Claudiovict
Junior Member
- Jul 21, 2025
- 5
- 1
- 36
They aiming for RTX 6090 or 5090?I'm pretty sure the big chungus gfx13 is a thing.
They aiming for RTX 6090 or 5090?I'm pretty sure the big chungus gfx13 is a thing.
They're aiming for a specific # of cloud gaming instances per board per watt.They aiming for RTX 6090 or 5090?
it's a 60W part wtaf did anyone expectLooks like almost linear gains in the 25w to 45w tdp range
Revit 2026 introduces a new graphicstechnology designed to significantlyimprove performance in both 3D and2D views. The ‘Accelerated Graphics TechPreview’, which is still in development,delivers smoother, more responsive modelnavigation when panning, orbiting, orzooming. Those working with particularlycomplex Revit models should experiencethe most noticeable performance benefit.
To get the best performance Autodeskrecommends a workstation with at least64 GB of RAM and a GPU with at least8 GB. In the past, this would have meant aworkstation with a discrete GPU. However,with the new AMD Ryzen AI Max PROprocessor — featured in the HP ZBook UltraG1a laptop and HP Z2 Mini G1a desktopworkstations — users now have access toan integrated AMD Radeon GPU with theequivalent performance of a mid-rangediscrete GPU. Coupled with up to 128 GB ofsystem memory, of which up to 96 GB can beassigned to the GPU, the chip is fully capableof handling Revit’s new graphics demands.When running Revit’s Snowdon Towersample project on the HP ZBook Ultra G1awith an AMD Ryzen AI Max+ PRO 395processor, the viewport feels significantlymore responsive compared to the traditionalgraphics engine. Model navigation is silkysmooth, and the model is not simplifiedin any way — a technique often used tomaintain full interactivity with largermodels. As shown in our benchmarks below,performance (measured in frames per second)was approximately 2.5 to 3 times faster.
According to Autodesk, AutoCAD2026 delivers major performanceimprovements over the 2025 release,including file open times up to 11x faster andapplication launch speeds up to 4x faster.One of the most impactful — yet lesstalked about — enhancements lies inthe graphics pipeline. Users can expectnoticeably smoother navigation of 3Dmodels and a significantly more responsivedesign experience when using wireframe,conceptual,and hiddenvisualstyles. Theseimprovementsare enabledby the latestenhancements toGraphics System Fabric(GSF), a next-gen graphics enginebuilt on Microsoft® DirectX® 12that optimises GPU utilisation. GSFoffloads more graphics processing from theCPU to the GPU and caches more data inGPU memory, allowing for faster access andbetter real-time performance.GSF represents a significant leapforward for AutoCAD, but it is still underdevelopment. Certain visual styles —such as realistic — are not yet supported,so AutoCAD will automatically switchbetween GSF and the previous GraphicsSystem One (GS1) engine, based on theactive view or model requirements.Another key advancement in AutoCAD2026 is improved support for processorswith integrated graphics. Historically,integrated GPUs often lacked sufficientmemory to be supported by GSF, whichtypically requires between 4 GB and 8 GBof dedicated GPU memory.That limitation is effectively removedwith the AMD Ryzen AI Max PROprocessor, which can allocate up to 96 GBof system memory to its integrated RadeonGPU — far exceeding the capacity of mostdiscrete GPUs in its class.In real-world testing on an HP ZBookUltra G1a powered bythe AMD RyzenAI Max+ PRO395 processor, thebenefits of GSF areclear. Navigatinga complexMastenbroek heavymachinery model (pictured)in hidden or conceptualview style is instant and fluidin AutoCAD 2026 with no loss ofdetail. In contrast, there is a noticeablelag in AutoCAD 2025: mouse movementslead to delayed on-screen responses, asthe software must first lower model detailin order to maintain interactivity, beforerestoring full detail once movement stops
tbh that's actually less than I expected.it's a 60W part wtaf did anyone expect
Strix is like 28W and double that is 60 so, yeah, that's platform-intended.tbh that's actually less than I expected.
Damn if so the power doesn't scale well as number of CUs increases in RDNA.Strix is like 28W and double that is 60 so, yeah, that's platform-intended.
Sure as hell they would, at like ~2.2G.but 160 CU in RDNA3 or 4 sure aint gonna be 280W.
HBM being production capacity/cost priced out of the consumer market aside, would that solve the power issue?Mind the considerably more power-pricey GDDR shoreline doe.
I mean it's not *that* power hungry.HBM being production capacity/cost priced out of the consumer market aside, would that solve the power issue?