380X is 28nmThat's the point. If the leak was from anyone else, it would be too easier to dismiss. But ChipHell was THE only site wth 3 consecutive generational leaks that were pretty damn close: Tahitis preview launch numbers (7950/7970) and even the most recent Volcanic Islands R200 series and Maxwell GM204 launch turned out to be quite accurate in terms of the performance numbers. Actually the GM204 leaks were underclocked engineering samples and the final cards performed even better than the leaks from ChipHell.
I think people fell for the GM 204 hype. While Maxwell is undoubtedly an awesome architecture from a perf/watt point of view, 970/980 aren't that awesome performance wise vs. 290X at high rez gaming.
If AMD bumps performance just 30-35% over 290X, you are already 15-18% faster than the 980. NV's 970/980 look really good because of the negative stigma of reference 290/290X, and because they are new tech with newer features and less power usage. As far as performance goes, they have hardly moved the bar. NV's performance also looks much better at 1080p/1200p/1440p which are the more popular resolutions.
NV hasn't even shown off the big guns. They are just milking the mid-range GM204 until they have to release 780/780Ti successors. I think 390X would need to be 25% faster than a 980 to compete favorably against the lower SKU of GM200. There is a lot more pressure on AMD with 390 series because let's face it 970/980 are really 960/960Ti.
Sounds hard to believe without 20nm + HBM + very large die -- a trifecta very unlikely this early in 20nm cycle. If true, it would sell out right away.
I'm in the same boat. They've had a good track record which means I'll be paying close attention with wallet in hand come Q1 2015 (February last I heard).That's the point. If the leak was from anyone else, it would be too easier to dismiss. But ChipHell was THE only site wth 3 consecutive generational leaks that were pretty damn close: Tahiti’s preview launch numbers (7950/7970) and even the most recent Volcanic Islands R200 series and Maxwell GM204 launch turned out to be quite accurate in terms of the performance numbers. Actually the GM204 leaks were underclocked engineering samples and the final cards performed even better than the leaks from ChipHell.
Ok so if chiphell is right at least perf/watt improves even slightly beyond nv. If thats the case doesnt that imply something new besides the hbm? Voltage regulation tech ...? I asume its still 28nm. What about gf soi?
All i care about is it being faster then a 980, performance needs to move forward.I'm not a hippy who needs a 150w tdp card that runs faster then a 970 while giving you butterfly kisses and a diet coke.I just want 980+ performance.
If this is true and launches for 350 or less with 4gb RAM I will be trading out the 290s I just got sooner than expected. Which is fine since I got 4 games with them that I wanted and 2 for 500 anyway.
All i care about is it being faster then a 980, performance needs to move forward.I'm not a hippy who needs a 150w tdp card that runs faster then a 970 while giving you butterfly kisses and a diet coke.I just want 980+ performance.
:biggrin:
While I do have a 970 on the way for my HTPC (long story short, it's the only card that fits the case) I agree with this.
I wouldn't mind efficiency taking the backseat for performance parts, I'll build a 4K system once SLI/crossfire can handle it without major compromises that are necessary now.
I honestly didn't even factor in the efficiency of the 970 in the purchase but heat output+cooling solution=physical dimensions aligned it with my quest for a best performing, < 10.3" length GPU.
These benchmarks and power consumption numbers look very competitive; more efficient than even GM204 if true.
GCN 2 needs to beat Maxwell on efficiency if AMD hopes to recover the all important notebook marketshare.
On the Desktop, that efficiency gain would be best put to use with a more aggressive Boost system to extract more performance.
If GCN 2 lacks efficiency, it really is all over for AMD. They won't be able to stay relevant with another architecture generation and many years of eroding marketshare.
I think AMD was bleeding market share even when they had better, more efficient designs. It will be nice if AMD once again comes up with something more efficient. This chip is quite promising, but AMD may keep losing notebooks though I think a reversal in desktop may happen. It can't come soon enough for both AMD and customers.Even parity would be a good start, especially as they're both on the same manufacturing node. NVIDIA has been touting notebook gaming growth for quite some time; I have long suspected that the efficiency focus we saw with Kepler and continuing into Maxwell is due to the company going after notebooks and HPC as the primary design targets -- both markets that highly value power efficiency.
The benchmarks here suggest that AMD has indeed delivered on actually beating NVIDIA on power efficiency. I'm very interested to see how the chips perform once they're launched.
I think AMD was bleeding market share even when they had better, more efficient designs. It will be nice if AMD once again comes up with something more efficient. This chip is quite promising, but AMD may keep losing notebooks though I think a reversal in desktop may happen. It can't come soon enough for both AMD and customers.
AMD already missed 2015 for major notebook design wins. 860M/970/980M will dominate notebook sales. Even if 300 series beats Maxwell in perf/watt, the next time AMD will be able to bid for major wins is for 2016.
Being late and now showing at all (HD8000/200M) is far worse than being behind in Perf/watt. There were plenty of excellent HD8770/8790M and various Pitcairn designs that were head-to-head in perf/watt against mobile Kepler but it didn't matter since AMD was so late with their launch, all the major designs were won by NV by default of AMD not bidding with GCN, relying on ancient VLIW chips.
Right now AMD lost the 2015 laptop market completely. Their best hope is to have an efficient GCN 2.0/HBM design ready as a foundation for 2016 product stack.
^ Look at laptops such as MSI Gaming GS60/70, Maingear, Origin, MacBook Pros or Alienware. They get refreshed 1-2 times a year at the most. That means for the next 6 months, whoever locked in design wins with them won't be kicked out even if 300 series launched January 1, 2015.
NV have extracted the most they can out of the 28nm node. If the leaked chiphell benchmarks are an indication of R9 380X with 3072 sp then Nvidia has nothing to beat the R9 380X when it launches in notebooks.Broadwell-K and Skylake are well behind schedule too so it won't be until 2nd half of 2015 most likely until there are new laptops with the newer Intel chips in large volumes. By June 2015, GM204 will be 10 months old which means NV can re-spin a faster version of 970/980M refresh and/or lower prices. Whatever minor gains R9 300 brings in mobile, by the time new gaming laptop designs roll out in 2H of 2015 with Broadwell, NV will have plenty of time to respond and take the wind out of AMD's sails.
Sorry but I think it all boils down to R9 390X vs GM200 and R9 380X vs GM204. I am thinking that GCN 2.0 with HBM is going to bring AMD back into competition in gaming desktops and notebooks. I think you are going to be surprised at how much the HBM time to market advantage is going to work in favour of AMD.OTOH, the desktop is completely different as people don't need to wait until BW/Skylake to buy a new GPU, unlike laptop buyers which might wait for major architecture changes when timing their new laptop purchases. Since on the desktop side NV has still not brought out 960/960Ti or GM200, and they have major holes in their product stack in the $120-300 and $350-500 levels, R9 300 series has a greater opportunity to strike back on the desktop.
I agree that the laptop market is more important for GPU market share but given that AMD couldn't even secure HD7970M/8970M in many high-end Intel-powered laptops, despite both of those chips being extremely competitive in price/performance and performance/watt, AMD's strategy/relationships/approach in getting OEM design wins in the mobile dGPU sector is clearly broken.
I agree that the laptop market is more important for GPU market share but given that AMD couldn't even secure HD7970M/8970M in many high-end Intel-powered laptops, despite both of those chips being extremely competitive in price/performance and performance/watt, AMD's strategy/relationships/approach in getting OEM design wins in the mobile dGPU sector is clearly broken.