AMD: Pirate Islands (R* 3** series) Speculation/Rumor Thread

It only makes sense to bring out a top to bottom new line of GPUs on the same process if you can significantly improve perf/mm2 and perf/W. The first Fiji numbers are inconclusive and a bit contradictory, but they don't point to a slam dunk. And there's no point refreshing a bunch of silicon just to add FreeSync support.
It does mean that AMD won't be able to play in the discrete notebook space, but they don't have much to lose there to begin with.
AMD needs to be very careful with their resources. Better to spend it straight on 16nm.
 
It only makes sense to bring out a top to bottom new line of GPUs on the same process if you can significantly improve perf/mm2 and perf/W.
When is the last time AMD did a "true" top to bottom new line of GPU's?
 
When is the last time AMD did a "true" top to bottom new line of GPU's?

Forget that, when was the last time any GPU manufacturer did a "true" top to bottom new line of GPUs?
 
When is the last time AMD did a "true" top to bottom new line of GPU's?
3 years ago?
But why is that relevant? If they're unable to come up with something that significantly improves on the current specs on the same process, there's no point in refreshing.
 
3 years ago?
But why is that relevant? If they're unable to come up with something that significantly improves on the current specs on the same process, there's no point in refreshing.
Just wondering since I wasn't sure.
Does this mean manufacturers of freesync monitors will only be able to advertise the "freesync" advantage with the R9 390x? Or is this irrelevant?
 

Attachments

  • upload_2015-3-22_19-21-33.png
    upload_2015-3-22_19-21-33.png
    1.1 MB · Views: 42
Last edited:
Forget that, when was the last time any GPU manufacturer did a "true" top to bottom new line of GPUs?

Nvidia is currently doing that with the 900 series. The entire series is maxwell..

If AMD 300 series is mostly a series of rebrands, it will be one of the most ill featured as far as confusion goes ever, considering the mix of chips with truesound, directx 12, freesync compatible in the various forms of gcn 1 to 1.3.

In otherwords, AMD could take the rebranding crown.
 
Nvidia is currently doing that with the 900 series. The entire series is maxwell..
Maxwell could be counted as two architectures, as GM1xx and 2xx don't share the same feature sets. GM1xx's are used along GM2xx's in mobile 900's and desktop 700's - of desktop 900's become top to bottom series, I'd bet the lowend will be GM1xx's.
 
Does this mean manufacturers of freesync monitors will only be able to advertise the "freesync" advantage with the R9 390x? Or is this irrelevant?
The manufacturers of the monitors are making a monitor that's FreeSync compatible. Whether or not the GPU can do it isn't their problem. And Tonga and Hawaii (and others?) can do FreeSync as well. I don't think there's a major overlap between the buyers of low end rebrands and buyers of FreeSync monitors.

Maxwell could be counted as two architectures, as GM1xx and 2xx don't share the same feature sets. GM1xx's are used along GM2xx's in mobile 900's and desktop 700's - of desktop 900's become top to bottom series, I'd bet the lowend will be GM1xx's.
The low end Maxwells don't have all the features of their high end brethren, but they share the perf/W and perf/mm2 improvements, and the SM architecture, which IMO the most important characteristic of a GPU family.

If AMD 300 series is mostly a series of rebrands, it will be one of the most ill featured as far as confusion goes ever, considering the mix of chips with truesound, directx 12, freesync compatible in the various forms of gcn 1 to 1.3.
As long as games are created with the lowest common denominator in mind, it only matter for the checkmarks on the box. All GPUs will support DX12, just not some HW feature levels that won't be used for a couple of years to come anyway. If Nvidia owners can survive without TrueSound, an AMD owner should be able to as well. And the high-end will support FreeSync, so that's ok too.

I'm not defending AMD for not having better technology (what have they been doing in last 3 years???), but if (!) this was the hand they were dealt, it's not unreasonable at all to do some rebrands.
 
Last edited by a moderator:
Forget that, when was the last time any GPU manufacturer did a "true" top to bottom new line of GPUs?
Well, I do remember that Nvidia did it with Geforce FX... :p (Fat lot of good that did any of us though, heh! ;))
 
Huh, saw a completely different rumor that specifically stated the 370 would be an all new card with a 6" PCB clocked at 925mhz. At this point I suspect there's just a bandwagon of false rumors that people have generated just to be on the bandwagon too. I'm not saying this is false, it could be true for all I know. But so could an entire refresh or etc. etc.

May just have to wait till June to actually find out, though as far as I know losless delta color compression isn't something that's a huge cost to implement in hardware and does have a huge benefit. So I personally wouldn't see a reason not to implement it on nearly every card, especially as it can be used to reduce costs for performance, as Nvidia has done across its line so far (Nee GTX 960 with a 128bit bus).
 
Yeah, here's another set of rumors from less than two weeks ago: http://www.maximumpc.com/xfx_radeon...leaks_web_higher_end_r300_series_cards_follow

The only thing consistent about any of these leaks, rumors, and what have you is that they are anything but consistent. It will and won't be an entire lineup refresh, the 390x will have anywhere from 4 to 16 gigs of ram, and the secret moon donkey may or may not be appearing at midnight. I like speculation and rumor but at this point I think both sites and people know this stuff will get clicks so they'll both put up anything at all just to get them.
 
Fiji XT with 3584 SPs? R9 390 with GDDR5? (Notice that the R9 390X has 4 GB "memory" instead. :D)

I'm going to guess that the Fiji part of the rumor is either 100% true or completely false (except for "obvious" parts like the 28 nm), most likely the latter. The rest of the lineup seems reasonable though.
 
That isn't really compatible with these news:
http://wccftech.com/xfx-radeon-r9-3...gddr5-vram-launches-april-2015/#ixzz3U4TxXZhL
I can't see how they'd lower Pitcairn's power consumption by at least 30% without somehow changing the chip. PCB optimization would only get you so far.


R370 being Pitcairn a.k.a. HD7870 a.k.a. R9 270 would be terrible for AMD's midrange.
If having the same chip for the same performance segment for 3 years in a row wasn't bad enough, then doing it for 4 years will mean nVidia may go with 4 different chips to tackle it: GF114 on release, GK106/GK104 a few months later, GM206 right now and a Pascal successor in early 2016.
 
That isn't really compatible with these news:
http://wccftech.com/xfx-radeon-r9-3...gddr5-vram-launches-april-2015/#ixzz3U4TxXZhL
I can't see how they'd lower Pitcairn's power consumption by at least 30% without somehow changing the chip. PCB optimization would only get you so far.


R370 being Pitcairn a.k.a. HD7870 a.k.a. R9 270 would be terrible for AMD's midrange.
If having the same chip for the same performance segment for 3 years in a row wasn't bad enough, then doing it for 4 years will mean nVidia may go with 4 different chips to tackle it: GF114 on release, GK106/GK104 a few months later, GM206 right now and a Pascal successor in early 2016.

Well, they could sell them for peanuts, like back in the days with the Radeon HD 4870/4890.
 
Back
Top