|
|
| Page 365 of 782 |
|
|
 Posted: Mon, 11th Apr 2016 14:58 Post subject: |
|
 |
http://www.sweclockers.com reporting nvidia has told their oems that 980ti gets replaced in may and 970 in june
|
|
| Back to top |
|
|
tonizito
VIP Member
Posts: 52316
Location: Portugal, the shithole of Europe.
|
| Posted: Mon, 11th Apr 2016 15:02 Post subject: |
|
|
Yeah, the ti is going to be replaced with the next $2000 derptan card 
| boundle (thoughts on cracking AITD) wrote: | | i guess thouth if without a legit key the installation was rolling back we are all fucking then |
|
|
| Back to top |
|
|
|
|
| Posted: Mon, 11th Apr 2016 15:07 Post subject: |
|
|
thats nothing new, it was rumored long time ago that the first pascal cards will be expensive highend cards (600-700 bucks) followed bei mainstream cards a few weeks later
|
|
| Back to top |
|
|
|
|
| Posted: Mon, 11th Apr 2016 16:27 Post subject: |
|
|
Day 1 man!
Enthoo Evolv ATX TG // Asus Prime x370 // Ryzen 1700 // Gainward GTX 1080 // 16GB DDR4-3200
|
|
| Back to top |
|
|
_SiN_
Megatron
Posts: 12108
Location: Cybertron
|
| Posted: Mon, 11th Apr 2016 17:45 Post subject: |
|
|
| Janz wrote: | | http://www.sweclockers.com reporting nvidia has told their oems that 980ti gets replaced in may and 970 in june |
I'd love to get Werelds perspective on this - is a May/June release realistic at all? For the 980/970 replacements? Doesn't seem probable..
Watercooled 5950X | AORUS Master X570 | Asus RTX 3090 TUF Gaming OC | 64Gb RAM | 1Tb 970 Evo Plus + 2Tb 660p | etc etc
|
|
| Back to top |
|
|
|
|
| Posted: Mon, 11th Apr 2016 17:58 Post subject: |
|
|
release yes, availibility for endusers? nope
may and june are the dates on which oems get their chips from nvidia for mass production
|
|
| Back to top |
|
|
Stige
Posts: 3579
Location: Finland
|
| Posted: Mon, 11th Apr 2016 18:13 Post subject: |
|
|
| _SiN_ wrote: | | Janz wrote: | | http://www.sweclockers.com reporting nvidia has told their oems that 980ti gets replaced in may and 970 in june |
I'd love to get Werelds perspective on this - is a May/June release realistic at all? For the 980/970 replacements? Doesn't seem probable.. |
And what makes Werelds OPINION any more valid than anyone elses? Is he some sort of god???????+++
|
|
| Back to top |
|
|
vaifan1986
Posts: 4645
Location: Birthplace of the necktie.
|
|
| Back to top |
|
|
Nui
VIP Member
Posts: 5720
Location: in a place with fluffy towels
|
| Posted: Mon, 11th Apr 2016 18:32 Post subject: |
|
|
| Stige wrote: | | And what makes Werelds OPINION any more valid than anyone elses? Is he some sort of god???????+++ |
So when it comes to opinions you value everyones opinions the same? Even if one usually comes across as clueless and the other as a very informed person?
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35846
Location: Sweden
|
| Posted: Tue, 12th Apr 2016 09:01 Post subject: |
|
|
http://www.guru3d.com/news_story/nvidia_gp100_gpu_architecture_recap_full_gpu_has_3840_shader_processors.html
| Quote: |
Last week Nvidia announced the GP100 GPU powering the Tesla P100 HPC module. More rumors are surfacing about GP104 as well. Since the full block diagrams for Gp100 are now available, we can also tell what a full GP100 looks like when fully enabled, in this post a little recap on the GP100 architecture and its positioning.
This year several GPUs are going to be released from Nvidia, all based on their new Pascal architecture in a wide variety of segments in the market channels. For consumers the first wave of graphics cards it will be the GP104 GPU, these are empowering the high-end products like 'GTX 980' class products, the current rumor is that the new GTX 1070 and 1080 albeit with a bit of weird Full HD like naming, will use that chip. These should be announced during Computex time in June with availability in the summer (likely July).
Then there is big Pascal, the big daddy Nvidia GPU developed under GPU codename GP100. This is the GPU that will empower (for the consumer side) the enthusiast class products e.g. the Titan etc. Make no mistake, this product will not launch anytime soon for consumers. Expect at the very best a launch late this year closer to the Christmas season, likely even later in Q1/Q2 2017 (we think).
All Pascal products are based on a 16nm FinFet design and the GP100 in particular comes with stacked HBM2 (16GB in four stacks). The Pascal based GPU driving the unit holds 15 Billion transistors which is roughly double that of the current biggest Maxwell chip. Gp100 is huge at 600mm^2. The prognosis performance (according to Nvidia) is 5.3TFLOPS using 64-bit floating-point numbers and is rated at 10.6TFLOPS using 32-bit and 21.2TFLOPS using 16-bit. P100 has 4MB of L2 cache and 14MB of shared memory for just the register file. The following table provides a high-level comparison of Tesla P100 specifications compared to previous-generation Tesla GPU accelerators, however I added the GP100 as a fully enabled product:
[image]
[graph]
As the block diagram now shows, the GP100 features six graphics processing clusters (GPCs). Just look at the diagram and count along with me - each GPC holds 10 streaming multiprocessors (SMs) and then each SM has 64 CUDA cores and four texture units. Do the math and you'll reach 640 shader processors per GPC and 3840 shader cores with 240 texture units in total.
6 (GPC) x (10x64) = 3840 Shader processor units in total.
Meaning the GP100 used on the Tesla P100 is not fully enabled. Nvidia is known to out GPU that have disabled segments, it helps them selling different SKUs, the Tesla P100 holds a shader count of 3584 and thus has 56 SMs enabled (from the 60).
GP100’s SM incorporates 64 single-precision (FP32) CUDA Cores. In contrast, the Maxwell and Kepler SMs had 128 and 192 FP32 CUDA Cores, respectively. The GP100 SM is partitioned into two processing blocks, each having 32 single-precision CUDA Cores, an instruction buffer, a warp scheduler, and two dispatch units. While a GP100 SM has half the total number of CUDA Cores of a Maxwell SM, it maintains the same register file size and supports similar occupancy of warps and thread blocks.GP100’s SM has the same number of registers as Maxwell GM200 and Kepler GK110 SMs, but the entire GP100 GPU has far more SMs, and thus many more registers overall. This means threads across the GPU have access to more registers, and GP100 supports more threads, warps, and thread blocks in flight compared to prior GPU generations.
Since the graphics memory is on-die HBM2, the VRAM amount is fixed. That means that ALL GP100 products will get 16GB of memory. Nvidia might disable two HBM2 units and sell 8GB parts as well, though thus is highly speculative. HBM2 will run a wide 4096-bit HBM2 (1024 bit per IC stack) memory interface running an effective bandwidth anywhere up-to a full 1 TB/s.
This is a big chip, very big at 600mm^2 hence it is interesting to see that 16nm can offer a lot in terms of clock frequency, The Tesla P100 is an enterprise part that ends up in servers, however this part already is clocked at 1328 MHz with Boost capabilities towards a frequency of 1480 MHz. Combined the TDP still remains to be under 300W.
|
http://www.guru3d.com/news_story/amd_polaris_11_in_shows_compubench_has_1024_shader_processors.html
| Quote: |
This is a pretty interesting find, in the CompuBench database some device entries of what seems to be AMD Polaris 11 (the mainstream SKU) has appeared. It's all a bit dodgy to dig out, but each graphics hard has an identifying device number, a Device_ID. These Device_ID have appeared in the CompuBench databases and as such you can extrapolate products from that.
The most recent find it AMD Polaris 11, Device ID 67FF:C8 codenamed “Goose”. This would be the base GPU for a several entry-level products. Now, the CompuBench database reports back that this device has 16 CUs with a maximum clock frequency of 1000 MHz. Multiple your CUs (compute units) by the number of shader processors per cluster (assuming that AMD keeps 64 per cluster) and you'll notice that Polaris 11 in this configuration has 1024 Shader processors tied to a 128-bit bus and 2048 MB of memory.
Polaris 10, codenamed "Ellesmere," would then feature over 2304 stream processors (36 CUs); and Vega 10 featuring 4096 stream processors, with 64 CUs. Things could end up looking like this:
[graph]
Mind you that the specs shown in the CompuBench database might not be the full unlocked GPUs, so the CU numbers might even be higher. Which makes this news-item very speculative. We are inclined to say that Polaris 10 really would get 2560 shader processors.
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Tue, 12th Apr 2016 10:11 Post subject: |
|
|
| _SiN_ wrote: | | Janz wrote: | | http://www.sweclockers.com reporting nvidia has told their oems that 980ti gets replaced in may and 970 in june |
I'd love to get Werelds perspective on this - is a May/June release realistic at all? For the 980/970 replacements? Doesn't seem probable.. |
I don't see why it isn't possible, Nvidia have been better at keeping their shit under wraps lately. Obviously this is about GP104 or the equivalent thereof (~300-400mm chip) and that's also the only chip feasible to launch in the short term. I highly doubt GP100 will happen for consumers at all and even an FP32 focused version of it I don't see happening until GP100 itself is widely available, so that'll be Q2'17.
That said, I still doubt it. I reckon it'll be a paper launch at best, purely for marketing reason. If it is, then we might see it in Q3 with GDDR5X; they won't go for HBM2 on that one. If it isn't and it's a soft launch with it becoming available within a month of said launch, I wouldn't expect GDDR5X unless Micron are ahead of their schedule. It's that or they "launch" it, with Fermi-like levels of availability for the first few months. I don't see them doing a proper hard launch in June.
Again, GDDR5X is not ramping up volume until mid-summer according to Micron themselves. When Micron ramp up, it still takes at least a month, if not 2, for it to appear in any product that's properly launched.
As I've said before, Nvidia weren't ready in January. Back then there were "rumours" of Pascal "launching" by March, I said then that that wasn't going to happen because Nvidia clearly weren't ready. When Nvidia have something ready, they will show it off. That's just what they do. At the same time, they're also not afraid of faking it, using woodscrews if necessary  . Anyway, that "March launch" became April...then May...now June. Realistically, Jan - June is enough time for them to move to production volumes if they got working silicon back then. And I'm not sure things are going that well for them.
My money is still on Aug-Sep for any real launch. Announcing a chip or architecture is not launching it; GP100 won't be launched until 2017, even though it has been announced already.
Even on the DGX-1 boards (the 125k super computer with 8 GP100s) they didn't actually have 8 full packages of GP100. Better yet, there seem to be 4 or 5 different chips on the board they showed. That reeks of them having to do multiple spins of it and thus points to yield issues with GP100. Considering what they announced is a cut chip straight off the bat, with availability some 8-9 months after announcing it, it is clearly not panning out the way they wanted it to. That's not a surprise, because they are attempting to build a bloody 610mm chip on a brand new process, using some very new technology (FinFETs), coupled with brand new memory (HBM2) that requires some brand new silicon (interposer)..see where this is going?
This has worked for them exactly..never. Haters gonna hate, but every time Nvidia have attempted this, it has not worked out. NV30 (FX series) and Fermi are the best examples. Both 6 months late, both underwhelming in performance..and downright bad when you compare them to what ATI/AMD had to offer. R300 (9700 <3) had 25% fewer transistors (HUEHUEHUE) and yet beat NV30 without problems. And that got better when you turned on AA/AF, it demolished NV30 then (that chip is the reason AA and AF became common good; it showed that you can apply them without sacrificing 50% performance). Same goes for Fermi: it didn't get beaten on raw performance, but it did on every single other aspect. And considering Cypress was almost 40% smaller, it makes that meager 10% performance advantage look even worse. Had AMD built Cypress to be around 400mm^2 it probably would've beaten GF100 already.
And before anyone tries to claim it's a "choice" by Nvidia to release GP100 next year: no, it isn't. The last and only time Nvidia announced an HPC product this far ahead of launch without having the architecture in the wild already was Fermi (GF100). Big Kepler (eventually turning out to be GK110; the announcement specified no name) was announced alongside GK104, came out 6 months after. As for Maxwell and the actual Tesla microarchitecture (GT200)? All announced and released after their consumer versions were out. Even Fermi's refresh (GF110) only came to Tesla 6 months after its consumer version. |
|
| Back to top |
|
|
_SiN_
Megatron
Posts: 12108
Location: Cybertron
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Wed, 13th Apr 2016 21:10 Post subject: |
|
|
I think you're underestimating just how much of a difference 16nm makes.
Some boring data first to know what we're looking at.
Chips:
- Nvidia = GK104, GK110, GM204, GM200
- AMD = Tahiti, Tonga, Hawaii, Fiji
Starting with transistor density (in millions per mm^2)
- 12.04, 12.62, 13.07, 13.31
- 12.25, 13.93, 14.16, 14.93
Clockspeeds (maximum boost):
- 1085, 1019, 1216, 1075
- 1000, 970, 1000, 1050
SPs (FP32 ALUs..or "CUDA Core" HUEHUEHUE):
- 1536, 2880, 2048, 2816
- 2048, 2048, 2816, 4096
And last but not least, SP (FP32) GFLOPS (# SPs * Clock * 2 / 1000):
- 3333, 5869, 4981, 6054
- 4096, 3973, 5632, 8602
So here's GP100: 25.08m transistors per mm^2, 1480 MHz, 3584 SPs, 10609 GFLOPS.
Things to note:
- Nvidia's big chips run slower than their mid-size ones; AMD doesn't differentiate much, but runs slower altogether.
- Transistor density has consistently gone up on both sides as 28nm matured; AMD has crammed 22% more transistors per mm in Fiji than in Tahiti (whether that was a good decision or not I won't touch on  )
- Even this early on 16nmFF Nvidia are already getting an 88.4% increase in transistor density over GM200 (98.7% over GK110!) - 100% by next year is possible, especially since TSMC are claiming pretty good defect rates already (much better than 28nm early on).
- GM200's 21.5% higher GFLOPS count translates almost 1:1 into performance compared to GM204
- GK110 with almost the same GFLOPS as GM200 performs substantially worse in comparison, even slower than GM204 at almost twice the size
First of all, boost speeds in excess of 1500 MHz on smaller chips are totally viable and perhaps even probable for Nvidia.
Simple bit of theory: take GM204 and Tonga, double the SPs and assume a 1400 MHz clock: both would end up with 4096 SPs, good for 4096 * 1400 * 2 / 1000 = 11468 GFLOPS. For GM204 this would mean a 130% increase, for Tonga that's almost 190%. No matter how you look at this, double the performance at this die size is absolutely possible. And that would mean an 80% increase over a (stock) 980 Ti.
Now, obviously, there are other changes that I'm not even taking into account. They'll also want more ROPs, TMUs, so some transistors will go to that. But at the same time, it's not unlikely for both to have more efficient SPs. And all this would be at somewhere near half the die size of a 980 Ti.
Don't forget that the numbers we've seen from GP100 waste (yes, for gaming it's a waste) a lot of space on FP64 ALUs. Especially for Nvidia, that's not going to happen in consumer cards. That 3584 SP count for GP100? Totally not representative of what "Big Pascal for Gaming" will have.
Here's a reversed bit of theory; let's say Nvidia and AMD both do a GPU at 350mm^2. Nvidia with the same 25m density as GP100, AMD probably closer to 27m. That means they will produce chips with about 8750m and 9450m transistors respectively - that's more transistors than GM200; this theoretical AMD chip would even beat Fiji in transistor count. Some of it will be lost to TMUs and ROPs as I said, but even at these numbers there's no reason they can't cram more SPs into the same size as before. Couple that with a clockspeed that's 40% higher and 100% performance increase is totally doable.
Note, I'm not saying this will happen, but technically it is 100% viable 
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35846
Location: Sweden
|
| Posted: Wed, 13th Apr 2016 22:09 Post subject: |
|
|
For business reasons though I expect that we'll see a big gain for the upcoming GPU's from both AMD and Nvidia but with a bit of a headroom so there can be another big gain for the refresh models coming later on as they get more experienced with the new die shrink and fine tune things a bit further.
Probably still over 50% or 60% for the top-end models compared to current GPU models but not a full 100% even if that's achievable.
I might be entirely wrong of course and it might be possible to wring even more out of these GPU's to still have a good performance gain for both this first generation 16nm and the second one later on.
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35846
Location: Sweden
|
| Posted: Fri, 15th Apr 2016 09:48 Post subject: |
|
|
http://www.guru3d.com/news_story/amd_radeon_r7_470_and_r9_480_at_computex_or_e3.html
| Quote: |
Both graphics card manufacturers AMD and Nvidia are gearing up for the new high-end graphics card releases. It looks like we're going to have a busy summer. New rumors on the AMD product launch surfaced as well, this round the Radeon R7 470 and R9 480 are topic of discussion.
The new Polaris based GPUs are based on the new and energy efficient 14nm FinFET process.
Let's break it down:
Radeon R7 470 - This SKU series will be based on 14 nm "Baffin" (aka Polaris 11), rumored to be a 50 Watt TDP card.
Radeon R9 480 - This SKU is based on 14 nm "Ellesmere" (aka Polaris 10), rumored to be a ~130 Watt TDP card.
Obviously the the avid PC Gamer the R9 480 will be an intersting SKU as rumors right now point to a GPU that holds 2560 shader processors (GCN iteration 4). The R9 480 would get an active 2304 shader processors (leaving room in the GPU for an 480X model.
That means 40 shader processor clusters (each holding 64 SPs):
36 x 64 = 2304 (Radeon R9 480 / Polaris 10)
40 x 64 = 2304 (Radeon R9 480X / Ploaris 10)
We are looking at a 256-bit wide memory bus, yet it is unclear if that'll hold GDDR5 or doubled bandwith with GDDR5X. Polaris 10 is expected to be clocked in the 1 GHz marker on the GPU core frequency.
It is expected that AMD will announce Polaris GPUs on Computex 2016 (May 31-June 4) or even more likely, E3 2016 (June 14-17), more in line with last years release.
|
|
|
| Back to top |
|
|
|
|
| Posted: Fri, 15th Apr 2016 11:45 Post subject: |
|
|
Wait, so we're not getting any 490 cards from AMD this year? With Nvidia releasing their x70, x80 and possibly x80Ti cards, does that mean AMD is basically handing the high end market to Nvidia for almost a year until Vega 10 releases, by which point the green team might also have something new to offer? Unless 480X is as powerful as the Pascal x80/x80Ti which wouldn't make much sense.
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Stige
Posts: 3579
Location: Finland
|
| Posted: Sat, 16th Apr 2016 14:49 Post subject: |
|
|
4GB is "meh" for anything higher than 1080p. And at 1080p you don't even need Fury so. The card is kinda meh all around really, 390X is better value. Seen so many people sell their Fury to buy a 390X, especially at higher res.
|
|
| Back to top |
|
|
|
|
| Posted: Sat, 16th Apr 2016 14:59 Post subject: |
|
|
I have no intention of playing at higher res, the 390x is not as powerful as the Fury and they're roughly the same price, and the claim "don't need at 1080p" is preposterous. There are plenty of games I can't play at 1080p60 right now because of either grossly inflated system requirements (shitass developers) or lousy optimisation (shitass developers)
That "don't need for 1080p" nonsense needs to die already.
|
|
| Back to top |
|
|
Stige
Posts: 3579
Location: Finland
|
| Posted: Sat, 16th Apr 2016 15:03 Post subject: |
|
|
Well it would definitely be an upgrade over that lowsy Nvidia card you have
 Spoiler: | |
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Stige
Posts: 3579
Location: Finland
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Neon
VIP Member
Posts: 18941
Location: Poland
|
| Posted: Sat, 16th Apr 2016 19:14 Post subject: |
|
|
Why is the memory clock on the R9-Fury so low?
|
|
| Back to top |
|
|
Stige
Posts: 3579
Location: Finland
|
| Posted: Sat, 16th Apr 2016 19:18 Post subject: |
|
|
|
|
|
| Back to top |
|
|
JBeckman
VIP Member
Posts: 35846
Location: Sweden
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Apr 2016 05:30 Post subject: |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Tue, 19th Apr 2016 06:31 Post subject: |
|
|
Maxwell nerf drivers coming in 3..2..1...
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
| Page 365 of 782 |
All times are GMT + 1 Hour |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB 2.0.8 © 2001, 2002 phpBB Group
|
|
 |
|
...
very few games that actually run good in relation to how good they look, meh. Quantum Garbage being a prime example of this awfulness really.
