# Understanding GPU Naming Conventions
GPU naming is complex, with different meaning for NVIDIA, AMD, and Intel. Understanding conventions helps you compare cards meaningfully and identify the right GPU for your needs.
NVIDIA GeForce Naming
Format: GeForce [RTX/GTX] [Generation][Tier][Variant]
Example: GeForce RTX 4070 Super
Breakdown:
- GeForce — Product line
- RTX — Ray Tracing capable (modern)
- 40 — Generation (RTX 40 series)
- 70 — Tier within generation
- Super — Enhanced variant
RTX vs GTX
RTX — Modern cards with ray tracing hardware (RT cores) and AI cores (Tensor cores).
GTX — Older cards without ray tracing (GTX 16 series is exception with modern features).
All new NVIDIA cards are RTX. GTX cards are legacy budget options.
NVIDIA Tier Explanations
xx50 — Entry-level (basic gaming, 1080p low-medium settings).
xx60 — Mainstream (1080p high, 1440p medium).
xx70 — High-end (1440p high, 1080p ultra with high frames).
xx80 — Premium (1440p ultra, 4K high).
xx90 — Enthusiast/professional (4K ultra, workstation).
Higher tier = better performance and higher price. Match to monitor resolution.
NVIDIA Variant Suffixes
Ti (Titanium) — Enhanced version of base model.
Example: RTX 4070 Ti (better than 4070, worse than 4080)
Super — Refresh with better specs than base.
Example: RTX 4070 Super (between 4070 and 4070 Ti)
Ti Super — Combined enhancement.
Example: RTX 4070 Ti Super (better than Ti, close to 4080)
No suffix — Standard model.
Modern NVIDIA lineup features many variants creating fine-grained performance tiers.
NVIDIA Generation History
GTX 900 series (Maxwell) — 2014, mature architecture.
GTX 1000 series (Pascal) — 2016, huge efficiency improvement.
GTX 1600 series (Turing) — 2019, mainstream Turing without RT.
RTX 2000 series (Turing) — 2018-2019, first RTX generation.
RTX 3000 series (Ampere) — 2020-2022, massive RT/AI improvement.
RTX 4000 series (Ada Lovelace) — 2022-2023, efficiency-focused.
RTX 5000 series (Blackwell) — 2024-2025, latest generation.
Each generation typically improves performance 30-50% at same tier.
AMD Radeon Naming
Format: Radeon [Family] [Generation][Tier] [Variant]
Example: Radeon RX 7900 XT
Breakdown:
- Radeon — Product line
- RX — Consumer gaming line
- 7900 — Generation and tier (RX 7000 series, top tier)
- XT — Enhanced variant
AMD Tier Explanations
RX x400 — Entry-level.
RX x500 — Budget mainstream.
RX x600 — Mainstream.
RX x700 — High-end.
RX x800 — Premium.
RX x900 — Enthusiast.
AMD tiers roughly parallel NVIDIA tiers but numbering differs.
AMD Variant Suffixes
XT — Enhanced version.
Example: RX 7900 XT (higher than 7900, if it existed)
XTX — Maximum enhanced.
Example: RX 7900 XTX (top consumer card)
GRE — Gaming Reference Edition (China market originally, later global).
Example: RX 7900 GRE
No suffix — Base model.
AMD Generation History
RX 400 series (Polaris) — 2016, budget-mainstream focus.
RX 500 series (Polaris refresh) — 2017, refined 400 series.
RX 5000 series (RDNA) — 2019, new architecture.
RX 6000 series (RDNA 2) — 2020-2021, competitive with RTX 3000.
RX 7000 series (RDNA 3) — 2022-2023, competitive with RTX 4000.
RX 8000 series (RDNA 4) — 2024-2025, latest generation.
Intel Arc GPU Naming
Format: Arc [Series] [Tier][Variant]
Example: Arc A770
Breakdown:
- Arc — Product line
- A — First generation architecture (Alchemist)
- 7 — Tier
- 70 — Model
Intel Arc Details
Arc A380 — Entry-level.
Arc A580 — Budget mainstream.
Arc A750 — Mainstream.
Arc A770 — Premium (of first-gen Arc).
Arc B-series — Second generation (Battlemage), launched 2024-2025.
Intel Arc excellent for content creation (AV1 encoding) but less mature drivers than NVIDIA/AMD.
Reading GPU Reviews
Focus on these metrics:
Rasterization performance — Traditional rendering (1080p/1440p/4K FPS).
Ray tracing performance — Modern lighting simulation.
Content creation — Video encoding, 3D rendering benchmarks.
Power consumption — Idle and load wattage.
Thermals — Temperature under load.
Noise levels — Cooler performance.
Match reviewer priorities to your use case.
Understanding VRAM
GDDR6/GDDR6X — Memory used on modern GPUs.
VRAM capacity:
- 4-6GB: Entry level, budget
- 8GB: Mainstream 1080p/1440p
- 12GB: High-end 1440p, 4K medium
- 16GB: Premium 4K, professional
- 24GB+: Enthusiast, workstation
VRAM speed: GDDR6X faster than GDDR6. Wide memory bus increases bandwidth.
More VRAM helps with:
- 4K resolution
- Ray tracing
- Content creation
- Machine learning workloads
Bandwidth Considerations
Memory bandwidth = memory speed × bus width.
Typical values:
- 128-bit bus: Lower bandwidth (RTX 4060, RX 7600)
- 192-bit bus: Mid bandwidth (RTX 4070 tier)
- 256-bit bus: Higher bandwidth (RTX 4070 Ti tier)
- 384-bit bus: High bandwidth (RTX 4080, 4090)
Bandwidth matters for higher resolutions and complex workloads.
RT Cores and Tensor Cores
RT Cores — Dedicated ray tracing hardware. More cores = better RT performance.
Tensor Cores — AI acceleration (DLSS, some content creation).
Generation matters:
- RT Cores 3rd gen (RTX 30 series)
- RT Cores 4th gen (RTX 40 series)
Newer generation cores work more efficiently.
AMD Ray Tracing
RDNA 2 — First AMD RT implementation, slower than NVIDIA.
RDNA 3 — Improved RT, still behind NVIDIA.
RDNA 4 — Improved further.
For maximum ray tracing performance, NVIDIA leads. AMD better price/performance in raster.
DLSS and FSR
DLSS (Deep Learning Super Sampling) — NVIDIA AI upscaling.
DLSS 3 — Adds frame generation.
DLSS 4 — Latest, more refined.
FSR (FidelityFX Super Resolution) — AMD's open-source alternative.
FSR 3 — Includes frame generation.
Both technologies boost frame rates. DLSS generally higher quality; FSR works on any GPU.
Reference vs AIB Cards
Reference/Founder's Edition — Made by NVIDIA/AMD directly.
AIB (Add-in Board) partners — Third-party manufacturers.
Common AIB brands:
- NVIDIA partners: ASUS, MSI, Gigabyte, EVGA (discontinued), Zotac
- AMD partners: ASUS, MSI, Gigabyte, PowerColor, Sapphire, XFX
AIB cards typically feature:
- Custom coolers (usually better)
- Slight factory overclock
- Different aesthetics
- Higher price sometimes
Choosing Between Cards
Same tier, different manufacturers (e.g., MSI 4070 vs ASUS 4070):
- Cooler quality varies
- Warranty terms differ
- Noise levels vary
- Aesthetic differences
Different generation, same tier (RTX 3080 vs 4080):
- Newer better
- Newer more expensive
- Similar tier positioning
Cross-brand comparison (RTX 4070 vs RX 7800 XT):
- Depends on workload
- Ray tracing: NVIDIA usually wins
- Raster: Often similar
- Price: AMD sometimes cheaper
Common Model Confusion
RTX 4070 vs RTX 4070 Ti — Ti is stronger, roughly RTX 4070 Super or slightly better.
RTX 4070 Super vs RTX 4070 Ti — Super is between 4070 and Ti. Ti is stronger.
RX 7900 GRE vs RX 7900 XT — GRE weaker than XT.
Founders Edition vs Custom — Usually FE cheaper but louder.
Power Requirements
Modern GPU TDP:
- RTX 4060: 115W
- RTX 4070: 200W
- RTX 4080: 320W
- RTX 4090: 450W (transient spikes higher)
Recommended PSU wattage:
- RTX 4060: 550W minimum
- RTX 4070: 650W
- RTX 4080: 750W
- RTX 4090: 850W minimum, 1000W recommended
Quality PSUs matter more than raw wattage.
Cooling Considerations
Standard cooler — 2-2.5 slot, adequate.
Premium cooler — 3-3.5 slot, quieter and cooler.
Extreme cooler — 4 slot, silent operation.
Higher-end AIB cards typically have better cooling. Consider case clearance.
PCIe Requirements
PCIe 3.0 — All modern GPUs work in Gen3 slots but potentially slight performance loss.
PCIe 4.0 — Standard for most modern platforms.
PCIe 5.0 — Latest, minimal benefit for current GPUs.
Older systems (PCIe 3.0) can use modern GPUs but may bottleneck highest-end cards slightly.
Recommended by Use Case
1080p gaming:
- Entry: RTX 4060, RX 7600
- Mainstream: RTX 4060 Ti, RX 7600 XT
- Enthusiast: Anything above overkill
1440p gaming:
- Entry: RTX 4060 Ti, RX 7700 XT
- Mainstream: RTX 4070, RX 7800 XT
- Enthusiast: RTX 4070 Ti, RX 7900 XT
4K gaming:
- Entry: RTX 4070 Ti, RX 7900 XT
- Mainstream: RTX 4080, RX 7900 XTX
- Enthusiast: RTX 4090
Content creation:
- Entry: RTX 4060, RX 7600
- Mainstream: RTX 4070, Intel Arc A750 (AV1)
- Enthusiast: RTX 4080, RTX 4090
Ray Tracing Performance Tiers
Ray tracing typically requires higher-tier cards:
- RTX 4060 for RT at 1080p with DLSS
- RTX 4070 for RT at 1440p with DLSS
- RTX 4080 for RT at 4K with DLSS
- RTX 4090 for RT at 4K without DLSS
AMD equivalent cards perform 30-40% worse in RT.
Content Creation Considerations
Video editing:
- NVIDIA NVENC excellent
- AMD AMF improving
- Intel Quick Sync + AV1 (Arc) excellent
3D rendering:
- Blender OptiX (NVIDIA) faster than HIP (AMD)
- Number of RT cores helps
- VRAM capacity matters
AI/ML:
- NVIDIA CUDA ecosystem dominant
- Higher VRAM for larger models
- Consumer cards work for hobbyist ML
Older Generation Value
RTX 3060 12GB — Excellent 1080p, decent 1440p.
RTX 3070/3080 — Still capable 1440p cards.
RX 6800 XT/6900 XT — Excellent price-performance now.
Used market offers great value on previous-gen cards.
Warranty and Support
Typical warranty:
- Standard: 3 years
- Some manufacturers: 5 years
- Registration sometimes required
Reliability differences between AIB partners exist but generally minor.
Customer service varies by brand. Research before buying premium cards.
Common Buying Mistakes
Buying for future needs — Buy what you need now.
Chasing latest generation — Previous generation often better value.
Ignoring PSU adequacy — Undersized PSU limits GPU performance.
Ignoring case clearance — Modern GPUs are massive.
Ignoring aesthetics — Ugly card in showcase build annoys forever.
The Bottom Line
Understanding GPU naming helps compare cards meaningfully. NVIDIA and AMD use different conventions but similar general patterns.
Key considerations:
- Match tier to resolution
- Verify PSU adequacy
- Check case clearance
- Consider workload (gaming vs creation)
- Balance ray tracing vs price
Buying advice:
- Read multiple reviews before purchasing
- Check current pricing (fluctuates with new releases)
- Consider your monitor's resolution and refresh rate
- Don't overspec beyond what monitor can display
GPU choice affects gaming experience more than any other component. Take time to make informed decisions.
Match card to actual needs. Chasing the highest tier without workload requiring it wastes money.
The best GPU is the one adequate for your use case at a price you can justify.