Search and filter Insights
Featured hardware analysis
Selected analyses and explainers, with their evidence limits stated.
Hardware analysis
How the GeForce RTX 5090 compares with the RTX 4090 on the UOS GPU Rendering Index, with neighbours, method and limits.
Measured evidence publishable · 31 ranked in domainUpdated
Upgrade insight
Where the RTX 4080 SUPER, RTX 4090, RTX 5080 and RTX 5090 sit relative to an RTX 3090 on one measured Blender rendering scale.
Measured evidence publishable · 31 ranked in domainUpdated
Methodology explainer
What each UpgradeOrSave purchase call means, which evidence it needs, and why most calls during the beta are NEED MORE DATA.
Method explainer · no measured valuesUpdated
GPU analysis
Graphics cards compared inside one measured workload domain at a time.
Hardware analysis
How the GeForce RTX 5090 compares with the RTX 4090 on the UOS GPU Rendering Index, with neighbours, method and limits.
Measured evidence publishable · 31 ranked in domainUpdated
Upgrade insight
Where the RTX 4080 SUPER, RTX 4090, RTX 5080 and RTX 5090 sit relative to an RTX 3090 on one measured Blender rendering scale.
Measured evidence publishable · 31 ranked in domainUpdated
Hardware analysis
Radeon RX 6000, 7000 and 9000 cards on the UOS HIP Rendering Index, measured on one AMD backend domain, with evidence depth noted.
Measured evidence publishable · 11 ranked in domainUpdated
Methodology explainer
Blender renders on different GPU backends. UpgradeOrSave compares only within one backend domain, and this page shows the effect.
Coverage states available · no measured valuesUpdated
Methodology explainer
Whether a local model runs well depends on model, quantization, context and backend. Without verified runtime profiles, we make no fit claims.
Method explainer · no measured valuesUpdated
CPU analysis
Processors compared on published, rights-cleared workload indices.
Upgrade insight
Where Ryzen 5000 processors land against a Ryzen 5 3600 on the UOS CPU Rendering Index, with verified socket facts only.
Measured evidence publishable · 43 ranked in domainUpdated
Upgrade insights
Where candidate hardware lands relative to what you own, with fit and price limits stated.
Hardware analysis
How the GeForce RTX 5090 compares with the RTX 4090 on the UOS GPU Rendering Index, with neighbours, method and limits.
Measured evidence publishable · 31 ranked in domainUpdated
Upgrade insight
Where the RTX 4080 SUPER, RTX 4090, RTX 5080 and RTX 5090 sit relative to an RTX 3090 on one measured Blender rendering scale.
Measured evidence publishable · 31 ranked in domainUpdated
Upgrade insight
Where Ryzen 5000 processors land against a Ryzen 5 3600 on the UOS CPU Rendering Index, with verified socket facts only.
Measured evidence publishable · 43 ranked in domainUpdated
Methodology explainer
What each UpgradeOrSave purchase call means, which evidence it needs, and why most calls during the beta are NEED MORE DATA.
Method explainer · no measured valuesUpdated
Benchmark methodology explainers
How UpgradeOrSave builds, gates and presents its evidence.
Methodology explainer
What each UpgradeOrSave purchase call means, which evidence it needs, and why most calls during the beta are NEED MORE DATA.
Method explainer · no measured valuesUpdated
Methodology explainer
Displaying a price and using it in advice are separate permissions. Until a source grants both, UpgradeOrSave shows no prices.
Method explainer · no measured valuesUpdated
Methodology explainer
Why a product can be ranked in one workload and absent from another, and what indexed, under review and not approved mean.
Method explainer · no measured valuesUpdated
Methodology explainer
Gaming, rendering, local AI and productivity answer different questions. Here is why each keeps its own scale and reference.
Method explainer · no measured valuesUpdated
Methodology explainer
Blender renders on different GPU backends. UpgradeOrSave compares only within one backend domain, and this page shows the effect.
Coverage states available · no measured valuesUpdated
Methodology explainer
Whether a local model runs well depends on model, quantization, context and backend. Without verified runtime profiles, we make no fit claims.
Method explainer · no measured valuesUpdated