

You are not understanding how the new architecture works on the new 980 and 970. It uses better memory compression, so whilst it only has a 256bit bus, it deals with textures far more efficiently than Kepler for instance. The bus size is pretty irrelevant for these cards, as has been shown to you a few times.
 
BAO isn't demanding(in comparison with the rest of the titles used) they are all high.

MLL isn't vram intensive?

It's very lean on vram, iirc Kaap posted the amount@4K but it isn't high.

No idea, although i'm not convinced that the lower bus is that much of an issue between the cards for it to make any significant difference. I'll wait until some real user benchmarks trickle through at 4K in the benchmark threads between the 290's/970's/980's/290X before I pass my judgement.
 
I understand fully greg, max up in game AA and we'll see how they all truly fare with performance drop off directly tied to bus/res.

@Gee,

It isn't an issue in the grand scheme of things tbph, 4K isn't going to see a mass adoption, gregg tried it and moved back down for example.

This all started with bru asking what's going to be argued next time, I stated it's a genuine argument and low and behold here we are.

It's not really the place to talk about it- it's way off topic and would be better served up in one of the 9 series threads.:)
 
Last edited:
I understand fully greg, max up in game AA and we'll see how they all truly fare with performance drop off directly tied to bus/res.

Well going from what Gee posted, let's have a look at these bench results....

Theif 3840x2160 HQ
290x Uber/Mantle 40.0 Fps vs 980, 47.9

Battlefield 4 3840x2160 HQ
290x Uber/Mantle 38.8 Fps vs 980, 42.2

Bioshock infinite 3840x2160 HQ
290x Uber mode 30.2 Fps vs 980, 43.2

Metro last light 3840x2160 HQ
290X Uber mode 30 Fps vs 980, 37.3

Crisis 3 3840x2160 HQ
290x Uber mode 26.6 Fps vs 980, 29.1

Company of heros 3840x2160 HQ
290x Uber mode 41.3 Fps vs 980, 42.8

Now Crysis 3 in "Uber mode" is getting 29.1 on a 980 and a slower 26.6 on a 290X. I don't know of any more games that are more demanding than Crysis 3, so you clearly are not seeing what is what. No point going further though, as we are getting off topic.
 
You are not understanding how the new architecture works on the new 980 and 970. It uses better memory compression, so whilst it only has a 256bit bus, it deals with textures far more efficiently than Kepler for instance. The bus size is pretty irrelevant for these cards, as has been shown to you a few times.

No it's not, it changes the effective bus size by reducing the bandwidth. A higher bus will be EXACTLY as effective on such a card. The previous cards already had MANY types of compression to make the bandwidth usage more efficient, more bandwidth simply means you can transfer more data, it's that simple. AS data usage increases so does bandwidth usage even with compression, once you get to the point there isn't enough bandwidth you get a slow down.

It may be that 384bit Kepler = 256bit Maxwell, but a 512bit Kepler would improve performance when bandwidth is required vs 384bit Kepler, and a 384bit Maxwell will perform better than 256bit Maxwell.

I pointed it out before these cards were released. Efficiency in usage of bandwidth changes from one generation to the next, the same size bus isn't directly comparable between generations though the improvement in efficiency for both Tonga and Maxwell has been impressive.

Some games have larger or better textures and thus at high resolution might start to slow down at the 300GB/s mark, other games with smaller textures might only required 250GB/s at the same res, if your card has 256GB/s it might do fine in the second game but suck in the first game.

Games are also different and compression will be different. It's fairly usual that compression will sometimes help a lot and sometimes not at all. Think to differing SSD performance between those that ran compression and didn't, sometimes the ssd's that relied on heavy compression were the same speed, in some situations they were much much slower, it depended entirely on the data.

Same with Maxwell/Tonga and in general all compression technology, some games will likely lend itself to the new compression other games may not, so some games will get significantly more or less bandwidth efficiency improvement than other games. Which could be why in particular those above results are as they are. Some games the compression works well and 980 sli works great, in other games the compression sucks and/or bandwidth requirements are higher and 980 sli gets severe slow down.
 
d56owW3.png


Playable!

So what is the reason for the 980 doing so much better in Last Light, valley and Heaven compared with the AMD cards? I don't see how you can draw any conclusions from that table at all about how bandwidth effects 4k + AA performance. : /
 
Last edited:
With no AA, can't you read what's already been established?

So add AA and bring it down to what, 15 fps at a guess? Yer, that will show what is what. With everything maxed and 3 Titans at 1440P, I was lucky to get 60 fps in places. Pointless debate when you clearly don't understand how the new Maxwell architecture works.

Might pay for you to have a look at this thread, as it might make it clearer for you to understand how Maxwell deals with memory compression.

http://forums.overclockers.co.uk/showthread.php?t=18626546

No point continuing unless you watch that really.
 
PR fluff.

It's ground hog day again except it's single 980-it's ok v's ignored higher bus saturation setting limitations with twin 980's.

I'll remind you all when T2 exposes the limitations and destroys AMD's current line up.;)
 
PR fluff.

It's ground hog day again except it's single 980-it's ok v's ignored higher bus saturation setting limitations with twin 980's.

I'll remind you all when T2 exposes the limitations and destroys AMD's current line up.;)

Well if you won't even bother to see how the 980/970 deals with memory compression, there is no point continuing this really is there. Fingers in ears to suit your way of thinking is a poor show really.

At anyone interested and this is the same for the 256bit bus on the 285 and possible future AMD/nVidia cards.

The fact is though, even the 285 is using a very similar tech and called it "Lossless Delta Color Compression", where as nVidia have called it their "Third generation Delta Color Compression". Both are achieving the same goal though. Maxwell works by giving more choice of delta calculations to the compressor and this in turn reduces the memory bandwidth needed for that DCC.
 
PR fluff.

It's ground hog day again except it's single 980-it's ok v's ignored higher bus saturation setting limitations with twin 980's.

I'll remind you all when T2 exposes the limitations and destroys AMD's current line up.;)

The only thing that will be getting destroyed, is the 290 series by the 970/980 cards in the Benchmark threads regardless of bus size at any Res.

That's my forecast anyway.
 
Bus width alone is absolutely irrelevant.

Tahiti LE: 1536 Shaders @ 975Mhz, 256Bit Bus (100%)
Tahiti Pro 1792 Shaders @ 850Mhz, 384Bit Bus (110%)

GTX 660TI: 1344 Shaders @ 980Mhz, 192Bit Bus (100%)
GTX 670: 1344 Shaders @ 980Mhz, 256Bit Bus (115%)

Clearly it isn't.
 
Last edited:
Back
Top Bottom