And it's a load of bullcrap... You can't just move a chip designed for Samsung 8nm fab over to TSMC 7nm fab. It would need a redesign to work on the 7nm TSMC node!! The Samsung 8nm is technically a tweaked 10nm process with a higher density hence they decided to call it 8nm. But in all honesty the nanometer naming we have for transistors now, has little bearing to the actual transistor gate size anymore. That kind of went out the window when they got smaller than 32nm and then jumped to FinFET designs. https://www.techpowerup.com/266351/...w-density-metric-for-semiconductor-technology https://www.techpowerup.com/272489/...-7-nm-node-using-scanning-electron-microscope Looking at the CPU side of things, the Intel 14 nm chip features transistors with a gate width of 24 nm, while the AMD/TSMC 7 nm one has a gate width of 22 nm. The gate width had been the metric for years up until the 32nm process when that metric really couldn't shrink down significantly.
Hi, I would but I don't run any monitoring software. In the past I found they gave me problems. All I can say is everything feels smooth, no stutters and load times seem faster.
On another note looks like the 20GB RTX 3080 isn't going to happen, same goes for the 16GB RTX 3070 card. https://videocardz.com/newz/nvidia-allegedly-cancels-geforce-rtx-3080-20gb-and-rtx-3070-16gb
We are not privy to nvidia's sourcing strategy. I read earlier that nvidia tried to use threat of Samsung 8nm to get a better price from TSMC on their 7nm fab and failed which resulted in Ampere going to Samsung. Nvidia would not need to move a chip designed for Samsung 8nm fab over to TSMC 7nm fab, if nvidia had dual sourced Ampere for both Samsung and TSMC. https://www.techpowerup.com/266656/tsmc-secures-orders-from-nvidia-for-7nm-and-5nm-chips
There are all kinds of fake news multiplied and circulated. As I said, the GPU has to be redesigned from scratch in order to be transferred to a new fab proccess. You can draw your own conclusions.
There isn't a single game I can think of that actually exceeds 10GB memory usage currently. Most of this nonsense that people think games are actually using anywhere close to 10GB is due to pretty much all the monitoring tools only tracking the Allocated Vram. The Allocated stat is basically just how much memory the game is asking to be reserved for the game so it can't be utilized by Windows or other applications. Most games don't use the entire poll of Allocated Vram. In some cases there can be as much as a 40-50% buffer of available Vram that's not being used, it's just being reserved. However the most recent Afterburner Beta (v4.6.3.15854 Beta 3) which added features to the Gpu.dll plugin to actually track the actual memory usage for the process that is running! Here is RDR2 on Ultra settings at 4K with 1.5x resolution scaling (So actually running at 5760x3240 resolution). Actual memory usage is 6.2GB with 7.3GB allocated, I only rode around for maybe 2-3 mins though. Usually memory usage will increase slightly after playing the game for maybe an hour as you load different areas. Most likely might increase 1GB more usage after playing a while and more textures get cached to Vram. http://shadowdane.com/public/RDR2/RDR2_6K_downscaled_4K.png
@Shadowdane : Breakpoint at 4K Ultimate settings does use more than 10GB VRAM, though I didn't tried it yet as it's not exactly game I really need to play
Have you tested that with the new Afterburner feature as I mentioned above?? There is a difference between Allocated Vram and Actual Vram being used! I don't own Breakpoint, but here's Rise of the Tomb Raider with 3 scenes from the built in bench. The game is basically reserving about 25% more memory than what's actually being used. It's basically like regular Ram if you give applications 32GB to work with they are going to allocate more to keep a bigger cache loaded in your Ram. I'm sure if you tested this game on a 3090 for example it might allocate more than 10GB but the actual used memory would be about the same. https://imgur.com/gallery/7I2E7uK Scene 1 Allocated: 10.77GB Used: 8.37GB Scene 2 Allocated: 10.75GB Used: 8.34GB Scene 3 Allocated: 10.08GB Used: 7.68GB
I managed to get 4k,120,10bit and HDR working on my C9. I used PC resolutions instead. Doom Eternal looks amazing, so smooth and oddly enough everything looks clearer, could be to do with the smoother motion.
Well, the design wouldn't change much, but it would have to be tweaked for the new density, then retaped for the 7nm fab, but it can be done. Would cost a bundle for the retaping, however.
No transfer required. Nvidia dual sourced Ampere to both TSMC 7nm and Samsung 8nm fab proccesses. Nvidia wanted to use Samsung threat to get a better price for TSMC 7nm and failed. From a May 5th 2020 techpowerup article, "TSMC has reportedly secured orders from NVIDIA for chips based on its 7 nm and 5 nm silicon fabrication nodes, sources tell DigiTimes". Cost of TSMC 7nm process has now gone down, which is what NVIDIA was hoping for.
Managed to get a 3080FTW3 Ultra Gaming, out of PURE luck. Now, I am trying to undervolt this beast to lower the power draw and temps, best I got is 0.900V@1995MHz, but it still sucks about 360-380W. What's the sweet spot in your opinion? Besides, does anybody know how to disable the rgb christmas lights? EVGA precision does not allow me, there is no option to customize the leds.
Yes Precision X1 controls all the LEDs. I can make the entire card go dark (no lights) or anywhere in between (static color, rainbow, wave, etc.) In other news, my goodness this card is butter. Flew around my area (Tampa, FL) and it's astonishing with the graphics ramped up. 3440x1440 Ultra settings. I've found out in the middle of the ocean (aka no scenery other than the current hurricane out in the Atlantic I flew in) I hit "GPU limit" instead of "Mainthread" (CPU) limit. It's roughly 92-98 FPS it can crank on FS 2020 without the CPU limiting it. EDIT: This is the GeForce software recording it in realtime and still rendering the game at this frame rate. Stupid. Costs only a few FPS tops over not recording.
I dl'ed the wrong version, didn't see there was a specific release for the 3000 series. I finally turned everything off. Thank you! edit: I am having a small issue with my FTW3: basically on facebook, when a video is being autoplayed in any newsfeed and I refresh the page keeping the video on screen, the nvlddmkm driver crashes and is restored. When this happens I noticed that the core frequency spikes to 1800MHz, the browser's window goes black for one second and afterwards everything works again. This happens only with facebook's videos, if for instance I am watching a 4k video on yt and try to refresh nothing unusual happens. This problem occurs with the card at default settings, with no customized oc and so on. Only thing I did was switching to OC mode from the card itself. Can maybe somebody who owns a ftw3 test this particular scenario and report back? Idk if it's the card causing problems, windows 10, or the way facebook is programmed. The card gives me no troubles while playing, drivers never crashed, even at the most demanding settings and resolutions.
Did that, including a clean install with DDU. As i wrote, it happened even if the card was at its stock settings, no OC, no downvolting, just switched to oc mode from the card itself. I also went back to "normal" mode, tested that particular case, and the video driver crashed again. Maybe could you try that too and tell me if you notice anything weird (since you own the very same model as mine)? As browser I am using the latest MS Edge based on Chromium. Just tested on Firefox, the problem doesn't occur.
