Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Sep 1, 2020.
It will beat it when DLSS is involved even without DX-R.
I didn't hear that mentioned. You reckon 8k upwards is DLSS?
And to add to other comments - went quite Star Trek towards the end lol. He missed out the future bit where humanity is nothing like in Star Trek though.
Since i have a grand total of two RTX title i'll pass. I might get it depending on Mechwarrior 5 which murders RTX cards they will need to be 4x faster to make it playable with shadows.
Yep! That's really good.
Dude... I kinda feel bad for the guy now. Waiting for reviews still. Maybe they're just comparing raytracing performance in all those charts, that's unclear.
New page up:
If the 3070 get's the same cooler as the 3080 i wonder how quiet it will be.
That's always my main concern with GPUs, the noise levels. It's why i ran Arctic Coolers on my GPUs for ages, cuz nothing could beat them in noise levels.
That PCB is small, hopefully a waterblock wont be as expensive as they have come to be generation by generation
3090 GEFORCE RTX
3080 GEFORCE RTX
GPU Engine Specs: NVIDIA CUDA® Cores 10496 8704 5888
Boost Clock (GHz) 1.70 1.71 1.73
Memory Specs: Standard Memory Config 24 GB GDDR6X 10 GB GDDR6X 8 GB GDDR6
Memory Interface Width 384-bit 320-bit 256-bit
Technology Support: Ray Tracing Cores 2nd Generation 2nd Generation 2nd Generation
Tensor Cores 3rd Generation 3rd Generation 3rd Generation
NVIDIA Architecture Ampere Ampere Ampere
Microsoft DirectX® 12 Ultimate Yes Yes Yes
NVIDIA DLSS Yes Yes Yes
PCI Express Gen 4 Yes Yes Yes
HDMI 2.1 Yes Yes Yes
DisplayPort 1.4a Yes Yes Yes
NVIDIA Encoder 7th Generation 7th Generation 7th Generation
NVIDIA Decoder 5th Generation 5th Generation 5th Generation
VR Ready Yes Yes Yes
Display Support: Maximum Digital Resolution (1) 7680x4320 7680x4320 7680x4320
Standard Display Connectors HDMI 2.1, 3x DisplayPort 1.4a HDMI 2.1, 3x DisplayPort 1.4a HDMI 2.1, 3x DisplayPort 1.4a
Multi Monitor 4 4 4
HDCP 2.3 2.3 2.3
Founders Edition Card Dimensions: Length 12.3" (313 mm) 11.2" (285 mm) 9.5" (242 mm)
Width 5.4" (138 mm) 4.4" (112 mm) 4.4" (112 mm)
Height 3-Slot 2-Slot 2-Slot
Founders Edition Thermal Power Specs: Maximum GPU Temperature (in C) 93 93 93
Graphics Card Power (W) 350 320 220
Recommended System Power (W) (2) 750 750 650
Supplementary Power Connectors 2x PCIe 8-pin 2x PCIe 8-pin 1x PCIe 8-pin
3070 is a bit faster than the 2080ti.
Assuming same clock speeds:
3080 is approx 48% faster than 3070.
3090 is 75% faster than 3070.
3080 is reportedly twice as fast as the 1080ti.
I'm getting one.
I was going to get a 3090 but at over twice the price of the 3080 for 20% extra performance, thats daft.
Are we saying the 3080 has over 8 thousand cuda cores?
VIDEO CARD COMPATIBILITY TIPS
When looking to purchase any video card, keep these compatibility points in mind:
Motherboard compatibility - Every modern GPU fits into a PCIExpress 16x slot (circled in red here). PCIExpress is forward and backward compatible, meaning a PCIe1.0 graphics card from 15 years ago will still work in your PCIe4.0 PC today, and your RTX 2060 (PCIe 3.0) is compatible with your old PCIe2.0 motherboard. Generational changes increase total bandwidth (16x PCIe1.0 provides 4GBps throughput, 16x PCIe4.0 provides 32GBps throughput) however most modern GPUs aren’t bandwidth constrained and won’t see large improvements or losses moving between 16x PCIe3.0 and 16x PCIe4.0. If you have a single 16x PCIe3.0 or PCIe4.0 slot, your board is slot compatible with any available modern GPU.
Size compatibility - To ensure your video card will fit in your case, it is good practice to compare the card’s length, width (usually # of slots) and height with your case's compatibility notes. Maximum GPU length is often listed in your case manual o. Remember to take into account front mounted fans and radiators which often reduce length clearance by 25mm to over 80mm. GPU height clearance is not usually explicitly listed, but can usually be compared to CPU tower height clearance. In especially slim cases, some tall GPUs may interfere with the side panel window. GPU width (or number of slots) compatibility is easy to visually assess. mITX cases typically support a max of 2 slots, mATX typically 4 slots, ATX focused cases typically 7 slots or more. Be mindful that especially wide GPUs may interfere with your ability to install other add in cards like WiFi or storage controllers.
Power compatibility - GPU TDP, while actually referring to thermals, often serves as a good estimation of maximum power draw in regular use cases at stock settings. GPUs may draw their TDP + 20% (or more!) under heavy load depending on overclock, boosting characteristics, partner model limitations, or CPU limitations. Total system power is primarily your CPU+GPU power consumption. Situations where both the CPU and GPU are under max load are rare in gaming and most consumer workloads but may arise in simulation or heavy render workloads. See GamersNexus' system power draw comparison for popular CPU+GPU combinations between production heavy workloads here and gaming here. It is always good practice to plan for maximum power draw workloads or power draw spikes. Follow your GPU manufacturer's recommendations, take into account PCPartPicker's estimated power draw and always ask for recommendations here or in the Buildapc Discord. When necessary, it is strongly recommended you use two SEPARATE 8-pin power connectors instead of a daisy-chain connector.
Edit: Apparently so
Specs leaks were so wrong, that's kinda satisfying. 12pins isn't for consumer market at all it's 2x8pins for 3090&3080.
2080-Ti: 4352 Cuda Cores
3070: 5888 Cuda Cores
3080: 8704 Cuda Cores
3090: 10496 Cuda Cores
edit: what's weird is that we clearly see the 12pins on the PCB but the datasheet from the website indicates 2x8pins for founder edition models.
Waw, the hell am i gonna do with 8704 cores?
I'm being serious now lol
I don't care about the price of the 3090. I'm just happy nVidia woke up with the price of the 70 card. RTX 2070 launched at 600$ which was a ridiculous price for what it was. It's had pretty weak shaders performance for its price and the ray tracing performance was also weak. The standard 2070 were supposed to be 500$ but (in Canada anyway) the only cards at the price were blowers models all others models were around 800$ CAD which is 600$ USD. It looks like this time 500$ USD will be the price of all cards including founders edition anyway this is what i got from the event. So it will be around 660$ CAD which is fine anyway better than last time. Still a 50$ USD more than i would like but certainly not as bad at the 2070.
In 1st part with introduction, they as always shown improvement coming from "newest and greatest" that affects performance. So they introduced boost from RT cores and Tensor cores.
By now, nVidia's standard performance comparison is with DLSS when available. And when Tensor performance goes up a lot, it is clear that smartest comparison is with DLSS ON against last generation.
Secondly, that lovely slide showing 2.7x/1.7x/2.7x performance improvement from "New"-shaders/RT/Tensors are not really about better units.
It was comparison of 3080 vs 2080. That's good improvement. But in case of FP32, nothing did change. 3080 has 30TFLOPs which is 2 * SM-count * clock as always. 2 * 5888 * 1.73 = 30TFLOPs.
It is that much better GPU, but way it was presented it was misleading about tech being better. Sure, I expect Tensors to perform better and RT cores as well, but those multiplier are comparison between 2 cards. And not between 1 work unit in Turing vs 1 work unit in Ampere.
No argument about 1st great performance per $ improvement in long time. Cards have good prices and I am happy about it.
3090 is so expensive. The sweet spot is 3080 or 3070. Or of course rdna2 based Navi. For 2020, unless someone can wait for 2021 for Intel or Nvidia's super cards. Interesting gpu power!
Anyone selling GTX 1080Ti?
Looking for one...