By IG Share Share Intel Arc B390 and AMD Radeon 8060S bring integrated graphics much closer to the performance and feature sets associated with dedicated GPUs. The Arc B390 uses Intel’s Xe3 architecture with XMX AI hardware, XeSS 3 upscaling and multi-frame generation, while the Radeon 8060S combines 40 RDNA 3.5 compute units with a 256-bit shared LPDDR5X memory interface. Note: If you buy something from our links, we might earn a commission. See our disclosure statement. This comparison puts the Intel Arc B390 and Radeon 8060S against the AMD Radeon RX 9070 XT and NVIDIA GeForce RTX 5090 to show how current iGPUs compare with desktop graphics cards. It covers GPU architecture, compute hardware, memory bandwidth, power, 1080p and 1440p gaming performance, 4K capability, ray tracing, path tracing, XeSS, FSR, DLSS and frame generation. You will also find lists of Core Ultra Series 3 and Ryzen AI Max processors that include these iGPUs, along with laptops, mini PCs and compact workstations using Arc B390 or Radeon 8060S. This makes it easier to compare the GPU itself and the systems you can actually buy or track in 2026. iGPUs vs Discrete GPUs Comparison | Faceofit.com F Faceofit.com PC hardware, mobile tech and buying comparisons GPU comparison guide iGPUs vs Discrete GPUs Comparison Intel Arc B390 and AMD Radeon 8060S sit at the top end of integrated graphics. This page compares them with Radeon RX 9070 XT and GeForce RTX 5090 across processor options, memory bandwidth, gaming, ray tracing and AI rendering features. Updated January 2026 No pricing used Mobile responsive Interactive filters and charts Summary iGPU processor list Hardware Memory Gaming Ray tracing AI features Scorecard Charts Caveats Formats FAQ Quick read Four GPUs, four different jobs B390 has the richer current iGPU feature set. Radeon 8060S has much more raster hardware and memory bandwidth. RX 9070 XT is built for high refresh 1440p and 4K gaming. RTX 5090 has the highest RT, path tracing and memory ceiling in this group. Intel iGPU Arc B390 12 Xe3 cores, XMX AI hardware, XeSS 3 and up to 4X output with supported multi-frame generation. AMD iGPU Radeon 8060S 40 RDNA 3.5 CUs, 256-bit shared LPDDR5X and much higher raw raster throughput than a standard laptop iGPU. AMD dGPU RX 9070 XT 64 RDNA 4 CUs, 16 GB GDDR6, 640 GB/s bandwidth and stronger RT hardware than RDNA 3.5. NVIDIA dGPU RTX 5090 32 GB GDDR7, 1,792 GB/s bandwidth and the highest RT and path tracing performance in this comparison. iGPUs vs Discrete GPUs Comparison | Faceofit.com F Faceofit.com PC hardware, mobile tech and buying comparisons GPU comparison guide iGPUs vs Discrete GPUs Comparison Intel Arc B390 and AMD Radeon 8060S sit at the top end of integrated graphics. This page compares them with Radeon RX 9070 XT and GeForce RTX 5090 across processor options, memory bandwidth, gaming, ray tracing and AI rendering features. Updated January 2026 No pricing used Mobile responsive Interactive filters and charts Summary iGPU processor list Laptops and mini PCs Hardware Memory Gaming Ray tracing AI features Scorecard Charts Caveats Formats FAQ Quick read Four GPUs, four different jobs B390 has the richer current iGPU feature set. Radeon 8060S has much more raster hardware and memory bandwidth. RX 9070 XT is built for high refresh 1440p and 4K gaming. RTX 5090 has the highest RT, path tracing and memory ceiling in this group. Intel iGPU Arc B390 12 Xe3 cores, XMX AI hardware, XeSS 3 and up to 4X output with supported multi-frame generation. AMD iGPU Radeon 8060S 40 RDNA 3.5 CUs, 256-bit shared LPDDR5X and much higher raw raster throughput than a standard laptop iGPU. AMD dGPU RX 9070 XT 64 RDNA 4 CUs, 16 GB GDDR6, 640 GB/s bandwidth and stronger RT hardware than RDNA 3.5. NVIDIA dGPU RTX 5090 32 GB GDDR7, 1,792 GB/s bandwidth and the highest RT and path tracing performance in this comparison. Processor availability Which processors include Arc B390 or Radeon 8060S? The iGPU name alone does not identify the processor. These tables list the processor SKUs discussed in this guide, so buyers can check whether a laptop, mini PC or compact desktop uses the full B390 or 8060S graphics configuration. Intel Xe3 Arc B390 processor list These Core Ultra Series 3 processors use the full Arc B390 iGPU with 12 Xe3 cores. Lower Series 3 processors can use Arc B370 or smaller Intel Graphics configurations. Processor CPU cores Max CPU boost Cache iGPU GPU architecture Launch Intel Core Ultra X7 358H164.8 GHz18 MBArc B390Xe3Q1 2026 Intel Core Ultra X7 368H165.0 GHz18 MBArc B390Xe3Q1 2026 Intel Core Ultra X9 378H165.0 GHz18 MBArc B390Xe3Q1 2026 Intel Core Ultra X9 388H165.1 GHz18 MBArc B390Xe3Q1 2026 For a gaming-led B390 system, Core Ultra X7 358H is worth checking because it keeps the full B390 graphics configuration without requiring the top X9 CPU tier. AMD RDNA 3.5 Radeon 8060S processor list The Ryzen AI Max+ 388, 392 and 395 keep the full 40-CU Radeon 8060S. The CPU core count changes, but the graphics configuration remains at 40 CUs in these listed models. Processor CPU cores / threads Max CPU boost Cache iGPU GPU CUs GPU clock cTDP Ryzen AI Max+ 3888 / 165.0 GHz40 MBRadeon 8060S402.9 GHz45 to 120 W Ryzen AI Max+ 39212 / 245.0 GHz76 MBRadeon 8060S402.9 GHz45 to 120 W Ryzen AI Max+ 39516 / 325.1 GHz80 MBRadeon 8060S40about 2.9 GHz45 to 120 W Ryzen AI Max+ PRO 39516 / 325.1 GHz80 MBRadeon 8060S40about 2.9 GHz45 to 120 W Ryzen AI Max+ 388 is the gaming-focused value point in this list because it keeps all 40 graphics CUs while using fewer CPU cores than the 395. Systems using these iGPUs Laptops, mini PCs and compact workstations with Arc B390 or Radeon 8060S As of 5 September 2026, there is a reasonably large set of shipping or announced systems using CPUs that carry Intel Arc B390 or Radeon 8060S. The AMD side has far more mini PCs, while Intel B390 appears in more conventional premium laptops. Amazon search links Affiliate search links. Availability, seller and configuration can change. No fixed product pricing is shown on this page. Check Intel Arc B390 on AmazonCheck Intel Core Ultra Series 3 on AmazonCheck AMD Radeon 8060S on AmazonCheck AMD Ryzen AI Max on Amazon Intel Arc B390 laptops LaptopB390 CPU configurationType and notesSourceASUS ExpertBook Ultra B9406CAACore Ultra X7 358H14-inch business ultraportable. ASUS India lists the X7 and B390 configuration through retail partners.ASUSASUS Zenbook DUO UX8407Core Ultra X9 388HDual 14-inch OLED. Full 12-Xe3 B390 configuration.ASUSDell XPS 14 2026Core Ultra X7 358HPremium thin laptop. B390 configuration has been independently tested.NotebookcheckDell XPS 16 2026Core Ultra X7 358H, select X9 388H configsDell lists Arc graphics with the 358H and 388H configurations. Available through Dell India.DellFramework Laptop 13 ProCore Ultra X7 358H or X9 388HUpgradeable 13-inch system. Both B390 configurations have appeared in testing.PhoronixMSI Prestige 14 Flip AI+ 2026Core Ultra X7 358HConvertible 14-inch system. B390 configuration has been independently tested.NotebookcheckMSI Prestige 16 AI+ C3MTGCore Ultra X7 358H16-inch productivity laptop. B390 appears in benchmark records and reviews.GeekbenchAcer Swift 16 AI SF16-71TUp to Core Ultra X9 388HAcer officially specifies up to Arc B390.AcerAcer Swift Go 16Core Ultra X7 358H or X9 388HB390 configurations reached selected Asian markets in September 2026.NotebookcheckAcer Aspire X 16 AIUp to Core Ultra X9 388H16-inch Copilot+ system, up to 12 Xe3 cores and B390.AcerLenovo ThinkPad X9-15p Gen 1 Aura EditionCore Ultra X7 368H or X9 388HBoth CPU options explicitly use Arc B390.Lenovo PSREFLenovo ThinkPad P1 Gen 9Core Ultra X7 368H or X9 388HMobile workstation. B390 is available alongside optional RTX Pro GPUs.Lenovo PSREFLenovo IdeaPad Pro 5i 14 Gen 11, 14IPH11Up to Core Ultra X9 388HTop Panther Lake configuration carries B390.Lenovo PSREF Additional Panther Lake notebooks are covered in a current roundup that identifies the X7 and X9 300H plus B390 tier. Ultrabook Review Panther Lake roundup. Intel Arc B390 mini PCs The list is considerably shorter than the B390 laptop list. Mini PCCPUMemory options and notesStatusSourceASUS NUC 16 ProCore Ultra X7 or X9, including X9 388HUp to 96 GB LPDDR5X on X7 and X9 modelsOfficial 2026 productASUSKhadas Mind ProCore Ultra X7 358H or X9 388H32 or 64 GB on X7, up to 96 GB on X9AvailableKhadasLenovo Yoga Mini 01IPH11, Yoga Mini i Gen 11Core Ultra X7 358H32 GB LPDDR5X in many marketsOfficial productLenovo PSREFGPD BOXCore Ultra X7 358H32 or 64 GB LPDDR5X, MCIO and USB4 v2ListedGPDNIMO Mini PC 388H 1LCore Ultra X9 388HDetails still emergingComing soon as of September 2026NIMO B390 mini PC to watch Khadas Mind Pro X7 358H Khadas allows the B390 to run in a relatively high-power implementation, and the X7 version has been independently tested. ASUS NUC 16 Pro is the more conventional business and mini-workstation option. Notebookcheck Khadas Mind Pro review AMD Radeon 8060S laptops The 8060S appears with Ryzen AI Max+ 388, 392, 395 and Max+ PRO 395. AMD lists the 388 and 392 with the full 40-CU Radeon 8060S despite their lower CPU core counts than the 395. LaptopCPUMemory and positioningSourceASUS TUF Gaming A14 FA401EA 2026Ryzen AI Max+ 39232 GB LPDDR5X. 14-inch 2.5K 165 Hz. Gaming-focused 8060S implementation.ASUSASUS ROG Flow Z13 GZ302EARyzen AI Max+ 39532, 64 or 128 GB depending on configuration. Gaming tablet and laptop hybrid.ASUS IndiaASUS ProArt GoPro Edition PX13 HN7306Ryzen AI Max+ 388 or 395, depending on configuration32 GB with some 388 configs, up to 128 GB with 395. 13-inch creator convertible.ASUSHP ZBook Ultra G1aRyzen AI Max+ PRO 395Up to 128 GB unified memory. 14-inch mobile workstation.HPNIMO Axis AI LaptopRyzen AI Max+ 395128 GB unified memory, up to 8 TB storage.NIMOSixunited XN77-160M-CSRyzen AI Max+ 395Up to 128 GB, 16-inch OLED. Availability outside China may be limited.TechRadar The ProArt PX13 with Max+ 388 is notable because the 388 retains all 40 Radeon 8060S CUs. Notebookcheck Radeon 8060S results. Radeon 8060S mini PCs and compact workstations This is the larger system category, with gaming-oriented mini PCs, modular desktops and local-AI workstations. SystemCPUMemoryCurrent positionSourceGEEKOM A9 MegaMax+ 388 or Max+ 39564 GB or 128 GBAvailable. One of the few 388 mini-PC choices. Both use 8060S.GEEKOMGMKtec EVO-X2Max+ 39564, 96 or 128 GBEstablished 395 mini PC.TechRadarFramework DesktopMax+ 39564 or 128 GBCompact modular desktop.TechRadarHP Z2 Mini G1aMax+ PRO 395Up to 128 GBMainstream workstation option.HPBeelink GTR9 ProMax+ 395Up to 128 GBDual 10GbE, local-AI-focused design.TechRadarBOSGAME M5 AIMax+ 395Up to 128 GB120 W-class implementation, dual USB4 and dual NVMe.TechRadarMINISFORUM MS-S1 MaxMax+ 395Up to 128 GBCompact workstation and server-style system. Independently benchmarked with 8060S.NotebookcheckCorsair AI Workstation 300Max+ 395 optionHigh-capacity LPDDR5XCompact local-AI workstation.TechRadarNIMO AI Mini PC 395, 2LMax+ 395128 GBCurrently listed by NIMO.NIMONIMO AI Mini PC 395, 4LMax+ 395128 GBLarger high-power NIMO version.NIMOAcer Veriton RA100 AI Mini WorkstationMax+ 395High-capacity unified memoryAnnounced at CES 2026.AcerAcemagic Tank M1A Pro+Max+ 395Up to 128 GB120 W-class 8060S implementation, already benchmarked.TechRadarFAVM FA-EX9Max+ 395Up to 128 GBAround 2 L, includes OCuLink.TechRadarAOOSTAR NEX395Max+ 395Up to 128 GBListed in China, global availability may be limited.TechRadarLCFC AI Mini PCMax+ 395128 GBLenovo-affiliated LCFC system, primarily China.TechRadarMINIX Elite ER939-AIMax+ 395High-capacity LPDDR5XAnnounced compact workstation.TechRadarPELADN YO1 395Max+ 395128 GBThree power modes, up to 120 W.TechRadarColorful Smart 900Max+ 39596 GBAnnounced 8060S mini PC.TechRadar TechRadar had catalogued more than 30 Ryzen AI Max+ 395 systems by January 2026, including both shipping and announced machines. TechRadar Ryzen AI Max+ system roundup. GEEKOM A9 Mega, Max+ 388 versus Max+ 395 A9 MegaCPUCPU coresiGPURAMLower configurationRyzen AI Max+ 3888C / 16TRadeon 8060S, 40 CU64 GBHigher configurationRyzen AI Max+ 39516C / 32TRadeon 8060S, 40 CU128 GB Dropping from the Max+ 395 to the Max+ 388 does not reduce the Radeon 8060S CU count. AMD specifies 40 CUs and a 2.9 GHz graphics frequency for the 388. For a machine bought mainly for gaming, the 388 or 392 can make more sense when GPU power limits and memory bandwidth are similar. AMD Ryzen AI Max+ 388 specifications · GEEKOM A9 Mega System shortlist B390 laptop gamingStart with Acer Swift 16 AIAlso compare MSI Prestige 16 AI+, ASUS Zenbook DUO and ThinkPad X9-15p. B390 mini PCKhadas Mind ProASUS NUC 16 Pro is the more conventional business and compact-workstation option. 8060S laptopASUS TUF Gaming A14The Max+ 392 version is aimed directly at gaming. ROG Flow Z13 offers Max+ 395 in a 2-in-1 format. 8060S mini PCGEEKOM A9 MegaAlso compare GMKtec EVO-X2, Framework Desktop, MINISFORUM MS-S1 Max and HP Z2 Mini G1a. Core specifications Hardware comparison Use the filters to hide GPUs you do not want to compare. The same filter selection also changes the interactive chart farther down the page. Show GPUs Arc B390 Radeon 8060S RX 9070 XT RTX 5090 Specification Intel Arc B390 Radeon 8060S RX 9070 XT RTX 5090 GPU typeIntegratedIntegratedDesktop dGPUDesktop dGPU ArchitectureXe3RDNA 3.5RDNA 4Blackwell Compute configuration12 Xe cores40 CUs64 CUs, 4,096 SP170 SM, 21,760 CUDA cores RT hardware12 RT units40 RT accelerators64 RT accelerators170 RT cores AI hardwareXMXRDNA AI acceleration128 AI accelerators680 Tensor cores Peak FP32about 7.7 TFLOPSabout 14.9 to 29.7 TFLOPS, depending on counting method48.7 TFLOPS104.8 TFLOPS Vendor AI figure122 INT8 TOPSNot directly comparable389 INT8 TOPS, 779 sparse3,352 AI TOPS MemoryShared LPDDR5XShared LPDDR5X16 GB GDDR632 GB GDDR7 Memory interface128-bit LPDDR5X-9600256-bit LPDDR5X-8000256-bit, 20 Gbps512-bit, 28 Gbps Peak bandwidthabout 153.6 GB/s shared256 GB/s shared640 GB/s1,792 GB/s Power envelopeabout 38 to 40 W GPU at full performance55 W default SoC, up to 120 W configurable304 W TBP575 W TGP Rendering suiteXeSS 3FSR 3.x, AFMF 2.1FSR RedstoneDLSS 4.5 and later RTX 50 features referenced in this comparison TFLOPS and TOPS are architecture-specific figures. RDNA 3.5 dual-issue accounting can produce a much larger FP32 headline number for 8060S than conventional single-issue counting. Memory subsystem Bandwidth changes the picture at 1440p, 4K and RT Both iGPUs share memory bandwidth with the CPU and other SoC blocks. The desktop cards use dedicated graphics memory, so more of their bandwidth is available to the GPU workload. Arc B390154 GB/s Radeon 8060S256 GB/s RX 9070 XT640 GB/s RTX 50901,792 GB/s 5090 vs B39011.7×Approximate peak bandwidth ratio. 5090 vs 8060S7×Approximate peak bandwidth ratio. 5090 vs 9070 XT2.8×Approximate peak bandwidth ratio. Real game targets Where each GPU fits by resolution Cyberpunk 2077 native-resolution figures show the broad spacing between the iGPUs and desktop GPUs. These are directional because the systems, CPUs and power limits differ. Cyberpunk 2077 nativeArc B390Radeon 8060SRX 9070 XTRTX 5090 1080p Ultraabout 45 to 48 FPSabout 87 to 90 FPSabout 188 FPSabout 185 FPS 1440p Ultraabout 28 FPSabout 54 FPSabout 134 FPSabout 182 FPS 4K UltraNot practicalabout 22 FPSabout 63 FPSabout 112 FPS Best native target1080p1080p and 1440p1440p and 4K4K Arc B3901080pMedium to high settings are the main target. XeSS can help at higher settings or with RT. Radeon 8060S1080p to 1440p1080p Ultra and 1440p High are realistic in many games when package power is available. RX 9070 XT1440p to 4KHigh refresh 1440p and 4K raster are the main use cases. RTX 50904K4K raster, heavy RT and path tracing are its main gaming targets. Desktop GPU normalized game-suite view This table compares the two desktop cards using the same broader game-suite view from the original report. Game-suite averageRX 9070 XTRTX 5090RTX 5090 advantage 1080p Ultra raster156.6 FPS203.8 FPSabout 30% 1440p Ultra raster116.5 FPS167.3 FPSabout 44% 4K Ultra raster65.8 FPS110.8 FPSabout 68% 1080p Ultra RT77.8 FPS125.7 FPSabout 62% 1440p Ultra RT55.5 FPS100.8 FPSabout 82% 4K Ultra RT30.1 FPS64.1 FPSabout 113% At 1080p, very fast desktop GPUs can become CPU limited. The gap grows at 1440p, 4K and RT because the GPU workload becomes the larger limit. Ray tracing RT separates the iGPUs from the desktop cards A heavy RT workload exposes the gap in RT hardware, memory bandwidth and total power. Speed Way figures below are approximate. Arc B390~932Light 1080p RTUse XeSS in demanding titlesHeavy path tracing is not a practical target Radeon 8060S~1,933Light to medium 1080p RTSome 1440p light RTUpscaling is useful RX 9070 XT~6,3701440p high RT4K RT with upscaling and FGRDNA 4 RT hardware RTX 5090~14,5704K high RTBest path tracing result hereRay Reconstruction support Rendering features Upscaling, frame generation and low latency Feature coverage and raw performance are separate issues. B390 is a good example because its XeSS feature list is strong even though its base render rate is much lower than a high-end desktop GPU. GeForce RTX 5090 DLSS Super Resolution→ Ray Reconstruction→ Multi Frame Generation→ Reflex Radeon RX 9070 XT FSR ML Upscaling→ Ray Regeneration→ ML Frame Generation→ Anti-Lag Intel Arc B390 XeSS Super Resolution→ XeSS Multi Frame Generation→ XeLL Radeon 8060S FSR 3.x Upscaling→ FSR 3 FG or AFMF 2.1→ Anti-Lag Generated frames can improve visible motion smoothness, but the base rendered frame rate still determines much of the latency, simulation response and the amount of newly rendered information. At a glance Capability scorecard These ratings summarize the practical use cases from the original comparison. They are qualitative, not benchmark scores. CategoryArc B390Radeon 8060SRX 9070 XTRTX 5090 1080p raster3 / 54.5 / 55 / 55 / 5 1440p raster2 / 54 / 55 / 55 / 5 4K raster1 / 52 / 54 / 55 / 5 Ray tracing2 / 52.5 / 54 / 55 / 5 Path tracing1 / 51.5 / 53 / 55 / 5 AI rendering features4.5 / 53 / 54.5 / 55 / 5 Performance per wattExcellentVery goodGoodLow Memory modelShared system memoryLarge shared system memory pool16 GB dedicated VRAM32 GB dedicated VRAM Interactive chart Switch between bandwidth, 1440p gaming and RT Use the buttons below to change the metric. GPU checkboxes in the hardware section also control what appears in this chart. Memory bandwidth Cyberpunk 1440p Speed Way RT Reading the numbers Caveats that matter for iGPU comparisons System power The same iGPU can perform differently Arc B390 and Radeon 8060S results depend on package power, memory speed, cooling, firmware and OEM settings. A low-power system can be materially slower than a fully powered one. Shared memory Peak bandwidth is shared The CPU and other SoC blocks also use the LPDDR memory fabric. Peak bandwidth is not reserved for the iGPU in the same way dedicated VRAM bandwidth is reserved for a desktop GPU. TFLOPS and TOPS Cross-vendor numbers need context Precision, sparsity, execution layout and dual-issue behavior differ by architecture. Headline compute figures are better used inside the same architecture family than as a direct cross-vendor ranking. Frame generation Displayed FPS is not base FPS Generated frames can make motion look smoother, but they do not raise the underlying simulation rate or create the same latency profile as rendering every frame natively. Template formats Three ways to read the comparison Different readers need different amounts of detail. These page formats keep the same data but present it for different decisions. Format 1 Quick spec table Use the hardware table when architecture, memory, bandwidth and power are the main filters. Open hardware view Format 2 Resolution guide Use the gaming and RT sections when the decision is based on 1080p, 1440p, 4K or ray tracing. Open gaming view Format 3 Feature flow Use the AI feature section when comparing upscaling, frame generation, RT reconstruction and low-latency tools. Open feature view FAQ Questions buyers ask about B390 and 8060S Which iGPU is faster, Arc B390 or Radeon 8060S? Radeon 8060S has much more raw raster hardware and a wider shared-memory interface. In demanding games, a fully powered 8060S can be close to twice as fast as B390 in some workloads. Which iGPU has the stronger current AI gaming feature list? Arc B390 has the stronger feature list in this comparison because XeSS 3 includes AI upscaling, frame generation, multi-frame generation and XeLL on supported games. Does Radeon 8060S support the full RDNA 4 FSR Redstone feature set? No. Radeon 8060S is RDNA 3.5. It can use FSR 3.x features and AFMF 2.1, but the full RDNA 4 path shown here belongs to newer discrete Radeon hardware. Can B390 or 8060S replace RX 9070 XT for 4K gaming? No. The dedicated RX 9070 XT has far more graphics throughput and 640 GB/s of dedicated memory bandwidth. It is built for 1440p and 4K gaming in a way these iGPUs are not. Why does RTX 5090 pull farther ahead when ray tracing is enabled? RT performance depends on RT hardware, compute throughput, memory bandwidth and the rendering stack. RTX 5090 has much more of each than the iGPUs, and it also supports Ray Reconstruction in compatible games. Does multi-frame generation make B390 equal to a GPU that renders the same FPS natively? No. A high displayed FPS from generated frames does not create the same base simulation rate or input latency as rendering every frame natively. What should I check before buying a laptop or mini PC with one of these iGPUs? Check the exact processor SKU, memory speed, package power limit, cooling design and whether the device uses the full B390 or 8060S graphics configuration. iGPU performance can move a lot between systems using the same processor family. Faceofit.com buying note Choose by workload, not by the feature checklist alone Arc B390 suits thin systems where XeSS features matter and 1080p is the main target. Radeon 8060S is the stronger iGPU when raster speed and a large shared-memory pool matter. RX 9070 XT is the step up for 1440p and 4K. RTX 5090 sits at the top for 4K RT, path tracing and memory bandwidth. Core specifications Hardware comparison Use the filters to hide GPUs you do not want to compare. The same filter selection also changes the interactive chart farther down the page. Show GPUs Arc B390 Radeon 8060S RX 9070 XT RTX 5090 Specification Intel Arc B390 Radeon 8060S RX 9070 XT RTX 5090 GPU typeIntegratedIntegratedDesktop dGPUDesktop dGPU ArchitectureXe3RDNA 3.5RDNA 4Blackwell Compute configuration12 Xe cores40 CUs64 CUs, 4,096 SP170 SM, 21,760 CUDA cores RT hardware12 RT units40 RT accelerators64 RT accelerators170 RT cores AI hardwareXMXRDNA AI acceleration128 AI accelerators680 Tensor cores Peak FP32about 7.7 TFLOPSabout 14.9 to 29.7 TFLOPS, depending on counting method48.7 TFLOPS104.8 TFLOPS Vendor AI figure122 INT8 TOPSNot directly comparable389 INT8 TOPS, 779 sparse3,352 AI TOPS MemoryShared LPDDR5XShared LPDDR5X16 GB GDDR632 GB GDDR7 Memory interface128-bit LPDDR5X-9600256-bit LPDDR5X-8000256-bit, 20 Gbps512-bit, 28 Gbps Peak bandwidthabout 153.6 GB/s shared256 GB/s shared640 GB/s1,792 GB/s Power envelopeabout 38 to 40 W GPU at full performance55 W default SoC, up to 120 W configurable304 W TBP575 W TGP Rendering suiteXeSS 3FSR 3.x, AFMF 2.1FSR RedstoneDLSS 4.5 and later RTX 50 features referenced in this comparison TFLOPS and TOPS are architecture-specific figures. RDNA 3.5 dual-issue accounting can produce a much larger FP32 headline number for 8060S than conventional single-issue counting. Memory subsystem Bandwidth changes the picture at 1440p, 4K and RT Both iGPUs share memory bandwidth with the CPU and other SoC blocks. The desktop cards use dedicated graphics memory, so more of their bandwidth is available to the GPU workload. Arc B390154 GB/s Radeon 8060S256 GB/s RX 9070 XT640 GB/s RTX 50901,792 GB/s 5090 vs B39011.7×Approximate peak bandwidth ratio. 5090 vs 8060S7×Approximate peak bandwidth ratio. 5090 vs 9070 XT2.8×Approximate peak bandwidth ratio. Real game targets Where each GPU fits by resolution Cyberpunk 2077 native-resolution figures show the broad spacing between the iGPUs and desktop GPUs. These are directional because the systems, CPUs and power limits differ. Cyberpunk 2077 nativeArc B390Radeon 8060SRX 9070 XTRTX 5090 1080p Ultraabout 45 to 48 FPSabout 87 to 90 FPSabout 188 FPSabout 185 FPS 1440p Ultraabout 28 FPSabout 54 FPSabout 134 FPSabout 182 FPS 4K UltraNot practicalabout 22 FPSabout 63 FPSabout 112 FPS Best native target1080p1080p and 1440p1440p and 4K4K Arc B3901080pMedium to high settings are the main target. XeSS can help at higher settings or with RT. Radeon 8060S1080p to 1440p1080p Ultra and 1440p High are realistic in many games when package power is available. RX 9070 XT1440p to 4KHigh refresh 1440p and 4K raster are the main use cases. RTX 50904K4K raster, heavy RT and path tracing are its main gaming targets. Desktop GPU normalized game-suite view This table compares the two desktop cards using the same broader game-suite view from the original report. Game-suite averageRX 9070 XTRTX 5090RTX 5090 advantage 1080p Ultra raster156.6 FPS203.8 FPSabout 30% 1440p Ultra raster116.5 FPS167.3 FPSabout 44% 4K Ultra raster65.8 FPS110.8 FPSabout 68% 1080p Ultra RT77.8 FPS125.7 FPSabout 62% 1440p Ultra RT55.5 FPS100.8 FPSabout 82% 4K Ultra RT30.1 FPS64.1 FPSabout 113% At 1080p, very fast desktop GPUs can become CPU limited. The gap grows at 1440p, 4K and RT because the GPU workload becomes the larger limit. Ray tracing RT separates the iGPUs from the desktop cards A heavy RT workload exposes the gap in RT hardware, memory bandwidth and total power. Speed Way figures below are approximate. Arc B390~932Light 1080p RTUse XeSS in demanding titlesHeavy path tracing is not a practical target Radeon 8060S~1,933Light to medium 1080p RTSome 1440p light RTUpscaling is useful RX 9070 XT~6,3701440p high RT4K RT with upscaling and FGRDNA 4 RT hardware RTX 5090~14,5704K high RTBest path tracing result hereRay Reconstruction support Rendering features Upscaling, frame generation and low latency Feature coverage and raw performance are separate issues. B390 is a good example because its XeSS feature list is strong even though its base render rate is much lower than a high-end desktop GPU. GeForce RTX 5090 DLSS Super Resolution→ Ray Reconstruction→ Multi Frame Generation→ Reflex Radeon RX 9070 XT FSR ML Upscaling→ Ray Regeneration→ ML Frame Generation→ Anti-Lag Intel Arc B390 XeSS Super Resolution→ XeSS Multi Frame Generation→ XeLL Radeon 8060S FSR 3.x Upscaling→ FSR 3 FG or AFMF 2.1→ Anti-Lag Generated frames can improve visible motion smoothness, but the base rendered frame rate still determines much of the latency, simulation response and the amount of newly rendered information. At a glance Capability scorecard These ratings summarize the practical use cases from the original comparison. They are qualitative, not benchmark scores. CategoryArc B390Radeon 8060SRX 9070 XTRTX 5090 1080p raster3 / 54.5 / 55 / 55 / 5 1440p raster2 / 54 / 55 / 55 / 5 4K raster1 / 52 / 54 / 55 / 5 Ray tracing2 / 52.5 / 54 / 55 / 5 Path tracing1 / 51.5 / 53 / 55 / 5 AI rendering features4.5 / 53 / 54.5 / 55 / 5 Performance per wattExcellentVery goodGoodLow Memory modelShared system memoryLarge shared system memory pool16 GB dedicated VRAM32 GB dedicated VRAM Interactive chart Switch between bandwidth, 1440p gaming and RT Use the buttons below to change the metric. GPU checkboxes in the hardware section also control what appears in this chart. Memory bandwidth Cyberpunk 1440p Speed Way RT Reading the numbers Caveats that matter for iGPU comparisons System power The same iGPU can perform differently Arc B390 and Radeon 8060S results depend on package power, memory speed, cooling, firmware and OEM settings. A low-power system can be materially slower than a fully powered one. Shared memory Peak bandwidth is shared The CPU and other SoC blocks also use the LPDDR memory fabric. Peak bandwidth is not reserved for the iGPU in the same way dedicated VRAM bandwidth is reserved for a desktop GPU. TFLOPS and TOPS Cross-vendor numbers need context Precision, sparsity, execution layout and dual-issue behavior differ by architecture. Headline compute figures are better used inside the same architecture family than as a direct cross-vendor ranking. Frame generation Displayed FPS is not base FPS Generated frames can make motion look smoother, but they do not raise the underlying simulation rate or create the same latency profile as rendering every frame natively. Template formats Three ways to read the comparison Different readers need different amounts of detail. These page formats keep the same data but present it for different decisions. Format 1 Quick spec table Use the hardware table when architecture, memory, bandwidth and power are the main filters. Open hardware view Format 2 Resolution guide Use the gaming and RT sections when the decision is based on 1080p, 1440p, 4K or ray tracing. Open gaming view Format 3 Feature flow Use the AI feature section when comparing upscaling, frame generation, RT reconstruction and low-latency tools. Open feature view FAQ Questions buyers ask about B390 and 8060S Which iGPU is faster, Arc B390 or Radeon 8060S? Radeon 8060S has much more raw raster hardware and a wider shared-memory interface. In demanding games, a fully powered 8060S can be close to twice as fast as B390 in some workloads. Which iGPU has the stronger current AI gaming feature list? Arc B390 has the stronger feature list in this comparison because XeSS 3 includes AI upscaling, frame generation, multi-frame generation and XeLL on supported games. Does Radeon 8060S support the full RDNA 4 FSR Redstone feature set? No. Radeon 8060S is RDNA 3.5. It can use FSR 3.x features and AFMF 2.1, but the full RDNA 4 path shown here belongs to newer discrete Radeon hardware. Can B390 or 8060S replace RX 9070 XT for 4K gaming? No. The dedicated RX 9070 XT has far more graphics throughput and 640 GB/s of dedicated memory bandwidth. It is built for 1440p and 4K gaming in a way these iGPUs are not. Why does RTX 5090 pull farther ahead when ray tracing is enabled? RT performance depends on RT hardware, compute throughput, memory bandwidth and the rendering stack. RTX 5090 has much more of each than the iGPUs, and it also supports Ray Reconstruction in compatible games. Does multi-frame generation make B390 equal to a GPU that renders the same FPS natively? No. A high displayed FPS from generated frames does not create the same base simulation rate or input latency as rendering every frame natively. What should I check before buying a laptop or mini PC with one of these iGPUs? Check the exact processor SKU, memory speed, package power limit, cooling design and whether the device uses the full B390 or 8060S graphics configuration. iGPU performance can move a lot between systems using the same processor family. Faceofit.com buying note Choose by workload, not by the feature checklist alone Arc B390 suits thin systems where XeSS features matter and 1080p is the main target. Radeon 8060S is the stronger iGPU when raster speed and a large shared-memory pool matter. RX 9070 XT is the step up for 1440p and 4K. RTX 5090 sits at the top for 4K RT, path tracing and memory bandwidth. Affiliate Disclosure: Faceofit.com is a participant in the Amazon Services LLC Associates Program. As an Amazon Associate we earn from qualifying purchases. Share What's your reaction? Excited 0 Happy 0 In Love 0 Not Sure 0 Silly 0
Intel iGPU Arc B390 12 Xe3 cores, XMX AI hardware, XeSS 3 and up to 4X output with supported multi-frame generation.
AMD iGPU Radeon 8060S 40 RDNA 3.5 CUs, 256-bit shared LPDDR5X and much higher raw raster throughput than a standard laptop iGPU.
AMD dGPU RX 9070 XT 64 RDNA 4 CUs, 16 GB GDDR6, 640 GB/s bandwidth and stronger RT hardware than RDNA 3.5.
NVIDIA dGPU RTX 5090 32 GB GDDR7, 1,792 GB/s bandwidth and the highest RT and path tracing performance in this comparison.
Intel iGPU Arc B390 12 Xe3 cores, XMX AI hardware, XeSS 3 and up to 4X output with supported multi-frame generation.
AMD iGPU Radeon 8060S 40 RDNA 3.5 CUs, 256-bit shared LPDDR5X and much higher raw raster throughput than a standard laptop iGPU.
AMD dGPU RX 9070 XT 64 RDNA 4 CUs, 16 GB GDDR6, 640 GB/s bandwidth and stronger RT hardware than RDNA 3.5.
NVIDIA dGPU RTX 5090 32 GB GDDR7, 1,792 GB/s bandwidth and the highest RT and path tracing performance in this comparison.
B390 laptop gamingStart with Acer Swift 16 AIAlso compare MSI Prestige 16 AI+, ASUS Zenbook DUO and ThinkPad X9-15p.
B390 mini PCKhadas Mind ProASUS NUC 16 Pro is the more conventional business and compact-workstation option.
8060S laptopASUS TUF Gaming A14The Max+ 392 version is aimed directly at gaming. ROG Flow Z13 offers Max+ 395 in a 2-in-1 format.
8060S mini PCGEEKOM A9 MegaAlso compare GMKtec EVO-X2, Framework Desktop, MINISFORUM MS-S1 Max and HP Z2 Mini G1a.
System power The same iGPU can perform differently Arc B390 and Radeon 8060S results depend on package power, memory speed, cooling, firmware and OEM settings. A low-power system can be materially slower than a fully powered one.
Shared memory Peak bandwidth is shared The CPU and other SoC blocks also use the LPDDR memory fabric. Peak bandwidth is not reserved for the iGPU in the same way dedicated VRAM bandwidth is reserved for a desktop GPU.
TFLOPS and TOPS Cross-vendor numbers need context Precision, sparsity, execution layout and dual-issue behavior differ by architecture. Headline compute figures are better used inside the same architecture family than as a direct cross-vendor ranking.
Frame generation Displayed FPS is not base FPS Generated frames can make motion look smoother, but they do not raise the underlying simulation rate or create the same latency profile as rendering every frame natively.
Format 1 Quick spec table Use the hardware table when architecture, memory, bandwidth and power are the main filters. Open hardware view
Format 2 Resolution guide Use the gaming and RT sections when the decision is based on 1080p, 1440p, 4K or ray tracing. Open gaming view
Format 3 Feature flow Use the AI feature section when comparing upscaling, frame generation, RT reconstruction and low-latency tools. Open feature view
System power The same iGPU can perform differently Arc B390 and Radeon 8060S results depend on package power, memory speed, cooling, firmware and OEM settings. A low-power system can be materially slower than a fully powered one.
Shared memory Peak bandwidth is shared The CPU and other SoC blocks also use the LPDDR memory fabric. Peak bandwidth is not reserved for the iGPU in the same way dedicated VRAM bandwidth is reserved for a desktop GPU.
TFLOPS and TOPS Cross-vendor numbers need context Precision, sparsity, execution layout and dual-issue behavior differ by architecture. Headline compute figures are better used inside the same architecture family than as a direct cross-vendor ranking.
Frame generation Displayed FPS is not base FPS Generated frames can make motion look smoother, but they do not raise the underlying simulation rate or create the same latency profile as rendering every frame natively.
Format 1 Quick spec table Use the hardware table when architecture, memory, bandwidth and power are the main filters. Open hardware view
Format 2 Resolution guide Use the gaming and RT sections when the decision is based on 1080p, 1440p, 4K or ray tracing. Open gaming view
Format 3 Feature flow Use the AI feature section when comparing upscaling, frame generation, RT reconstruction and low-latency tools. Open feature view
8 Anti-Aliasing Techniques Explained: Compare FXAA, TAA, DLAA and GPU Performance Anti-aliasing techniques reduce jagged edges and spatial aliasing in rendered images to produce smoother graphics. ...
PC USB-C to Dual HDMI Adapters: 4K 60Hz and 120Hz Hardware Guide Pushing two 4K monitors from a single laptop port requires massive bandwidth. As multi-monitor workstations ...
PC Best GPU Support Brackets and Anti-Sag Stands for Heavy Graphics Cards Modern desktop graphics processing units exceed physical lengths of 330 millimeters and weights of three ...
PC MeLE Quieter 4C vs. 3Q/3C Review: N100 Benchmarks, Thermal Analysis & Linux Compatibility Fanless x86 architecture has matured, moving beyond sluggish Atom processors to capable silicon suitable for ...
PC EDID Emulator & Passthrough Adapter Guide: Fixing HDMI Handshakes, KVM Scrambling, and Ghost Monitors Modern display chains—involving KVM switches, splitters, and AV receivers—often break the digital handshake required for ...
PC Ceramic Shield vs. Gorilla Glass Victus 2 vs. Dragontrail Star 2 Your phone screen faces two persistent threats: concrete sidewalks and pocket sand. While manufacturers like ...