Again: monolithic Zen 2 CPUs would not be as efficient as Chiplet based designs. They would not clock as high, or would consume more power.
Again: monolithic Zen 2 CPUs would not be as efficient as Chiplet based designs. They would not clock as high, or would consume more power.Thats absurd. There is nothing decreasing with die size except yield.
Again: monolithic Zen 2 CPUs would not be as efficient as Chiplet based designs. They would not clock as high, or would consume more power.
Im currently on a gtx770 and was watching navi. It's out of my desired pricing so what is the best choice for 1080p on a 200-300$ budget now? 1660 or 1660ti? rx 590?
Anyone that thinks GPU pricing will ever return to pre mining levels....no..not happening..
Mostly I agree, apart from one thing. There will be better options, especially, when Nvidia will lower MSRPs of RTX GPUs and this, GTX 1660 Ti, will also be hit with lower MSRP(potentially...).The GTX 1660 ti for this generation is completely unappreciated in the general landscape.
For the most part, it resembles premining prices. That is it present a 36% jump in performance vs the GTX 1060 for 30 more dollars or 12% more money which basically pays for the price of the memory. Far better than anything released today and in the near future for a new release card.
Navi coming in at 350, 279 would have absorbed a tonne of good will and marketshare for AMD like the 4870/ 5870(40-48% marketshare) opposed to the 18-20% marketshare.
Yes this always existed, but as the structure sizes get smaller, more quantum effects dominate and the variation balloons. We might be at the point where it's almost certain we'll get slower sections on a larger die.Thats absurd. There is nothing decreasing with die size except yield.
The process is characterized for best and worst corners. There is nothing particularly unsual for TSMC N7. Only if your plan is to heavily rely on binning your chances of getting an exceptional good die increases with smaller die sizes. But again this is applicable to any other process.
AMD appears to be not caring about marketshare, but caring about margins.
So why release that much larger, earlier process, 16 GB HBM2 card for $700?More about preserving margins because of the crazy high 7 nm wafer costs.
Im currently on a gtx770 and was watching navi. It's out of my desired pricing so what is the best choice for 1080p on a 200-300$ budget now? 1660 or 1660ti? rx 590?
So why release that much larger, earlier process, 16 GB HBM2 card for $700?
More about preserving margins because of the crazy high 7 nm wafer costs.
The GTX 1660 ti for this generation is completely unappreciated in the general landscape.
For the most part, it resembles premining prices. That is it present a 36% jump in performance vs the GTX 1060 for 30 more dollars or 12% more money which basically pays for the price of the memory. Far better than anything released today and in the near future for a new release card.
What makes it a good value even in today's market is that it competes with the Vega 56 at fire sale loss prices in price/performance and it moved the market forward and caused a cascade of price drops.
The reason it is unappreciated is because of a hype train that navi was going to come at $250 dollars with largely the performance stated which was unrealistic to begin with(7nm + Gddr6).
On top of this, there was a unrealistic attitude that the GTX 1660 ti should have been an gtx 2050 /1650. (140% faster) without a real change in node, to continue the hype train momentum.
Basically the market was led to believe a messiah was coming and the current products on the market were the devil including the well priced ones. Now that Navi is here and we have even less excuses for the price increase, AMD is somehow skirting the blame because of the poor excuse that AMD doesn't get more marketshare even when they have better value( Radeon 4870/5870 says hello). What I have said before and what youtubers are finally realizing after being made fools from AMD's so called leaks is AMD is after your money like every one else and they are trying to turn into the premium brand based on a mountain of hate. Some of it deserved but considering AMD's pricing as of late(not pricing related to firesale, but actual launch and official prices), they are not that much better when it comes to hurting consumers. They are slowly tricking people into buying more for less while getting people to wait months to more than a year for their product. While Nvidia is guilty of it, it is mostly the performance doing the talking(the GTX 1060 outperfomed the rx480 at launch) It's weird to see something like people selling their GTX 1080 ti's to get Vega 64's at 600 dollars. They are trying to turn into the premium brand when they are not.
People are trying to use the R and D excuse for AMD for the price increase(which applies much more to nvidia) but AMD is a very lean company when it comes to R and D because of their long product lifecycle and extensive use of engineers in China and India. That means they can afford to charge less because they amortize IP at a much lower rate because the R and D expense is lower and it's done over a much longer period of time due to rebrands. E.g 500 million over 3 years vs 2 billion over 18month-24 months. AMD does not need Nvidia like margins to be profitable because they are a lean company that can be profitable with lower margins. Nvidia spent 674 million on the last quarter on R and D while AMD spent 374 million which is mostly spent on CPU development. So AMD should be the value company when their expenses reflect this(outside of terrible decisions like HBM in gaming parts).
The reason why the optics are so different AMD guerilla marketing team is amazing. They have a strong presence with strategic leaks and also online posters. Do you notice posters disappearing with a strong AMD bias right after the true details of the card come out? I do and I suspect others if they inspected this thread, the radeon VII one or Vega one, would also see it too. Even the free marketers, do a good job of harassing youtubers if they say anything not so positive about AMD. Steve from techspot has mentioned it, Adoredtv talked about it, Not an apple Fan mentioned it too. Radeon Rebellion and the team red compaign are some of the best marketing I have seen because it has turned peoples logic backwards. E.g RX 590 with same die(refresh), same architecture, same memory is okay for 280 dollars, but a GTX 1660 ti, with a larger die, new architecture, new memory is not okay at 280 dollars. It boggles me.
People need to reward cards like the GTX 1660 ti, while punish card like the GTX 2080/2080 ti/2070/ 5700xt/5700/rx 590 that promote a price fixing environment. Card's that raise graphic card cost because they not only pass on the increase in cost due to BOM(GTX 1660 ti), but a little or lot extra because it is beneficial to both producers in increasing their marginthe list above).
Navi more than any card before it needs to fail at 450 to 390 pricing. When the value producers can get away with nearly doubling the cost of what used to be 240 dollars, it worsens or encourages higher pricing from the premium player. Expecting the premium dominant player to value price their products is unreasonable and bad for competition in long run because it bankrupts the competition (it pushes the competition to sell at a loss since they have to see for less than the dominant player) while the increase in volume does not offset the lower margin(they posses most of the marketshare so less room for improvement). I stress again, both companies on the products I listed above need to fail. Not just one. Navi succeeding at the prices above would only help Nvidia raise their own prices.
IF AMD wants to become the premium brand, they will have to do it with products that make the competition look bad like with their CPU's(potentially with ryzen 2). Navi made RTX 2070/2060 look better by comfirming what Nvidia said(Nvidia's 12nm products compete very well with AMD's 7nm products).
Navi coming in at 350, 279 would have absorbed a tonne of good will and marketshare for AMD like the 4870/ 5870(40-48% marketshare) opposed to the 18-20% marketshare.
That GPU has good enough margins in HPC and server space .So why release that much larger, earlier process, 16 GB HBM2 card for $700?
Interesting, I must have been reading about another MRAM type when I picked up the thing about its cell size, even the more recent SOT-MRAM (160 F2) is close to SRAM (160-200 F2) in that regard.Last time i checked MRAMs have larger periphery (word line decoder, sense amplifiers, write circuit) such that area overhead for small slices is quite large. Chances are that if you use MRAM as L1$, area would even increase.
Yes this always existed, but as the structure sizes get smaller, more quantum effects dominate and the variation balloons. We might be at the point where it's almost certain we'll get slower sections on a larger die.
I can speculate on a plausible (to me) reason as to how this might be.
Lets put it to rest for now, and wait for navi 14 to be released. We should get back to this discussion, then.It might be that the variance is slightly higher but just looking at the cell library and process characterization there is nothing un-usual i am aware of.
Interesting, I must have been reading about another MRAM type when I picked up the thing about its cell size, even the more recent SOT-MRAM (160 F2) is close to SRAM (160-200 F2) in that regard.
There is however another MRAM type called VC-MRAM which is much smaller (50-60 F2), albeit with a higher voltage requirement (2.2V to SRAM's 0.6-1.2V).
A recent advance made in MRAM called 'field free' may mitigate that voltagfe requirement though.
Relatively speaking, none of the current MRAM variants have had even a small fraction of the research investment invested in SRAM, and I would expect many improvements to the field as they seek to end the scaling problems that SRAM brings with it - especially in the area of mobile which can benefit most from persistent memory techniques.
Yes, the most significant benefit of more recent memory types/concepts over SRAM, DRAM and Flash seems to be scaling down beyond 10nm.In any case, (from my point of view) the single biggest issue of SRAM for mobile is leakage - as it is increasing with each smaller process node. Here MRAM showing factor 4 and more lower leakage!
This makes me wonder what their tessellation performance is like now that they finally have it enabled, its typically been worse than nVidia's in heavy loads I think?P.S. Does anyone of you know wheter Navi has 4 Prim Shaders Out/8 In, per Shader Engine, or Total?
Pretty much it has been worse in anything geometry related, vs Nvidia.This makes me wonder what their tessellation performance is like now that they finally have it enabled, its typically been worse than nVidia's in heavy loads I think?
I can easily see Intel pulling miracle and making GTX 1660 Ti performance class GPU that fits in 75-90W power draw scenario. What I cannot see is Intel pricing it below 200$ mark, if everybody else has similarly performing GPUs, at the same price, but consuming more power.
Yeah, you still fail to comprehend what I'm saying.
Hint: It had nothing to do with defending a product's price, and it had nothing to do with the actual volume of future sales.
You just don't understand what I'm trying to say... after all these posts...
Im currently on a gtx770 and was watching navi. It's out of my desired pricing so what is the best choice for 1080p on a 200-300$ budget now? 1660 or 1660ti? rx 590?
Anyone that thinks GPU pricing will ever return to pre mining levels....no..not happening..
Considering there will be RTX 2060 Super, they won't hurt their margins that much. TU106 that goes into RTX 2070, and RTX 2060 is essentially the same. They still be selling non-Super GPUs at standard MSRP prices.
The fact that only LLVM patches are coming out and not AMDGPU makes it more interesting.
The GTX 1660 ti for this generation is completely unappreciated in the general landscape.
Navi coming in at 350, 279 would have absorbed a tonne of good will and marketshare for AMD like the 4870/ 5870(40-48% marketshare) opposed to the 18-20% marketshare.
They shifted whole stack up one tier and doubled prices.
So rx480/470 replacement cost as vega and selling as high-end but we see only 10-15% performance uplift vs vega.
rx460/560 replacement will cost about same as 570/580 with again only small performance gain vs them.
The 1660ti is not a good value card. If it has a good value, then so does the 2060. They've got the same price/performance, however the 2060 also has RTX (if that's worth anything to you), so I'd argue that the 2060 is the card with the better value.
Using European prices, computerbase.de says that the 2060 has marginally better perf/€, both in 1080p and 1440p. According to TPU, the 1660ti has marginally better perf/$, 2-3% depending on resolution. TPU actually puts the 1060 at 4-7% better perf/$ than the 1660ti (using 1660ti launch day prices)!