R9 380x rumor and speculation thread

[Flapdrol1337]

Does it matter if its 384 or 256 bit bus? It's bandwidth compression tech and fast GDDR5 is ample for the performance target, potentially 380 + 15%? That can't be a bandwidth limited scenario.

I don't think 256 bit will be a problem. If it is they can probably push the clocks a bit, 290 launched with 5 "GHz" memory, 390 is at 6.

 

[happy medium]

What if 380x is full Tonga and performs at around 90% of a 390. That would be an interesting card at $239.99

I would definitely be interested in a card like that.

looks like it is full Tonga, but is it 256 or 384 bit?
http://www.guru3d.com/news-story/amd-r9-380x-tonga-spotted-with-2048-shader-processors.html

 

[LTC8K6]

I have been waiting to buy this card since the 285 came out.

 

[f2bnp]

I seriously hope this isn't a Tahiti XT rebadge and we finally get Tonga XT.

 

[Rvenger]

If you compare the XFX R9 380 and 380X leaked pictures, you will notice they both have the clip on fans and the same PCB. My guess is either full tonga or cut down but higher clocks like the R9 270 and 270x were. Definitely Tonga though.

 

[Elixer]

If you compare the XFX R9 380 and 380X leaked pictures, you will notice they both have the clip on fans and the same PCB. My guess is either full tonga or cut down but higher clocks like the R9 270 and 270x were. Definitely Tonga though.

Looks like the PCB would support a bigger heatsink, looking at the holes that aren't used...(or maybe a backplate?)

 

[f2bnp]

If you compare the XFX R9 380 and 380X leaked pictures, you will notice they both have the clip on fans and the same PCB. My guess is either full tonga or cut down but higher clocks like the R9 270 and 270x were. Definitely Tonga though.

You're right, it can't be Tahiti, which is assuring. But I don't think the market needs a higher clocked 380 either. A Tonga XT with 4GB priced at the right price point would be utterly killer.
Can't wait to see what those buggers can really do compared to GTX 970/R9 Hawaii

 

[burninatortech4]

Why would they design tonga with 384bit capability and not use it?

I agree that 256bit and 4gb ram is much more likely. I hope it (overclocked) benches 90% of a 390 for ~$250. That will be my next card.

 

[Rvenger]

Why would they design tonga with 384bit capability and not use it?

I agree that 256bit and 4gb ram is much more likely. I hope it (overclocked) benches 90% of a 390 for ~$250. That will be my next card.

Tonga is ROP limited so it will see little to no benefit on 384bit. It also costs less to put on 256bit interface with 2gb less VRAM, especially if they price at $250-$270.


Also, reading what others are saying around the net, Tonga should help AMD in their falling marketshare since the $250 price point is the sweet spot for most PC gamers, this is what they needed and the full Tonga is like a brand new card not a refresh since it was only used for Apple.

 

[3DVagabond]

Why would they design tonga with 384bit capability and not use it?

I agree that 256bit and 4gb ram is much more likely. I hope it (overclocked) benches 90% of a 390 for ~$250. That will be my next card.

Tonga is ROP limited so it will see little to no benefit on 384bit. It also costs less to put on 256bit interface with 2gb less VRAM, especially if they price at $250-$270.


Also, reading what others are saying around the net, Tonga should help AMD in their falling marketshare since the $250 price point is the sweet spot for most PC gamers, this is what they needed and the full Tonga is like a brand new card not a refresh since it was only used for Apple.

Which is exactly why they'll price it @ $299.

 

[Elixer]

Doesn't this boil down to a consumer version of the AMD FirePro W7100 card then?
So the specs would be something along the lines of

Memory Bus: 256 bit
Bandwidth: 160 GB/s
Shading Units: 1792
TMUs: 112
ROPs: 32
Compute Units: 28
Pixel Rate: 29.4 GPixel/s
Texture Rate: 103 GTexel/s
Floating-point performance: 3,297 GFLOPS

So, it could be 6G or 4GB, hoping it is 6GB (or 8)

 

[arandomguy]

Doesn't this boil down to a consumer version of the AMD FirePro W7100 card then?
So the specs would be something along the lines of

So, it could be 6G or 4GB, hoping it is 6GB (or 8)

28 CUs is what we have now in the 285/380.

 

[Elixer]

28 CUs is what we have now in the 285/380.

Right, and the M295X had 32, but, I don't think that was a full tonga either.
So, it needs to be higher than the 285 & 380, but lower than the 290 which is 44, that could mean we are looking at 40.

 

[arandomguy]

m295x is speculated to not be full Tonga because of the memory bus due to some questions over what the right side of the leaked die shot is supposed to represent.

http://forums.anandtech.com/showpost.php?p=37469786&postcount=304

But the die shot does show what appears to be 32 CUs which would make that the upper limit.

In terms of real world performance differences we can probably use Tahiti LE comparisons as a best case scenario there is indeed a 384-bit bus. (https://www.techpowerup.com/reviews/Club_3D/HD_7870_jokerCard_Tahiti_LE/28.html). Bear in mind Tahiti LE had even less CUs at 24.

If bus width is the same then probably 7950 to 7970 in terms of performance gains. (https://www.techpowerup.com/reviews/AMD/HD_7950/27.html)

 

[Red Hawk]

Better late than never. Just hope it's an actual full Tonga.

 

[Chicken76]

Or maybe we'll see two variants:

• Variant 1: 2048 shaders 256bit 4GB targeting price-conscious 1080p market.
• Variant 2: 2048 shaders 384bit 6GB targeting 1440p. It will be more efficient and cheaper to produce than Hawaii.

 

[burninatortech4]

Or maybe we'll see two variants:

• Variant 1: 2048 shaders 256bit 4GB targeting price-conscious 1080p market.
• Variant 2: 2048 shaders 384bit 6GB targeting 1440p. It will be more efficient and cheaper to produce than Hawaii.

Variant one is much more likely. Hopefully we see it priced around ~$240 and have it bench at about 80-90% of a 390 when overclocked. It should crap on 4gb 960's in the same price bracket.

 

[3DVagabond]

Or maybe we'll see two variants:

• Variant 1: 2048 shaders 256bit 4GB targeting price-conscious 1080p market.
• Variant 2: 2048 shaders 384bit 6GB targeting 1440p. It will be more efficient and cheaper to produce than Hawaii.

AMD's 256bit MC is a far more efficient design than their 384bit. Surely don't need more than 4GB for 1440.

 

[redzo]

I am thinking that the 380x will put the 280x to sleep for good and maybe manage to touch a stock 290 a little if they increase the stock clock higher than 380's. The 290 needs to go also as it is disruptive for the current AMD gpu line. The 290x will eventually go away as it is being eaten alive price wise by the 390.

 

[3DVagabond]

With the relative performance increases the 300 series is showing above the 200 series, it's likely we'll get close to 290 performance but with lower power consumption.

 

[RussianSensation]

I am thinking that the 380x will put the 280x to sleep for good and maybe manage to touch a stock 290 a little if they increase the stock clock higher than 380's. The 290 needs to go also as it is disruptive for the current AMD gpu line. The 290x will eventually go away as it is being eaten alive price wise by the 390.

What do you mean 290 needs to go? It's an amazing value right now, for sure way better than a $220-250 R9 380X would be.

Computerbase has an after-market 290 beating a 380 by 55% at 1080P HQ.

Sweclockers uses a reference thermal throttling 290 in their charts (an after-market version would = 290X on their chart):

34% faster than an after-market 380, and if we use 290X reference to indicate an after-market 290, we get 43% more performance at 1080P:

Naturally because 290 has ample memory bandwidth, 64 ROPs it performs even better with higher GPU loads/resolution. Using the same as above, a reference 290 beats a 380 by 37% and an after-market 290 would win by 47%.

Considering it's still possible to buy an after-market 290 in the US for $230 with Dirt Rally too, there is no way a 380X at $220-250 would be a good deal. Even at $200 it would still be a bad deal.

It's amazing how certain PC gamers wait for 10-12 months to get some marginal increase in performance in a mythical GTX960Ti or R9 380X when we have had 12 months of incredible R9 290 deals.

Even if 380X eventually comes out, it's never going to be as good as the 290 deals that were available over the last 12 months around many countries in the world.

We are also starting to see GTX970 regularly drop to $280, R9 390 has also been on sale for $270. It's just a matter of time before these cards move down to the $250-260 price levels. R9 380X needs to launch ASAP because every day it's not on the market is every day we get closer to $250-260 R9 390/970 cards at which point a $200-240 R9 380X will be DOA. AMD should have launched the R9 380X a LONG time ago.

Having a $269-279 380X starting Sept 2014 would have allowed AMD to drop prices and stay competitive with a 960 onslaught but instead they are contemplating launching such a product near the end of the 28nm generation? Yawn.

 

[3DVagabond]

@Russian: People fall for newer is better all of the time. It's what killed AMD's market share. People could have been buying 290/X's all along but they'd rather pay more for 390/X's. They need to come out with it because people will buy it and they'll make more money than they do on a similar priced 290. Even if it's slower. Plus green team won't have to deal with the better perf/$ arguments where 290/X's are concerned.

 

[redzo]

What do you mean 290 needs to go? It's an amazing value right now, for sure way better than a $220-250 R9 380X would be.

Sweclockers uses a reference thermal throttling 290 in their charts (an after-market version would = 290X on their chart):

34% faster than an after-market 380, and if we use 290X reference to indicate an after-market 290, we get 43% more performance at 1080P:

I don't challenge 290's value. In fact, I support/ed it a lot at that budget.
My point is that AMD has to replace it with something cheaper to manufacture in order to optimize/maximize profits. I challenge 290's capability to achieve that; it is probably a lot more expensive both chip and board wise to manufacture than a possible 380x.
edit: wow! I haven't read this before posting. You're a mind reader.

@Russian: People fall for newer is better all of the time. It's what killed AMD's market share. People could have been buying 290/X's all along but they'd rather pay more for 390/X's. They need to come out with it because people will buy it and they'll make more money than they do on a similar priced 290. Even if it's slower. Plus green team won't have to deal with the better perf/$ arguments where 290/X's are concerned.

 

[raghu78]

I don't challenge 290's value. In fact, I support/ed it a lot at that budget. My point is that AMD has to replace it with something cheaper to manufacture in order to optimize/maximize profits. I challenge 290's capability to achieve that; it is probably a lot more expensive both chip and board wise to manufacture than a possible 380x.

well said. R9 290, while being a very good deal for the consumer is not a good sale for AMD. AMD AIBs are selling through their inventory of R9 290 by price cuts and discounts. A 438 sqmm GPU on a 512 bit memory bus PCB being sold for USD 250-280 is really bad for AMD. R9 380X 4 GB will launch at USD 240-250 and be on par with R9 280X 3GB. It will beat the GTX 960 OC easily at 1080p and stomp it at 1440p. Power efficiency will be R9 380X's weakness against GTX 960 OC as we all know already that Maxwell is vastly more efficient.

http://www.computerbase.de/2015-08/...h/2/#diagramm-rating-1920-1080-hohe-qualitaet

 

[RussianSensation]

@Russian: People fall for newer is better all of the time. It's what killed AMD's market share. People could have been buying 290/X's all along but they'd rather pay more for 390/X's. They need to come out with it because people will buy it and they'll make more money than they do on a similar priced 290. Even if it's slower. Plus green team won't have to deal with the better perf/$ arguments where 290/X's are concerned.

Ya but then 380X will be crapped on for not having HDMI 2.0, 4K HVEC, and worse perf/watt against Maxwell. At least with a 290 the consumer gets amazing performance for the price. NV loyal customers won't move over to a 380X as they'll either save money and get 960 or pay more for a 970. The only way is if AMD offers more of everything which means they would need to price 380X 4GB at $159. Only then, NV loyalists might even consider it over the 950/960.

Think about it, all the same marketing that sells 950/960 will be intact against the 380X, but 380X will be even slower than a 290. How in the world would it prove to be more successful in sales than a $250 290 that failed to do so for 8-9 months?