How much I love this guy's posts. Cold hard facts backed with numbers, benchmarks, links, images. Solid 100%
On paper specs, Titan X has 3.3X more pixel shading power in games over a 925mhz HD7970, but its performance is about 2.13X faster at 1440P. We can't compare NV vs. AMD ROPs like that.
Tonga has 40% higher memory bandwidth efficiency, which means if Fiji has the same tech, its 512GB/sec is equivalent to 717GB/sec on a 290X, which is more than double.
HD7870 launch:
580 is 3% faster at 1080P with AA @ Computerbase
Today,
7870/R9 270X is 30% faster than the 580 at 1080P with AA @ Computerbase
http://www.anandtech.com/show/7465/...-gtx-780770-price-cuts-gtx-780-ti-launch-dateFirst and foremost, both GeForce GTX X80 and GeForce GTX X70 are getting price cuts, effective tomorrow (October 29th). GTX X80 will be reduced by $150 to $499, and meanwhile GTX X70 will be getting smaller $70 trim, bringing the price of that card down to $329.
So according to you since NV had new stuff to sell, it's OK to neglect older products, despite 780/780Ti being only 1.5-2 years old. WOW, just wow, I guess it's now OK for a $650-700 card to become as slow as a $400 R9 290, that today sells for just $220. :whiste:
That's hilarious coming from NV followers who bashed AMD's frame times with HD7000 series but AMD actually worked to fix the problem instead of closing their eyes because R9 200 series was release. I guess it's OK that R9 270X crushes 580 by 30%, that 680 can't even beat an R9 280 today and 960 OC coming close to 780Ti in some titles.
You should tell a 780/780Ti owner that it's OK that their cards are now destroyed by a $220 R9 290 because a GPU's shelf-life is now < 2 year according to you!
I am pretty sure the guy who paid $700 for a 780Ti finds it amusing that his card is 50% slower in The Witcher 3.
Do you own a computer shop or something because only an NV shareholder or a person who sells Maxwell GPUs would start defending Fermi/Kepler performance in modern titles.
Just imagine how many of his relatives and close friends received his biased advice in the last couple of decades of their PC upgrades. I bet any $ that he is one of those guys who would recommend a 960 over spending $50 extra for an R9 290 but then once a next gen x60 card comes out, he'll recommend the gamer to spend yet another $200 in 2 years from now just to end up at $220 R9 290 level of performance of 2 years ago. Sad, just sad.
Looking forward to your criticism of NV next generation once they copy AIO CLC and HBM tech. Wow, NV is so unoriginal, couldn't even invent their own memory technology, had to copy AMD's HBM design ideas with Pascal. / sarcasm. Also, copied tessellation from Radeon 8500 and EyeFinity, etc.
Unfortunately there is a growing group that wants PC gaming to become a monopoly so all games have GameWorks - you know GeForceMasterRace vs. AMDpeasantPCgamers.
Meh, these tech new sites are in the dark like the rest of us, only difference is they lack logic and speculate like NUTs, completely wildly with so many different outcomes.
Anyone with some common sense will have realized immediately, AMD simply cannot re-badge Hawaii, it's not competitive due to the massive power consumption delta. The R290/X SKUs are selling for $200-280. If they are releasing a new series of SKUs and want to sell it for $329/389, it has to be significantly improved to compete* against 970/980.
** My meaning of competitive 390/X
They need to shave 50W off the TDP if performance is the same, putting it at ~200W gaming load. Or boost performance by 10-20% with a small power reduction. If its faster, it can justify the higher power usage. But it cannot be slower and much more power hungry. That's a dealbreaker.
Can't believe nobody has linked Warsam71's very fitting picture on twitter here yet:
About 8K, this pops to my mind.
http://www.engadget.com/2015/04/07/lg-leaks-8k-imac/
If Any technology is rumored to be revolutionary - this has to be done by Apple.
Ha, just play CiV and D3 like me, then you'll be able to use a 7970 on your 1440p monitor forever...
Honestly by now AMD representative credibility is in same bucket as TSMC's process marketing. We all want AMD to succeed with products, but their credibility was bulldozed into pile of crap. Only benchmarks from reliable 3rd parties count now.
fury for apple 8k then later?
fury for apple 8k then later?
I was wrong to use the term "pixel fillrate" to describe the impact of ROPs. 3Dmark's color fill test is not actually a test of pure ROP performance, but is heavily-influenced and limited by effective memory throughput (notice how Tonga excels at the test with its color compression techniques despite a significant disadvantage in both peak bandwidth and ROPs relative to Hawaii and why Hawaii barely moves past Tahiti on the test despite double the ROPs), rasterizers, and even shader configuration as well. All the peak numbers we can see on GPU-Z are theoretically real in an ideal scenario, but in tests will never be reached (including bandwidth) because of other factors of the GPU. However, the specifications are still real and important, it's just difficult to measure why with consumer tools. Even though more/faster ROPs do not linearly mean a higher score in that pixel fill test because of those other limitations (like Hawaii relative to Tahiti and, to a lesser degree due to HBM's bandwidth and greater shader disparity, the upcoming Fiji relative to Tonga), the higher number of ROPs are still useful for pixel sampling.I forgot to address your point. We can never compare AMD vs. NV ROPs on paper, it has never worked that way.
Titan X has 96 ROPs = 43.6
285 has 32 ROPs = 19.9
290 has 64 ROPs = 16.6
My point was not to compare ROPs of NV to AMD but rather Fiji's to Hawaii's by taking into account that Tonga with only 32 ROPs already clobbers Hawaii in pixel fill-rate.
On paper specs, Titan X has 3.3X more pixel shading power in games over a 925mhz HD7970, but its performance is about 2.08X faster at 1440P per TPU's testing. We can't compare NV vs. AMD ROPs on paper.
I was just mentioning bus width with regards to clock speed to simply demonstrate a misunderstanding people have when ignoring the latter to assume the former alone means anything; you seem to be under the impression I'm taking potshots at Fiji or something, I'm not. I'm fully aware of delta color compression, but it's not, for example, relevant to peak bandwidth resulting from memory bus width in tandem with frequency. Peak bandwidth is peak bandwidth and HBM could be implemented with or without lossless color compression that ultimately just increases the processor's efficient usage of that bandwidth.Tonga has 40% higher memory bandwidth efficiency, which means if Fiji has the same tech, its 512GB/sec is equivalent to 717GB/sec on a 290X, which is more than double.
The performance of the 290X isn't really that far off of the 980. If they keep power usage the same and improve performance by 20% it'll be a good value @ $390. In the end gross performance trumps efficiency if the pricing is good. performance
A 5% bump in clocks is going to equal a 20% increase in performance how?!
because not everyone is expecting a direct rebadge, as you are with zero evidence.
Ouch. That's pretty sad. I'd be disappointed if I had shelled out for a 580.
The overwhelming majority of gamers out there are still at 1080P. 8K isn't going to be a concern for at least five years, probably longer.How the heck are you supposed to do 8K on 4 GB? I mean you could, but the framerate won't be pretty.
How the heck are you supposed to do 8K on 4 GB? I mean you could, but the framerate won't be pretty.
The overwhelming majority of gamers out there are still at 1080P. 8K isn't going to be a concern for at least five years, probably longer.
Did you read the thread? People are hyping that Fiji will be able to run games at 8K, all because the AMD pages posted some pictures of games in 8K resolution.
Did you read the thread? People are hyping that Fiji will be able to run games at 8K, all because the AMD pages posted some pictures of games in 8K resolution.
More hype for disappointment, I'm afraid...