Ah, I see you've edited your post after-the-fact to try to cover it up. No further need to go into it, I see how you are...
I hope you all realize the R285 is actually the cut down Tonga, the full die Tonga is all being sold to Apple for Macs. Thus the 285 is more akin to the 7950, it should have no right beating a 7970Ghz.
But there are improvements in there which when taken advantage of by games (Bioshock, FC4 with enhanced tessellated godrays), lead to a major performance gain compared to Tahiti.
If all R390X is was a 25% bigger Hawaii with Tonga's front end improvements (more throughput & better tessellation), memory compression and HBM (saving die space & TDP) low latency/high bandwidth vram, it should easily be ~50% faster than R290X.
ps. The R285 is still a crap product, at that price, that performance, and 2GB vram. No go either way. I feel the same about the 960, eventhough it has better efficiency. For a bit more, one can get the R290, and some more, the 970.
Ohh lala - Looking deep into his mind ....
(Edit what do you see in the mist? A father obsessed with gpu and a caring mother trying to call for supper ?)
Last edited:
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
Nope. I never edited anything regarding HD7970 vs. 285. For anyone who followed this thread closely, I never once tried to claim that 285 is better than an HD7970 overall OR on average. The discussion was on showing that AMD improved GCN already on the same 28nm node and I provided proof of that. Bioshock Infinite isn't the only game where Tonga's architectural improvements show up. If you read reviews on Tonga you would have remembered this.
R9 285 beats HD7950 Boost by up to 30% in Thief and it bests the R9 280X which shouldn't even be possible because on paper specs 280X wins in everything. Obviously not all games bottleneck a card the same way which is why we don't see the situation of Bioshock Infinite or Thief in all games. But what this tells us is that if AMD includes the colour/memory bandwidth compression, geometry improvements and pixel fill-rate efficiency into future R9 300 products around a chip that has way more SPs and TMUs than Tonga, these architectural advantages that Tonga incorporated are more likely to matter in a larger GPU that won't be bottlenecked as much by shader and TMU performance.
Thief uses Phong Tessellation.
Exactly. In situations where R9 280X/290X cards become geometry limited, their performance tanks. Enhanced Godrays have the benefit of using tessellation to render in FC4 and 290X suffers a 17.5% performance hit.
R9 285 only experiences a 4% performance hit when moving from Volumetric God Rays to Enhanced God Rays.
Just because most games aren't like Bioshock Infinite, Thief and FC4 doesn't mean AMD can just increase TMUs, SPs ROPs and forget about the other aspects of GPU design. Doubling geometry performance isn't free. You need transistors to do it.
R9 285 has 2.15X the geometry performance of the HD7950 Boost despite almost the same die size. If AMD wants R9 300 series to last 2-3 years, they need to address various weaknesses of GCN 1.0/1.1 design and Tonga showed AMD has focused on some weaknesses already. Why to this date Tonga's architectural improvements keep getting ignored in an R9 300 thread is something I can't explain.
This means R9 390X should double R9 290X's tessellation performance on perf/mm2 basis which is a big deal for future games. Tonga keeps up with R9 280 despite having just 176GB/sec memory bandwidth. This+HBM should fix any memory bandwidth bottlenecks R9 390X could have had. HBM would ensure even more efficiency memory controller and huge improvement in bandwidth/mm2. AMD already doubled the ROP count with R9 290X.
That leaves AMD with SPs, TMUs, ROP efficiency, L2 cache and Asynchronous Compute scheduling as major areas of improvements. This is a lot of work for 390 but that means the engineers no longer need to worry about memory bandwidth and tessellation. Since AMD doesn't have the financial resources, this is probably why they used Tonga as a test-bed for some of these improvements, while leaving the remainder for R9 390 series launch.
The 1792 SP Tonga is just 1% behind GTX680 and 5% behind GTX770 at 1440P, despite 176GB/sec memory bandwidth. Amazing how people call this chip a failure when it costs $180 today but GTX680 was $500 and GTX770 $400. While I would not recommend 285 for its 2GB of VRAM and worse price/performance vs. R9 280X/290, Tonga was a decent chip on its own if we remove its inadequate VRAM and way out of line $249 launch price.
Last edited:
The prior launches from gf have been delayed so i think thats a obvious place it could go wrong.
Amd amd ati have prior been good at adopting new nodes at tsmc - think eg 4770. And they seems to have worked on hbm for a long time go eg read the old sa articles.
So my bet is, the guilty is the butler.
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
It's unfortunate that PC gamers have such a short memory. AMD often uses niche and low volume products to test out new things, whether its new memory, newer node or architectural changes. HD4770 would get absolutely destroyed by an HD4890 series card in terms of perf/transistors and most PC gamers don't know that the Radeon HD 4770 was far from a simple affair as the GPU had a transistor count of 826 million, or approximately just 14% less than the Radeon HD 4890, ATI’s most complex GPU at that time.
On paper HD4770 was a terrible product compared to HD4850/4870/4890. Despite a lower node and transistor complexity approaching HD4890 (!) the 4770 card lost to the HD4850 in performance. However, without HD4770 40nm, HD5800 wouldn't be as good as it was because it gave AMD a 2nd chance to optimize 40nm. As a $249 2GB card, Tonga failed but there are hidden things in Tonga that imo are important for AMD's future products.
Last edited:
very true. AMD has laid the foundation for performance scaling in R9/R7 3xx. Now with GCN 2.0 they need to introduce some significant architectural efficiency improvements by making big leaps in perf/sp, perf/watt and perf/sq mm. :thumbsup:
People who think AMD are planning/trying to move away from GCN any time soon haven't been reading the memos. It's nothing to do with being able to afford anything. AMD plans on continued improvement with GCN, but there is no new arch planned any time soon.
you are right. As Raja Koduri said GCN is the most scaleable GPU architecture built. AMD needs to keep making signficant architectural efficiency improvements at a more regular cadence. A yearly architectural revision with significant gains in perf/sp , perf/sq mm and perf/watt (atleast 15 - 20%) not counting breakthroughs from HBM and 2.5D die stacking is required.
The scary part is Nvidia has achieved all this power efficiency without HBM and when they bring HBM to their GPUs by late 2016 they will get another massive boost in perf/watt. The massive power efficiency gains due to HBM combined with Maxwell's bandwidth efficiency (third gen color compression) and further improvements which Nvidia will make in Pascal means AMD needs big improvements in 2015, 2016 and 2017. The Pascal big die flagship will be here by H2 2017 and that will be a monster GPU with 1TB bandwidth and 16 GB HBM. AMD needs to expect even better than Maxwell efficiency in perf/watt and perf/sq mm on upcoming Nvidia GPUs and thus plan and design their products accordingly.
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
As long as AMD can provide good options in the low-end, mid-range and mid-high end for future 300/400/500 series, they will be fine. If AMD instead goes for the performance crown at all costs and neglects 95% of sub-$450 dGPU cards, that will be a dangerous strategy.
Being late with R9 200 and 300 series is proving very costly. AMD may need to rethink its future launch strategy and switch to smaller gains on a per generational basis but have something new every 12 months. Because NV adopted the "bifurcating a generation strategy" since Maxwell, AMD cannot afford to wait 15-20 months between new top-to-bottom series releases. NV showed with 670/680 and with 750/750Ti that you don't need to wait 6+ months to launch everything top-to-bottom. You can just release newer products when they are ready and slowly fill your line-up over time. AMD needs to use that lesson starting with 400 series. If you miss a gamer's upgrade path, you are going to be waiting 2-4 years when the next time comes for his/her upgrade. NV understands this better than AMD.
AMD got lucky that Skylake got delayed to August 2015. This means they have a legitimate shot of getting into OEM desktop and laptop design wins with R9 300 cards around back-to-school season. Windows 10 should finally get people with ancient Windows XP rigs to upgrade to Skylake+W10. That's why it's important that R9 390 is not the only new product in 300 series.
Being late with R9 200 and 300 series is proving very costly. AMD may need to rethink its future launch strategy and switch to smaller gains on a per generational basis but have something new every 12 months. Because NV adopted the "bifurcating a generation strategy" since Maxwell, AMD cannot afford to wait 15-20 months between new top-to-bottom series releases. NV showed with 670/680 and with 750/750Ti that you don't need to wait 6+ months to launch everything top-to-bottom. You can just release newer products when they are ready and slowly fill your line-up over time. AMD needs to use that lesson starting with 400 series. If you miss a gamer's upgrade path, you are going to be waiting 2-4 years when the next time comes for his/her upgrade. NV understands this better than AMD.
AMD got lucky that Skylake got delayed to August 2015. This means they have a legitimate shot of getting into OEM desktop and laptop design wins with R9 300 cards around back-to-school season. Windows 10 should finally get people with ancient Windows XP rigs to upgrade to Skylake+W10. That's why it's important that R9 390 is not the only new product in 300 series.
Last edited:
It was unfortunate, the delay, but on could hardly blame AMD for the same. First, it was the process which delayed 290s, which were supposedly to be on a smaller node. Now again with the 3xx series, it became clear some time ago that nothing smaller than 28nm could be used now, and given that the turn around time hasn't been bad really. It hurt them, but that's life. What hurt them more is the fact that benchmarketeers spun the reference cooling to be really bad, affecting sales a bit. What was worse was the people who couldn't anticipate demand with the coinhunt.
You caught him cherrypicking again. :awe:
And to add to it:
285 loses to 7970 in performance/watt as well.
Wholy smokes. Clearly there is a wall full of make believe.
I can not believe that people are actually supporting post with such bogus..........claims!!!
But he is right. I guess that's why the gears have shifted to attacking the poster
Last edited:
waffleironhead
Diamond Member
- Aug 10, 2005
- 7,054
- 561
- 136
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
You are now the 3rd poster in this thread for whom the architectural comparisons of 285 vs. 7970 flew well over your head when the discussion had 0 to do with 285's perf/watt vs 7970 or their respective average performance standing.
As I said to others who clearly missed the whole point of Tonga
vs. 280x/290x comparison that you take some time to carefully read the last 3 pages of this thread to actually understand what's being discussed. You should actually read what's stated and not assume things and ideas that aren't even being debated.
He was 100% incorrect since his posts couldn't at all comprehend how AMD could improve with R9 300 series on the same node. He doesn't even acknowledge real architectural advancements in Hawaii's memory controller or Tonga's tessellation, colour/memory compression. His rebuttal of linking 285's perf/watt is a waste of time because 285 is not a fully unlocked product and because 390X will have a completely different memory controller and memory type. Basically he wasted space by posting 285's perf/watt as some "proof" that AMD cannot improve perf/watt with 390 cards by resting his argument on the point that 285 couldn't improve upon 7970. What happens when AMD takes all the architectural advancements in Tonga, add far more efficient HBM on a bandwidth/mm2 and watt basis and incorporates other changes they didn't have time to add to Tonga? Do you honestly believe 390X won't have superior perf/watt to a 290X? Since 290X uses about 270W, this is a done deal unless you think 390X will only be 11% faster?
The thing with that poster is that he never posted even 1 positive thing about any AMD products since he joined. Most people who followed his posts for the last 2.5 years pretty much know he will post nothing positive regarding anything AMD. In every thread on AMD he posts negativity.
----
I am interested in seeing how AMD will address the active DP transistors for gaming. NV has solved this issue but it seems this has burdened Hawaii and Tahiti significantly. If AMD manages to improve finer voltage granularity and keep those DP units idle in games, even with no changes to the actual architecture they will already achieve a good increase in perf/watt.
Last edited:
wand3r3r
Diamond Member
- May 16, 2008
- 3,180
- 0
- 0
You caught him cherrypicking again. :awe:
And to add to it:
285 loses to 7970 in performance/watt as well.
Way to miss the point. Read his post.
wand3r3r
Diamond Member
- May 16, 2008
- 3,180
- 0
- 0
Agreed. ATF is full of it. Goal post shifting every generation ftw. At least some of us still are interested in tech and the consumer pov.
xthetenth
Golden Member
- Oct 14, 2014
- 1,800
- 529
- 106
It's not even a difficult concept that's flying over heads. When the stats of a card are a literal superset of those of another, a time when the latter outperforms the former is indicative of an architectural advantage. Therefore to show an architectural advance, he posts an instance where the 285 outperforms the 7970 despite being comprehensively outperformed on paper.
Words have meaning and shape the context for the information the pretty little pictures contain unless we're just measuring whose benches are longer. For stats jousting outliers are cherry picking, for architecture discussions, outliers are a way of demonstrating changes, they don't just spring forth from the lumineferous aether but are the effects of some cause.
There's no shame in having a point go over your head, but purposely ducking it is ridiculous.
Words have meaning and shape the context for the information the pretty little pictures contain unless we're just measuring whose benches are longer. For stats jousting outliers are cherry picking, for architecture discussions, outliers are a way of demonstrating changes, they don't just spring forth from the lumineferous aether but are the effects of some cause.
There's no shame in having a point go over your head, but purposely ducking it is ridiculous.
wand3r3r
Diamond Member
- May 16, 2008
- 3,180
- 0
- 0
Well said. If they'd read the post instead of skimming the pictures they'd have realized the point they missed made their posts look ridiculous.
Ok. That might be it. I never understanded the reaction. I thought the point from rs was so obvious for all, but they reacted because they were angry because the other things he wrote? You know this rs have an agenda "argument". But you might be right. I guess the information and data level from rs can get overwhelming at times, but i really like those solid post. Keep em comming !
They're trolling you. Every reasonable person in this thread knows what you said and pointed out.
Nobody missed the point. He's twisting the point. Not the same thing, although it can look similar.
While picking individual games and showcasing Tonga's improvements do highlight the architectural strides AMD has made this is largely pointless if, in the grand scheme of things, there is little reflection in average performance.
Edit:
Picking games where Tonga does well is fine. But if for every game that Tonga does well, there is a game that tonga does poorly then things balance out.
Edit:
Picking games where Tonga does well is fine. But if for every game that Tonga does well, there is a game that tonga does poorly then things balance out.
Last edited:
It matters in the context it was presented in. RS explained it perfectly. You might need to reread what was said.
Just wanted to correct this, but HBM doesn't actually bring any power efficiency improvements compared to GDDR5, quite the contrary actually (at the setups that will be used in GPUs).
I know that most people have probably seen the claims of a 40% reduction in power usage for HBM compared to GDDR5, however that reduction is on a mW/Gbps/pin basis (source), and since HBM features a massive increase in the number of I/O pins, the net result can very easily be an increase in power.
So in the case of Nvidia they will be going from 256 pins, to most likely 4096 pins (4 1GB HBM modules), and from 7 Gbps to 1Gbps. This will result in a 33% increase in power, not decrease (0.58 * 4096pins/256pins * 1Gbps/7Gbps = 1.33).
For AMD it will be a 7% decrease in power, since they already use a very wide 512 bit bus with Hawaii (0.58 * 4096pins/512pins * 1Gbps/5Gbps = 0.93).
A 33% increase for Nvidia and a 7% decrease for AMD will probably result in the current power gap shrinking by about 10W or so (assuming 4GB of GDDR5 at 5-7 Gbps uses around 20-30 watts).
Now the above is just the memory modules, and obviously the PHY will also change significantly with HBM, which should result in significant power reduction for both AMD and Nvidia. However AMD should once again see the bigger gain, since their current 512 bit interface is probably more power hungry than Nvidias 256 bit interface.
All in all the switch to HBM could on it's own shrink the gap in power usage between AMD and Nvidia, by some 15-20 watts.
While picking individual games and showcasing Tonga's improvements do highlight the architectural strides AMD has made this is largely pointless if, in the grand scheme of things, there is little reflection in average performance.
Edit:
Picking games where Tonga does well is fine. But if for every game that Tonga does well, there is a game that tonga does poorly then things balance out.
Oh well, since Tonga is slow in that game, I'm sure that they won't incorporate the 2x tesselation performance upgrade architectural changes into a future 390x. It'll just be Hawaii but bigger, no architectural differences, no HBM, just bigger Hawaii. Maybe they'll go back to GCN 1.0 or hell, why not kick it old school and go all the way back to VLIW5?
/sarcasm
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 353
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter