Tuesday, March 14th 2023
NVIDIA Plans Different Review Post Dates for MSRP and non-MSRP GeForce RTX 4070
NVIDIA is planning to launch its performance-segment GeForce RTX 4070 "Ada" graphics in about a month from now, with the official launch date set for April 13, and sales commencing on that day. Reviews for new graphics cards typically go up a day or two ahead of product availability. With the RTX 4070, however, NVIDIA is planning to try something new. The company is setting April 12 as the review NDA for the cards that are selling at the MSRP set by NVIDIA; and April 13 for graphics cards priced at a premium (think overclocked designs). This probably incentivizes NVIDIA's board partners to have at least one custom-design card that they're selling at the NVIDIA MSRP; so that their brand can get exposure sooner.
At this point we don't know what NVIDIA's MSRP for the RTX 4070 is. The RTX 4070 Ti launched at an MSRP of $799. The RTX 4070 is rumored to be based on the same "AD104" silicon as the RTX 4070 Ti, but heavily cut-down. The card is expected to feature 5,888 CUDA cores, out of the 7,680 physically present on the silicon, which the RTX 4070 Ti maxes out. The memory sub-system is surprisingly the same, with 12 GB of 21 Gbps GDDR6X memory across a 192-bit wide memory interface, working out to 504 GB/s of memory bandwidth. Most premium ("non-MSRP") boards are expected to resemble RTX 4070 Ti cards, as their board designs are the same; whereas the ones meant to sell at MSRP could see some cost-cutting to suit the lower typical graphics power (TGP) of the RTX 4070.
Source:
VideoCardz
At this point we don't know what NVIDIA's MSRP for the RTX 4070 is. The RTX 4070 Ti launched at an MSRP of $799. The RTX 4070 is rumored to be based on the same "AD104" silicon as the RTX 4070 Ti, but heavily cut-down. The card is expected to feature 5,888 CUDA cores, out of the 7,680 physically present on the silicon, which the RTX 4070 Ti maxes out. The memory sub-system is surprisingly the same, with 12 GB of 21 Gbps GDDR6X memory across a 192-bit wide memory interface, working out to 504 GB/s of memory bandwidth. Most premium ("non-MSRP") boards are expected to resemble RTX 4070 Ti cards, as their board designs are the same; whereas the ones meant to sell at MSRP could see some cost-cutting to suit the lower typical graphics power (TGP) of the RTX 4070.
18 Comments on NVIDIA Plans Different Review Post Dates for MSRP and non-MSRP GeForce RTX 4070
To be fair none of the new GPUs are what I would call bad products it's just the pricing that makes them terrible to me.... At the same time people seem to have purchased a decent amount of 6500XTs and that was an actual terrible product regardless of price.
I'm sure this will be the christmas it's finally worth upgrading... We're almost there...
Thankfully the big channels seem to have caught on and are evaluating subsequent competing products based on the actual prices of cards at retailers, and not Nvidia's "totally real MSRP you guys we swear". That doesn't really solve the day 1 issue though.
wccftech.com/nvidia-geforce-rtx-4070-allegedly-starts-at-749-us-price-custom-models-with-prices-close-to-4070-ti-msrp/
Couple that with this:
Ada GPUs are practically a scam at this point in time, and I'm willing to lump even the 4090 on that, given how neutered it is against AD102's true potential. Shame RDNA 3 failed to meet its power and performance targets, it utterly missed the mark.
And while I'm not betting against a fully enabled AD102 it will only have 12.5% more cuda cores and likely faster memory for an extra 400 usd. The 4080 only performs about 30% worse than the 4090 with 60% ish less cuda cores so I'm not expecting much from a fully unlocked chip
The 4080 wouldn't be so bad if it costed what it should $999 ish usd becuase lets be real even at 1200 I would still buy one over the 1000 usd 7900XTX. The card I really dislike in the ada lineup is the 4070ti it should have been the 4070 at most and priced $599 or lower.
The 7900XTX is ok not good not terrible but I'm not a fan of the 7800XT I mean 7900XT....
3080 /10 or 3090 is just misleading and beating around the bush.
So 4070 ti being 4-9% faster means 4070 is slower, still cheaper or equal value to the 1 year old 3080/12, that nobody could buy at msrp, because it didn't have one, $1700, but later reduced to 800.
My biggest issue with the 4070ti is they gimped it in a way where I doubt it will age well. Other than their flagship cards they've been really stingy on Vram and now it seems they are really skimping on the bus sizes. As much as I dislike the 7900XT I'm guessing it will age substantially better.
It performs so great comparably because most of the Ampere GPUs were heavily power limited, the other Ada cuts are ridiculously awful (compared to full AD102, the full AD104 is around 30-35% relative performance - and has exactly half the resources of the RTX 4090 cutdown as it is), add some intentionally gated features, heavy marketing, sunk-cost fallacy, "pride in the purchase" factor and AMD royally screwing up RDNA 3, suddenly the 4090 "is a good deal".
NVIDIA has basically zero regard for the enthusiast crowd and has been in relentless pursuit of increasing ASP, that is all they care about and they happily took advantage of their market leader position to forsake Ada's generational leap capabilities to sell the worst cuts they could in order to maximize profit, to the point they have entirely done away with performance per dollar improvements vs. the already bad 30 series in that regard.
From the 3090, the only GPU that would make the slightest sense (but not enough to put down $1800+ on one) to me would be the 4090. The 4080 is too expensive for what it offers, and the 4070 Ti is basically trading half my VRAM for the ability to use DLSS 3. And that is exactly what NVIDIA intended, the days of upgrading GPUs every generation are over, you'll be buying one every 6 to 8 years at this rate, requiring at least 5 to break even on a performance per dollar improvement.