9-24-2024 Results will be updated as they are tested.
Objective
This test examines the thermal performance of an Asus RX 5700 XT ROG Strix OC at three different power levels: 220W, 260W, and 300W. The focus is on the impact of die size and thermal compound on heat dissipation.
Test Setup
The GPU’s die measures 251mm², and different power levels were tested to observe temperature variations. Premium thermal compounds were used for optimal thermal conductivity.
Results:
- 220W Test: PTM 7950 leads at 108.4c combined GPU and Hotspot over ambient.
- 260W Test: Thermalright TF-7 just barely edged out PTM 7950 as the leader with 116.2c combined GPU and Hotspot over ambient.
- 300W Test: At this power level, the GPU operated at near its thermal limit, with temperatures peaking over 100c. Unexpectedly Noctua’s older formula NT-H1 leads with 127.8c combined GPU and Hotspot over ambient.
Conclusion
The results show a clear relationship between power consumption and heat generation, with higher wattage levels causing significant temperature increases especially with a smaller die size. Premium thermal compounds helped mitigate overheating, but the die size limited overall thermal performance. We will continue to add more thermal interfaces to this list and introduce more testing scenarios as time becomes available.
Test Card: Asus RX 5700 XT ROG Strix OC – 251mm sq die size
Thermal Density and Test Card The die size plays a crucial role in determining the thermal load. According to a Steam chart of the top 50 GPUs, the 5700 XT ranks among the worst for heat density. This makes it an ideal test card, as it can operate between 110W and 330W. However, I’ve never seen it exceed 320W in testing.
GPU Die mm-sq TDP Density w/mm-sq
RTX 2060 445 160 0.359550562
GTX 1650 200 75 0.375
RTX 2060 SUPER 445 175 0.393258427
RTX 2070 445 175 0.393258427
RTX 2070 SUPER 545 215 0.394495413
RTX 2080 545 215 0.394495413
GTX 1660 284 120 0.422535211
GTX 1660 Ti 284 120 0.422535211
ARC a580 406 175 0.431034483
GTX 1660 SUPER 284 125 0.440140845
RTX 2080 SUPER 545 250 0.458715596
R9 FURY 596 275 0.461409396
RTX 3050 276 130 0.471014493
GTX 1070 314 150 0.477707006
RX 7900 GRE 529 260 0.491493384
RTX 3080 628 320 0.50955414
RTX 3060 Ti 392 200 0.510204082
GTX 1080 Ti 471 250 0.530785563
ARC a750 406 225 0.554187192
ARC a770 406 225 0.554187192
RX 6600 237 132 0.556962025
RTX 3080 Ti 628 350 0.557324841
RTX 3090 628 350 0.557324841
RTX 3070 392 220 0.56122449
RX 7900 XT 529 300 0.56710775
GTX 1050 132 75 0.568181818
GTX 1050 Ti 132 75 0.568181818
GTX 1080 314 180 0.573248408
GTX 1060 200 120 0.6
RTX 3060 276 170 0.615942029
RX 570 232 150 0.646551724
GTX Titan Z 561 375 0.668449198
RX 7900 XTX 529 355 0.671077505
RTX 4070 294 200 0.680272109
RX 6700 XT 335 230 0.686567164
RX 7700 XT 346 245 0.708092486
RTX 4060 159 115 0.72327044
RTX 4090 609 450 0.738916256
RTX 3070 Ti 392 290 0.739795918
RTX 4070 SUPER 294 220 0.74829932
RTX 4070 Ti SUPER 379 285 0.751978892
RX 7800 XT 346 263 0.760115607
RX 580 232 185 0.797413793
RX 7600 204 165 0.808823529
RTX 4080 379 320 0.844327177
RTX 4080 SUPER 379 320 0.844327177
RTX 4060 Ti 188 160 0.85106383
RX 5700 XT 251 225 0.896414343
RX 7600 XT 204 190 0.931372549
RTX 4070 Ti 294 285 0.969387755
Titan Xp 471 250 0.530785563
Vega 64 495 295 0.595959596
NOTE: *Fake PTM* I say fake because I ordered it on Aliexpress and have no way to validate its authentic.
NOTE: I received the Kooling Monster KOLD-01 thermal paste as a free sample but initially got poor results. After contacting the company, I discovered the issue stemmed from inconsistent clamping pressure on the GPU. Outside of this one sample, all other thermal compounds were purchased by me at retail prices.
For testing, I’m comparing the highest GPU and hotspot temperatures after 10 minutes at designated wattages. I still have five pastes from my collection to test and four more from Aliexpress, plus Halnziye products arriving soon. I’ve had to discard some early results due to clamping pressure issues, but after adding nylon washers, the problem was fully resolved. I also tried testing liquid metal solutions, but their results were inconsistent. I’ll continue testing more compounds and keep this page updated with new findings
Future Testing Plans I have upcoming tools that will allow for more scientific and consistent testing of thermal compounds. Once they arrive, I’ll create a new testing series focusing on different die sizes and processors, as these factors can significantly change the thermal rankings. This new method will also let me evaluate heat loads across various surface areas more efficiently.
220 Watt Results
Name - GPU - Hotspot - GPU+Hotspot Over Ambient
GD900 71 89 117.6
Sub Zero C14 71 88 117.2
Arctic MX-4 70 88 117
be quiet! DC2 69 88 115.4
Gelid GC-Extreme 70 88 114.8
Arctic MX-6 69 87 114.2
Nuomi TP-133 70 86 113.6
Thermalright TF-7 69 85 113.2
Kooling Monster KOLD-01 69 86 112.6
Iceberg Fuzeice 70 85 112.2
Noctua NT-H2 68 84 111.6
Noctua NT-H1 68 84 111.4
Arctic Silver 5 69 85 111.2
TG Kryonaut 69 84 110.4
Bitspower TP-1 69 84 110.2
TG Hydronaut 68 84 108.6
Fake PTM 7950 68 83 108.4
260 Watt Results
Name - GPU - Hotspot - GPU+Hotspot Over Ambient
Gelid GC-Extreme 73 95 126
GD900 71 96 125
Sub Zero C14 71 94 122.2
Arctic MX-6 71 93 121.6
Arctic Silver 5 70 94 121.6
Arctic MX-4 71 93 121.4
be quiet! DC2 70 93 120.6
Noctua NT-H2 72 91 120.6
Nuomi TP-133 71 92 120
Kooling Monster KOLD-01 70 92 119.8
Bitspower TP-1 71 91 119.8
Iceberg Fuzeice 70 92 119.6
TG Hydronaut 71 91 119.2
Noctua NT-H1 71 90 119
TG Kryonaut 70 91 118.8
Fake PTM 7950 70 89 116.6
Thermalright TF-7 69 90 116.2
300 Watt Results
Name - GPU - Hotspot - GPU+Hotspot Over Ambient
Gelid GC-Extreme 76 105 139.2
GD900 75 106 138.8
Sub Zero C14 76 103 136.4
Arctic Silver 5 74 103 134.8
Iceberg Fuzeice 75 101 133.6
be quiet! DC2 73 101 131.6
Kooling Monster KOLD-01 74 100 131.6
TG Hydronaut 75 99 131.2
Bitspower TP-1 74 99 130.6
Arctic MX-6 73 100 130.6
Arctic MX-4 74 99 130.6
TG Kryonaut 74 99 130.6
Noctua NT-H2 75 98 130.6
Fake PTM 7950 74 99 130.4
Thermalright TF-7 74 98 129.6
Nuomi TP-133 73 99 129
Noctua NT-H1 73 97 127.8