AMD R9 390 Series To Launch Alongside Computex 2015

[GhostRyder]

Likely it will be modified for DX12 but still mostly a rebrand. Without that would be worthless buy.

They sounds like they are doing more than just that but who knows until we have more details... Point though is I doubt this would be a straight up re-badge.

Though not comfirmed, using amd's most efficient silicon which is tonga with 2048 gcn and 256bit bus, that is still about 200watt TDP part. they are doubling # of gcn cores and memory bus. so 390x could be in the 400watt TDP range. Which go back to the rumored AIO water cooler that leaked AMD was lookin' was workin on, that would seem to fit that they would need something along those lines to cools this. It will come down to how well it performance compared to titan X which could figure uses only aroud 225-250watt TDP tops.

Well if scaling in this area was linear then yes, but basing off the predictions the TDP is going to be ~350 but then again TDP is not power consumption and that could be much higher. Depends on how they push it core wise and what the improvements actually yield since this is a new chip. But just basing off the fact they are using an AIO does not mean much because they are likely just evolving to the next area for gaming card reference coolers since most people want everything as quiet as possible.

The average card of the next gen will have more bandwith compared to previous gen, yes, this has been happening for a long time already. Resolution raises, and so does the bandwith needs to raise. But the alledged 600gb/s bandwith? Not yet, not until GPUs can take full advantage of it. It is amazing for SLI/Crossfire users, since it will def be a step in the right direction for 4K.

I'm actually wondering what would be the ideal bandwith for the current "dream", that being 60FPS @ 4K.

Well taking advantage of it means different things, I think though these high end cards are aimed at the extreme range which is where CFX/SLI users and extreme resolution users are. It maybe a bit overkill in the lower areas and yes it will probably be a bit more than needed but this just means memory bandwidth won't be holding it back. Plus we have to start somewhere with new tech...

But how much ram?! I don't think 4Gb is enough for a flagship card since GTX 980 will soon be nVidia midrange card.

I agree, I think 4gb might be the problem overall as its going to be running on the edge. Right now its enough at 4K, but for how long is up in the air and honestly its making the Titan X (or eventual 1080/1080ti or whatever its going to be called) more appealing. Going to be interesting none the less especially if we see some way they manage 8gb versions of the card.

 

[eroldru]

You're worng! Here are the specs for ya:
http://techreport.com/review/26997/amd-radeon-r9-285-graphics-card-reviewed

This speaks for itself and it's taken from techreport
http://techreport.com/r.x/radeon-r9-285/power-load.gif

 

[sergionography]

This whole issue isn't about technology, it is about inventory.
The manufacture of the cards wouldn't be a problem, what is a problem is that AMD under Rory Read stuffed the channel with products to make the books look good. Before any new cards can be accepted AIB/AIC's need to clear inventory. If "new" rebrands enter the channel alongside current SKUs what happens to sales of 200 series cards? The other alternative to order mass recalls, BIOS reflashing, and repackaging which would cut profit to the bone - assuming there was any to be made. The Kitguru article specifically mentions the unsold inventory as being a major reason AMD hasn't released new cards

They are linked. Downsizing a company ( scaling back/cutting projects, cutting down R&D - lest we forget Synopsys gained 150 AMD R&D engineers in their deal), and prioritizing console APU designs have put the company under considerable constraints. It is also the reason that AMD haven't fielded a top-to-bottom new GPU series since Evergreen in 2009-10.

I see where you are coming from and it's part of the reason why it could be new chips. Because if amd are having so much trouble selling this old inventory you think a rebrand of those same exact chips into r300 will make them sell any better? Idk about that. And if amd meant Fiji will gain them market share well that also is unlikely because it's gonna be an expensive top tier card that sure can make them money because of higher margin, but it won't be a high demand card as the numbers usualy go for the sub 300$ cards so there must be something drastic in that sector to even speak of market share.

You mean that they should not put low prices?.

No I meant investing in free open standard developer tools and other perks that don't directly translate to sales. R&D instead at this point should be going solely on new architecture and socs. amd is putting too much effort into software when they only make their money out of hardware so while software is complimentary to the hardware I feel like they r a bit extravagant with it at such rough times.

 

[arbiter]

But how much ram?! I don't think 4Gb is enough for a flagship card since GTX 980 will soon be nVidia midrange card.

Would probably expect maybe 8gb if they are targeting 4k but, if you look at 290x 4gb vs 8gb. Last i checked the 8gb have a 120$ price premium, 310$(4gb) vs 430$(8gb).

They sounds like they are doing more than just that but who knows until we have more details... Point though is I doubt this would be a straight up re-badge.

If they don't add DX12 in to it, well it would be pretty much worthless buy knowing DX12 is coming in new few months after the gpu is out.

Well if scaling in this area was linear then yes, but basing off the predictions the TDP is going to be ~350 but then again TDP is not power consumption and that could be much higher. Depends on how they push it core wise and what the improvements actually yield since this is a new chip. But just basing off the fact they are using an AIO does not mean much because they are likely just evolving to the next area for gaming card reference coolers since most people want everything as quiet as possible.

They are doubling gcn cores, most likely doubling the memory path, so 400watts is very possible less they underclock it which probably what they will have to do. I would say it would be bit more linear on AMD side then would be nvidia. Also going by what hawaii which is 300watts and this new one has almost 50% more gcn cores so.... Those leaked pictures of cooler amd was supposed to working on, would support the idea could be that high.

 

[Prima.Vera]

So basically only the 390 and 390X wont be re-brands?? Or the 380(X) also?

 

[RejZoR]

No, even R9-380 will be brand new silicon. Maybe not as radical as 390/390X, but they won't be just old GPU's with new name in VGA BIOS.

 

[mastrdrver]

Its not all about memory bandwidth. Even with that doesn't mean it will be able to.

Yes, but up until the bandwidth gets addressed, you can throw all the horsepower you want at 4k 60fps and it will not matter. This has been known for a while since multi monitor users have had the same problem.

 

[arbiter]

So basically only the 390 and 390X wont be re-brands?? Or the 380(X) also?

AMD said 380(x) will be matured version of hawaii silicon. So they could do some upgrades to it which would hope means DX12 to it but could be stuck with 11.2

Yes, but up until the bandwidth gets addressed, you can throw all the horsepower you want at 4k 60fps and it will not matter. This has been known for a while since multi monitor users have had the same problem.

As another person said, need the GPU horsepower to be able to take advantage of it.

 

[dj-electric]

AMD said 380(x) will be matured version of hawaii silicon

When AMD say that, they mean nothing.
280X was supose to be "mature". It ended up being far less efficient than HD 7970 becuase all they did is jack the core clock up while taking 0 effort in efficiency improvements.

Same for R9 285.
http://tpucdn.com/reviews/Gigabyte/GTX_960_G1_Gaming/images/perfwatt_1920.gif
It was supose to be a more "mature" silicon and it ended up being less efficient than its 2.5 year-old (at the time) brothers.

That leads me to believe that if a rebranded R9 290 and 290X will come out, they are most likely to consume the same or even more than the already ridicules Hawaii

 

[mastrdrver]

AMD said 380(x) will be matured version of hawaii silicon. So they could do some upgrades to it which would hope means DX12 to it but could be stuck with 11.2



As another person said, need the GPU horsepower to be able to take advantage of it.

AMD said all GCN gpus will support DX12.

I'm not saying bandwidth alone will solve the problem. Though, I'm not disagreeing with the need for GPU horsepower, but without the bandwidth it's not going to help.

 

[DeadSkull]

When AMD say that, they mean nothing.
280X was supose to be "mature". It ended up being far less efficient than HD 7970 becuase all they did is jack the core clock up while taking 0 effort in efficiency improvements.

Same for R9 285.
http://tpucdn.com/reviews/Gigabyte/GTX_960_G1_Gaming/images/perfwatt_1920.gif
It was supose to be a more "mature" silicon and it ended up being less efficient than its 2.5 year-old (at the time) brothers.

That leads me to believe that if a rebranded R9 290 and 290X will come out, they are most likely to consume the same or even more than the already ridicules Hawaii

Last time I saw a gpu architecture really "mature" was nVidia GF100 > GF110 improvement. Lower temp, reenabled shaders, same power draw with higher core clocks.

 

[Sony Xperia S]

Last time I saw a gpu architecture really "mature" was nVidia GF100 > GF110 improvement. Lower temp, reenabled shaders, same power draw with higher core clocks.

That's why new revisions of the silicon are being made. To use an improved process and its potential for general improvements everywhere.

The thing is how much all those GCN GPUs (from the first one) will support DirectX 12.

Maybe only partially?

Have you heard about the upcoming Arctic Islands? Any comments on them?

 

[HumanSmoke]

The thing is how much all those GCN GPUs (from the first one) will support DirectX 12. Maybe only partially?

All versions of GCN support DirectX 12 fully ( Tier 3)

Have you heard about the upcoming Arctic Islands? Any comments on them?

There is no concrete information regarding Arctic Islands at the moment.

 

[Sony Xperia S]

All versions

Ich spreche kein Deutsch.

Seriously though, Arctic Islands are interesting on their own, there are many sites already reporting about them:

http://news.softpedia.com/news/AMD-Prepares-Radeon-R9-400-Arctic-Islands-GPUs-471557.shtml

In any case, the GPUs should be designed for either the 16 nm or 14nm FinFET fabrication process technology, which means much higher efficiency than 28nm.
Globalfoundries will likely handle the manufacturing, since AMD appears to have cut ties with TSMC after repeated failure to ramp new nm nodes on time.

Potential chip names include New Siberian XT, Wrangel XT, Herald XT, maybe even Victoria? There are many islands in the Arctic Ocean. Unless AMD meant that all islands near the arctic circles are up for grabs, in which case we're better off not even trying to guess.

 

[64K]

So basically only the 390 and 390X wont be re-brands?? Or the 380(X) also?

The last thing I read on the 380x is that it will have the same specs as a 290x but with improved efficiency so that leaves room to boost the clocks a little to make it faster than the 290x. The 390/390x will be the most interesting new silicone from AMD but may be pretty expensive at launch. We'll see.

 

[Sony Xperia S]

The 390/390x will be the most interesting new silicone from AMD...

Mhm, come 20nm babies, come.

 

[GhostRyder]

The last thing I read on the 380x is that it will have the same specs as a 290x but with improved efficiency so that leaves room to boost the clocks a little to make it faster than the 290x. The 390/390x will be the most interesting new silicone from AMD but may be pretty expensive at launch. We'll see.

Yea, sounds like they are going to boost it up a bit higher to make it compete with the GTX 980 cards at an improved efficiency though its still going to use more power but will be a close competitor performance wise. I think the question will be though if the boost clocks are all its still going to be capable of or will it overclock even further as that would be what makes it an even better change.

Ich spreche kein Deutsch.

Seriously though, Arctic Islands are interesting on their own, there are many sites already reporting about them:

http://news.softpedia.com/news/AMD-Prepares-Radeon-R9-400-Arctic-Islands-GPUs-471557.shtml

In any case, the GPUs should be designed for either the 16 nm or 14nm FinFET fabrication process technology, which means much higher efficiency than 28nm.
Globalfoundries will likely handle the manufacturing, since AMD appears to have cut ties with TSMC after repeated failure to ramp new nm nodes on time.

Potential chip names include New Siberian XT, Wrangel XT, Herald XT, maybe even Victoria? There are many islands in the Arctic Ocean. Unless AMD meant that all islands near the arctic circles are up for grabs, in which case we're better off not even trying to guess.

Its all mostly rumor at this point, I mean we don't even have much information regarding the new Pirate Islands GPU's coming soon so the Arctic Islands are far from giving us any real information. Would love some of course but that is not going to have much for at least another 6+ months I would guess.

When AMD say that, they mean nothing.
280X was supose to be "mature". It ended up being far less efficient than HD 7970 becuase all they did is jack the core clock up while taking 0 effort in efficiency improvements.

Same for R9 285.
http://tpucdn.com/reviews/Gigabyte/GTX_960_G1_Gaming/images/perfwatt_1920.gif
It was supose to be a more "mature" silicon and it ended up being less efficient than its 2.5 year-old (at the time) brothers.

That leads me to believe that if a rebranded R9 290 and 290X will come out, they are most likely to consume the same or even more than the already ridicules Hawaii

Well that is only if you are comparing it to the 7970 reference instead of the 7970ghz edition, it shows a little bit of improvement over the Ghz edition though not much. 285 does consume less power but that is mostly a similar way of balancing load done with PowerTune similar to how Maxwell handles load more than anything. Though the card was mostly designed to make a cheaper R9 280 is the end of the day which is what they were hoping with the change in ram, bus, and circuitry and they achieved that while offering similar performance. As for the 290X, even if it is matured your probably right because they will likely bump the clocks up to overtake whatever efficiency gains they get immediately.

Meh, ill wait until more details arise before I make final judgments though I do like where this is going...

 

[jabbadap]

This speaks for itself and it's taken from techreport
http://techreport.com/r.x/radeon-r9-285/power-load.gif

Most likely confusing use of tdp,tbp etc. useless marketing power usage spec. Neither amd or nvidia uses real tdp(thermal design power) for marketing, amd uses TBP(typical board power) which is the power card sips average in typical use(i.e. gaming). See tpu:s review(r9-285 tbp 190W, gaming avarage 189W):
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

Nvidia uses some vague term called graphics card power, which usually means peak power while gaming (all boost 2.0 cards from titan to gtx960, except gtx980/gtx970, which uses average)...

 

[Initialised]

I hope we see the 980Ti by then.

Whers da popkorns?

 

[xvi]

Still, I am an impatient man and waiting on anything whether it be a new GPU or CPU just drives me nuts especially when we have been waiting for so long already.

The longer we wait, the more time they have to make improvements. As long as it's a good card, it might be worth it.

 

[EarthDog]

1. HBM... Am I the only one that is indifferent there? 512b bus is seemingly wide enough to pump AA @ 4k. So what will more do if that isn't the bottleneck? It sure won't help 2560x1440 or even 5760x1080/1200. So to me, yippee on the bandwidth...
2. It comes with an AIO. Scary. Scary because I'm afraid it's a necessity and not more of a novelty. I own a 295x2 so I know what 500w of heat will do to radiator...or two. It better not be 300w tdp. Frankly, if it's over 250, I will be disappointed...but it wouldn't prevent me from buying it.

 

[arbiter]

2. It comes with an AIO. Scary. Scary because I'm afraid it's a necessity and not more of a novelty. I own a 295x2 so I know what 500w of heat will do to radiator...or two. It better not be 300w tdp. Frankly, if it's over 250, I will be disappointed...but it wouldn't prevent me from buying it.

It will be 300+watts, even if it uses 20nm which i heard stories of 20nm might not been ready but could be. doing 50% jump in shaders, probably be north of that less AMD under clocks it.

 

[Sony Xperia S]

1. HBM... Am I the only one that is indifferent there? 512b bus is seemingly wide enough to pump AA @ 4k. So what will more do if that isn't the bottleneck? It sure won't help 2560x1440 or even 5760x1080/1200. So to me, yippee on the bandwidth...

Yes, I think so. I cannot wait to see what HBM brings to the table.

 

[GLD]

My main rig in my specs tops out about 275 watts max when gaming. It plays all my games on high, some ultra, with out skipping a beat. Reviews shows my 7850 is a 100w or less card. I couldn't make the jump to a 300w card and fell good about it. A 150w card maybe, as I might step up to a 125w 8350 cpu, because they do kick arse for the money.

My 2nd rig, my daily driver pulls less then 65w all day long. Now that is what I am talking about. AMD FTW.

Anyway... I am sure the next gen of AMD cards will sell great.

 

[EarthDog]

Yes, I think so. I cannot wait to see what HBM brings to the table.

You have more than 1 4k monitor or 3 1440p ones to actually have a horse in the race? Awesome! Otherwise, it really doesn't matter for you friend...

What's the point of using a firehose when a gardenhose allows adequate flow. It won't flow faster just because it's a bigger pipe.