Synthetic Benchmarks¶

CPU Benchmarks¶

front node: Intel Core i9-13900H CPU (see its topology in the Description section)
az4-n4090 and az4-a7900 partitions: AMD Ryzen 9 7945HX CPU (see its topology in the Description section)
iml-ia770 partition: Intel Core Ultra 9 185H CPU (see its topology in the Description section)
az5-a890m partition: AMD Ryzen AI 9 HX 370 CPU

Memory Throughput¶

Memory throughput is measured with the bandwidth benchmark and, each time, one thread is explicitly pinned to one CPU core (even if a core possesses more than one PU).

AMD Ryzen 9 7945HX¶

az4-mixed RAM is: Corsair Vengeance SO-DIMM 96 GB (2 x 48 GB, dual channel) DDR5 5200 MT/s CL44.

Intel Core i9-13900H¶

front node RAM is: Corsair Vengeance SO-DIMM 96 GB (2 x 48 GB, dual channel) DDR5 5200 MT/s CL44.

Intel Core Ultra 9 185H¶

iml-ia770 RAM is: 32 GB (2 x 16 GB, dual channel) DDR5 5600 MT/s.

AMD Ryzen AI 9 HX 370¶

az5-890m node RAM is: 4 x 8 GB, quad channel LPDDR5x 7500 MT/s.

Throughput Comparison between the different Core Types¶

In the following graphics, c, p, e and LPe mean "core", "p-core", "e-core" and "LP e-core", resp. They characterize the type of cores used in the benchmark. Just before each abbreviation, there is a number that represents the number of cores involved in the benchmark. For instance 4e means that 4 e-cores have been used.

L1d cacheL2 cacheL3 cacheRAM

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "CPU memory throughput depending on the CPU", "width": 1000, "height": 300, "data": { "values": [ {"category": "Ryzen 9 7945HX 1c", "group": "1-read", "value": 328}, {"category": "Ryzen 9 7945HX 1c", "group": "2-write", "value": 175}, {"category": "Ryzen 9 7945HX 1c", "group": "3-copy", "value": 348}, {"category": "Ryzen 9 7945HX 1c", "group": "4-scale", "value": 350}, {"category": "Ryzen 9 7945HX 1c", "group": "5-add", "value": 412}, {"category": "Ryzen 9 7945HX 1c", "group": "6-triad", "value": 465}, {"category": "Core i9-13900H 1p", "group": "1-read", "value": 277}, {"category": "Core i9-13900H 1p", "group": "2-write", "value": 243}, {"category": "Core i9-13900H 1p", "group": "3-copy", "value": 384}, {"category": "Core i9-13900H 1p", "group": "4-scale", "value": 260}, {"category": "Core i9-13900H 1p", "group": "5-add", "value": 310}, {"category": "Core i9-13900H 1p", "group": "6-triad", "value": 283}, {"category": "Core i9-13900H 1e", "group": "1-read", "value": 86}, {"category": "Core i9-13900H 1e", "group": "2-write", "value": 71}, {"category": "Core i9-13900H 1e", "group": "3-copy", "value": 160}, {"category": "Core i9-13900H 1e", "group": "4-scale", "value": 150}, {"category": "Core i9-13900H 1e", "group": "5-add", "value": 125}, {"category": "Core i9-13900H 1e", "group": "6-triad", "value": 125}, {"category": "Core Ultra 9 185H 1p", "group": "1-read", "value": 459}, {"category": "Core Ultra 9 185H 1p", "group": "2-write", "value": 307}, {"category": "Core Ultra 9 185H 1p", "group": "3-copy", "value": 517}, {"category": "Core Ultra 9 185H 1p", "group": "4-scale", "value": 338}, {"category": "Core Ultra 9 185H 1p", "group": "5-add", "value": 409}, {"category": "Core Ultra 9 185H 1p", "group": "6-triad", "value": 376}, {"category": "Core Ultra 9 185H 1e", "group": "1-read", "value": 121}, {"category": "Core Ultra 9 185H 1e", "group": "2-write", "value": 121}, {"category": "Core Ultra 9 185H 1e", "group": "3-copy", "value": 230}, {"category": "Core Ultra 9 185H 1e", "group": "4-scale", "value": 224}, {"category": "Core Ultra 9 185H 1e", "group": "5-add", "value": 182}, {"category": "Core Ultra 9 185H 1e", "group": "6-triad", "value": 177}, {"category": "Core Ultra 9 185H 1LPe", "group": "1-read", "value": 80}, {"category": "Core Ultra 9 185H 1LPe", "group": "2-write", "value": 80}, {"category": "Core Ultra 9 185H 1LPe", "group": "3-copy", "value": 152}, {"category": "Core Ultra 9 185H 1LPe", "group": "4-scale", "value": 143}, {"category": "Core Ultra 9 185H 1LPe", "group": "5-add", "value": 120}, {"category": "Core Ultra 9 185H 1LPe", "group": "6-triad", "value": 116}, {"category": "AI 9 HX 370 1p", "group": "1-read", "value": 280}, {"category": "AI 9 HX 370 1p", "group": "2-write", "value": 280}, {"category": "AI 9 HX 370 1p", "group": "3-copy", "value": 554}, {"category": "AI 9 HX 370 1p", "group": "4-scale", "value": 554}, {"category": "AI 9 HX 370 1p", "group": "5-add", "value": 420}, {"category": "AI 9 HX 370 1p", "group": "6-triad", "value": 418}, {"category": "AI 9 HX 370 1e", "group": "1-read", "value": 212}, {"category": "AI 9 HX 370 1e", "group": "2-write", "value": 212}, {"category": "AI 9 HX 370 1e", "group": "3-copy", "value": 421}, {"category": "AI 9 HX 370 1e", "group": "4-scale", "value": 410}, {"category": "AI 9 HX 370 1e", "group": "5-add", "value": 318}, {"category": "AI 9 HX 370 1e", "group": "6-triad", "value": 318} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "ubench"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

Buffer of 16 KB (higher is better).

For all the core types, the L1d cache is owned by a single core.

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "CPU memory throughput depending on the CPU", "width": 1000, "height": 300, "data": { "values": [ {"category": "Ryzen 9 7945HX 1c", "group": "1-read", "value": 175}, {"category": "Ryzen 9 7945HX 1c", "group": "2-write", "value": 173}, {"category": "Ryzen 9 7945HX 1c", "group": "3-copy", "value": 171}, {"category": "Ryzen 9 7945HX 1c", "group": "4-scale", "value": 173}, {"category": "Ryzen 9 7945HX 1c", "group": "5-add", "value": 174}, {"category": "Ryzen 9 7945HX 1c", "group": "6-triad", "value": 174}, {"category": "Core i9-13900H 1p", "group": "1-read", "value": 151}, {"category": "Core i9-13900H 1p", "group": "2-write", "value": 62}, {"category": "Core i9-13900H 1p", "group": "3-copy", "value": 100}, {"category": "Core i9-13900H 1p", "group": "4-scale", "value": 90}, {"category": "Core i9-13900H 1p", "group": "5-add", "value": 131}, {"category": "Core i9-13900H 1p", "group": "6-triad", "value": 127}, {"category": "Core i9-13900H 4e", "group": "1-read", "value": 149}, {"category": "Core i9-13900H 4e", "group": "2-write", "value": 42}, {"category": "Core i9-13900H 4e", "group": "3-copy", "value": 83}, {"category": "Core i9-13900H 4e", "group": "4-scale", "value": 81}, {"category": "Core i9-13900H 4e", "group": "5-add", "value": 115}, {"category": "Core i9-13900H 4e", "group": "6-triad", "value": 115}, {"category": "Core Ultra 9 185H 1p", "group": "1-read", "value": 201}, {"category": "Core Ultra 9 185H 1p", "group": "2-write", "value": 78}, {"category": "Core Ultra 9 185H 1p", "group": "3-copy", "value": 133}, {"category": "Core Ultra 9 185H 1p", "group": "4-scale", "value": 127}, {"category": "Core Ultra 9 185H 1p", "group": "5-add", "value": 169}, {"category": "Core Ultra 9 185H 1p", "group": "6-triad", "value": 161}, {"category": "Core Ultra 9 185H 4e", "group": "1-read", "value": 227}, {"category": "Core Ultra 9 185H 4e", "group": "2-write", "value": 84}, {"category": "Core Ultra 9 185H 4e", "group": "3-copy", "value": 149}, {"category": "Core Ultra 9 185H 4e", "group": "4-scale", "value": 149}, {"category": "Core Ultra 9 185H 4e", "group": "5-add", "value": 226}, {"category": "Core Ultra 9 185H 4e", "group": "6-triad", "value": 225}, {"category": "Core Ultra 9 185H 2LPe", "group": "1-read", "value": 119}, {"category": "Core Ultra 9 185H 2LPe", "group": "2-write", "value": 43}, {"category": "Core Ultra 9 185H 2LPe", "group": "3-copy", "value": 74}, {"category": "Core Ultra 9 185H 2LPe", "group": "4-scale", "value": 73}, {"category": "Core Ultra 9 185H 2LPe", "group": "5-add", "value": 89}, {"category": "Core Ultra 9 185H 2LPe", "group": "6-triad", "value": 88}, {"category": "AI 9 HX 370 1p", "group": "1-read", "value": 275}, {"category": "AI 9 HX 370 1p", "group": "2-write", "value": 273}, {"category": "AI 9 HX 370 1p", "group": "3-copy", "value": 275}, {"category": "AI 9 HX 370 1p", "group": "4-scale", "value": 276}, {"category": "AI 9 HX 370 1p", "group": "5-add", "value": 268}, {"category": "AI 9 HX 370 1p", "group": "6-triad", "value": 266}, {"category": "AI 9 HX 370 1e", "group": "1-read", "value": 200}, {"category": "AI 9 HX 370 1e", "group": "2-write", "value": 166}, {"category": "AI 9 HX 370 1e", "group": "3-copy", "value": 209}, {"category": "AI 9 HX 370 1e", "group": "4-scale", "value": 209}, {"category": "AI 9 HX 370 1e", "group": "5-add", "value": 203}, {"category": "AI 9 HX 370 1e", "group": "6-triad", "value": 205} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "ubench"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

Buffer of 512 KB (higher is better).

The configuration of the L2 cache depends on the core type:

Ryzen 9 7945HX cores, Core i9-13900H, Core Ultra 9 185H p-cores and Ryzen AI 9 HX 370 p-/e-cores: a L2 cache is dedicated to each core
Core i9-13900H and Core Ultra 9 185H e-cores: 4 cores are sharing the L2 cache together
Core Ultra 9 185H LP e-cores: 2 cores are sharing the L2 cache

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "CPU memory throughput depending on the CPU", "width": 1000, "height": 300, "data": { "values": [ {"category": "Ryzen 9 7945HX 8c", "group": "1-read", "value": 683}, {"category": "Ryzen 9 7945HX 8c", "group": "2-write", "value": 752}, {"category": "Ryzen 9 7945HX 8c", "group": "3-copy", "value": 780}, {"category": "Ryzen 9 7945HX 8c", "group": "4-scale", "value": 651}, {"category": "Ryzen 9 7945HX 8c", "group": "5-add", "value": 673}, {"category": "Ryzen 9 7945HX 8c", "group": "6-triad", "value": 679}, {"category": "Core i9-13900H 6p", "group": "1-read", "value": 234}, {"category": "Core i9-13900H 6p", "group": "2-write", "value": 183}, {"category": "Core i9-13900H 6p", "group": "3-copy", "value": 226}, {"category": "Core i9-13900H 6p", "group": "4-scale", "value": 205}, {"category": "Core i9-13900H 6p", "group": "5-add", "value": 238}, {"category": "Core i9-13900H 6p", "group": "6-triad", "value": 237}, {"category": "Core i9-13900H 8e", "group": "1-read", "value": 62}, {"category": "Core i9-13900H 8e", "group": "2-write", "value": 72}, {"category": "Core i9-13900H 8e", "group": "3-copy", "value": 90}, {"category": "Core i9-13900H 8e", "group": "4-scale", "value": 99}, {"category": "Core i9-13900H 8e", "group": "5-add", "value": 89}, {"category": "Core i9-13900H 8e", "group": "6-triad", "value": 81}, {"category": "Core Ultra 9 185H 6p", "group": "1-read", "value": 296}, {"category": "Core Ultra 9 185H 6p", "group": "2-write", "value": 195}, {"category": "Core Ultra 9 185H 6p", "group": "3-copy", "value": 273}, {"category": "Core Ultra 9 185H 6p", "group": "4-scale", "value": 270}, {"category": "Core Ultra 9 185H 6p", "group": "5-add", "value": 234}, {"category": "Core Ultra 9 185H 6p", "group": "6-triad", "value": 240}, {"category": "Core Ultra 9 185H 8e", "group": "1-read", "value": 100}, {"category": "Core Ultra 9 185H 8e", "group": "2-write", "value": 74}, {"category": "Core Ultra 9 185H 8e", "group": "3-copy", "value": 145}, {"category": "Core Ultra 9 185H 8e", "group": "4-scale", "value": 145}, {"category": "Core Ultra 9 185H 8e", "group": "5-add", "value": 151}, {"category": "Core Ultra 9 185H 8e", "group": "6-triad", "value": 151}, {"category": "AI 9 HX 370 4p", "group": "1-read", "value": 500}, {"category": "AI 9 HX 370 4p", "group": "2-write", "value": 450}, {"category": "AI 9 HX 370 4p", "group": "3-copy", "value": 459}, {"category": "AI 9 HX 370 4p", "group": "4-scale", "value": 457}, {"category": "AI 9 HX 370 4p", "group": "5-add", "value": 493}, {"category": "AI 9 HX 370 4p", "group": "6-triad", "value": 463}, {"category": "AI 9 HX 370 8e", "group": "1-read", "value": null}, {"category": "AI 9 HX 370 8e", "group": "2-write", "value": null}, {"category": "AI 9 HX 370 8e", "group": "3-copy", "value": null}, {"category": "AI 9 HX 370 8e", "group": "4-scale", "value": null}, {"category": "AI 9 HX 370 8e", "group": "5-add", "value": null}, {"category": "AI 9 HX 370 8e", "group": "6-triad", "value": null} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "ubench"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

Buffer of 20 MB (higher is better). Except for AI 9 HX 370 p-cores where a buffer of 10 MB is used.

The configuration of the L3 cache depends on the core type:

Ryzen 9 7945HX cores: a L3 cache is shared between 8 cores
Core i9-13900H p-cores, Core Ultra 9 185H p-cores and Core Ultra 9 185H e-cores: the L3 cache is shared for all these cores
Core Ultra 9 185H LP e-cores do not have a L3 cache
Ryzen AI 9 HX 370 p-cores: a L3 cache is shared between 4 cores
Ryzen AI 9 HX 370 e-cores: a L3 cache is shared between 8 cores

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "CPU memory throughput depending on the CPU", "width": 1000, "height": 300, "data": { "values": [ {"category": "Ryzen 9 7945HX 16c", "group": "1-read", "value": 48}, {"category": "Ryzen 9 7945HX 16c", "group": "2-write", "value": 56}, {"category": "Ryzen 9 7945HX 16c", "group": "3-copy", "value": 49}, {"category": "Ryzen 9 7945HX 16c", "group": "4-scale", "value": 48}, {"category": "Ryzen 9 7945HX 16c", "group": "5-add", "value": 47}, {"category": "Ryzen 9 7945HX 16c", "group": "6-triad", "value": 48}, {"category": "Core i9-13900H 6p", "group": "1-read", "value": 60}, {"category": "Core i9-13900H 6p", "group": "2-write", "value": 73}, {"category": "Core i9-13900H 6p", "group": "3-copy", "value": 68}, {"category": "Core i9-13900H 6p", "group": "4-scale", "value": 67}, {"category": "Core i9-13900H 6p", "group": "5-add", "value": 66}, {"category": "Core i9-13900H 6p", "group": "6-triad", "value": 66}, {"category": "Core i9-13900H 8e", "group": "1-read", "value": 35}, {"category": "Core i9-13900H 8e", "group": "2-write", "value": 73}, {"category": "Core i9-13900H 8e", "group": "3-copy", "value": 61}, {"category": "Core i9-13900H 8e", "group": "4-scale", "value": 44}, {"category": "Core i9-13900H 8e", "group": "5-add", "value": 41}, {"category": "Core i9-13900H 8e", "group": "6-triad", "value": 41}, {"category": "Core Ultra 9 185H 6p", "group": "1-read", "value": 74}, {"category": "Core Ultra 9 185H 6p", "group": "2-write", "value": 73}, {"category": "Core Ultra 9 185H 6p", "group": "3-copy", "value": 71}, {"category": "Core Ultra 9 185H 6p", "group": "4-scale", "value": 70}, {"category": "Core Ultra 9 185H 6p", "group": "5-add", "value": 72}, {"category": "Core Ultra 9 185H 6p", "group": "6-triad", "value": 72}, {"category": "Core Ultra 9 185H 8e", "group": "1-read", "value": 42}, {"category": "Core Ultra 9 185H 8e", "group": "2-write", "value": 73}, {"category": "Core Ultra 9 185H 8e", "group": "3-copy", "value": 58}, {"category": "Core Ultra 9 185H 8e", "group": "4-scale", "value": 58}, {"category": "Core Ultra 9 185H 8e", "group": "5-add", "value": 51}, {"category": "Core Ultra 9 185H 8e", "group": "6-triad", "value": 51}, {"category": "Core Ultra 9 185H 2LPe", "group": "1-read", "value": 17}, {"category": "Core Ultra 9 185H 2LPe", "group": "2-write", "value": 29}, {"category": "Core Ultra 9 185H 2LPe", "group": "3-copy", "value": 29}, {"category": "Core Ultra 9 185H 2LPe", "group": "4-scale", "value": 27}, {"category": "Core Ultra 9 185H 2LPe", "group": "5-add", "value": 25}, {"category": "Core Ultra 9 185H 2LPe", "group": "6-triad", "value": 24}, {"category": "AI 9 HX 370 4p", "group": "1-read", "value": 61}, {"category": "AI 9 HX 370 4p", "group": "2-write", "value": 98}, {"category": "AI 9 HX 370 4p", "group": "3-copy", "value": 80}, {"category": "AI 9 HX 370 4p", "group": "4-scale", "value": 80}, {"category": "AI 9 HX 370 4p", "group": "5-add", "value": 73}, {"category": "AI 9 HX 370 4p", "group": "6-triad", "value": 72}, {"category": "AI 9 HX 370 8e", "group": "1-read", "value": 61}, {"category": "AI 9 HX 370 8e", "group": "2-write", "value": 86}, {"category": "AI 9 HX 370 8e", "group": "3-copy", "value": 76}, {"category": "AI 9 HX 370 8e", "group": "4-scale", "value": 77}, {"category": "AI 9 HX 370 8e", "group": "5-add", "value": 70}, {"category": "AI 9 HX 370 8e", "group": "6-triad", "value": 70} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "ubench"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

Buffer of 2 GB (higher is better).

The RAM throughput mainly depends on the memory module types:

Ryzen 9 7945HX and Core i9-13900H cores: Corsair Vengeance SO-DIMM 96 GB (2 x 48 GB, dual channel) DDR5 5200 MT/s CL44
Core Ultra 9 185H: 32 GB (2 x 16 GB, dual channel) DDR5 5600 MT/s

Raw Data¶

Benchmark	# Cores	AMD Ryzen 9 7945HX	Intel Core i9-13900H - p-cores	Intel Core i9-13900H - e-cores	Intel Core Ultra 9 185H - p-cores	Intel Core Ultra 9 185H - e-cores	Intel Core Ultra 9 185H - LP e-cores	AMD Ryzen AI 9 HX 370 - p-cores (Zen5)	AMD Ryzen AI 9 HX 370 - e-cores (Zen5c)
`bandwidth`	1	download	download	download	download	download	download	download	download
`bandwidth`	2	download	download	download	download	download	download	download	download
`bandwidth`	3	download	download	download	download	download	-	download	download
`bandwidth`	4	download	download	download	download	download	-	download	download
`bandwidth`	5	download	download	download	download	download	-	-	download
`bandwidth`	6	download	download	download	download	download	-	-	download
`bandwidth`	7	download	-	download	-	download	-	-	download
`bandwidth`	8	download	-	download	-	download	-	-	download
`bandwidth`	9	download	-	-	-	-	-	-	-
`bandwidth`	10	download	-	-	-	-	-	-	-
`bandwidth`	11	download	-	-	-	-	-	-	-
`bandwidth`	12	download	-	-	-	-	-	-	-
`bandwidth`	13	download	-	-	-	-	-	-	-
`bandwidth`	14	download	-	-	-	-	-	-	-
`bandwidth`	15	download	-	-	-	-	-	-	-
`bandwidth`	16	download	-	-	-	-	-	-	-

Peak Performance¶

The cpufp benchmark is used. The CPU peak performance is measured according to different operations:

FMA - Fused Multiply–Add, performs the following operation: $d = a \times b + c$ on 64-bit or 32-bit floating-point numbers (reffered as f64 & f32 here)
DPA2 - Performs the dot product of two 16-bit integers (i16) and accumulates the result in a 32-bit integer (i32): $c^{i32} = c^{i32} + \sum^2_{s = 1}{ a_s^{i16} \times b_s^{i16}}$
DPA4 - Performs the dot product of four 8-bit integers (i8) and accumulates the result in a 32-bit integer (i32): $c^{i32} = c^{i32} + \sum^4_{s = 1}{ a_s^{i8} \times b_s^{i8}}$

In the following graphics, c, p, e and LPe mean "core", "p-core", "e-core" and "LP e-core", resp. They characterize the type of cores used in the benchmark. Just before each abbreviation, there is a number that represents the number of cores involved in the benchmark. For instance 4e means that 4 e-cores have been used.

Single-coreMulti-coreMulti-core Accumulated

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "Peak performance on CPU (mono-core) depending on the type of operation.", "width": 1000, "height": 300, "data": { "values": [ {"category":"Ryzen 9 7945HX 1c", "group": "1-FMA f64", "value": 87}, {"category":"Ryzen 9 7945HX 1c", "group": "2-FMA f32", "value": 174}, {"category":"Ryzen 9 7945HX 1c", "group": "3-DPA2 i32/i16", "value": 347}, {"category":"Ryzen 9 7945HX 1c", "group": "4-DPA4 i32/i8", "value": 695}, {"category": "Core i9-13900H 1p", "group": "1-FMA f64", "value": 61}, {"category": "Core i9-13900H 1p", "group": "2-FMA f32", "value": 121}, {"category": "Core i9-13900H 1p", "group": "3-DPA2 i32/i16", "value": 242}, {"category": "Core i9-13900H 1p", "group": "4-DPA4 i32/i8", "value": 484}, {"category": "Core i9-13900H 1e", "group": "1-FMA f64", "value": 22}, {"category": "Core i9-13900H 1e", "group": "2-FMA f32", "value": 43}, {"category": "Core i9-13900H 1e", "group": "3-DPA2 i32/i16", "value": 43}, {"category": "Core i9-13900H 1e", "group": "4-DPA4 i32/i8", "value": 86}, {"category":"Ultra 9 185H 1p", "group": "1-FMA f64", "value": 76}, {"category":"Ultra 9 185H 1p", "group": "2-FMA f32", "value": 153}, {"category":"Ultra 9 185H 1p", "group": "3-DPA2 i32/i16", "value": 305}, {"category":"Ultra 9 185H 1p", "group": "4-DPA4 i32/i8", "value": 611}, {"category":"Ultra 9 185H 1e", "group": "1-FMA f64", "value": 30}, {"category":"Ultra 9 185H 1e", "group": "2-FMA f32", "value": 60}, {"category":"Ultra 9 185H 1e", "group": "3-DPA2 i32/i16", "value": 121}, {"category":"Ultra 9 185H 1e", "group": "4-DPA4 i32/i8", "value": 240}, {"category":"Ultra 9 185H 1LPe", "group": "1-FMA f64", "value": 18}, {"category":"Ultra 9 185H 1LPe", "group": "2-FMA f32", "value": 35}, {"category":"Ultra 9 185H 1LPe", "group": "3-DPA2 i32/i16", "value": 70}, {"category":"Ultra 9 185H 1LPe", "group": "4-DPA4 i32/i8", "value": 139}, {"category":"AI 9 HX 370 1p", "group": "1-FMA f64", "value": 70}, {"category":"AI 9 HX 370 1p", "group": "2-FMA f32", "value": 140}, {"category":"AI 9 HX 370 1p", "group": "3-DPA2 i32/i16", "value": 280}, {"category":"AI 9 HX 370 1p", "group": "4-DPA4 i32/i8", "value": 559}, {"category":"AI 9 HX 370 1e", "group": "1-FMA f64", "value": 53}, {"category":"AI 9 HX 370 1e", "group": "2-FMA f32", "value": 106}, {"category":"AI 9 HX 370 1e", "group": "3-DPA2 i32/i16", "value": 211}, {"category":"AI 9 HX 370 1e", "group": "4-DPA4 i32/i8", "value": 422} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Performance (Gop/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "Operation"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

CPU single-core peak performance depending on the type of operation (higher is better).

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "Peak performance on CPU (multi-core) depending on the type of operation.", "width": 1000, "height": 300, "data": { "values": [ {"category":"Ryzen 9 7945HX 16c", "group": "1-FMA f64", "value": 1281}, {"category":"Ryzen 9 7945HX 16c", "group": "2-FMA f32", "value": 2574}, {"category":"Ryzen 9 7945HX 16c", "group": "3-DPA2 i32/i16", "value": 5176}, {"category":"Ryzen 9 7945HX 16c", "group": "4-DPA4 i32/i8", "value": 10350}, {"category": "Core i9-13900H 6p", "group": "1-FMA f64", "value": 363}, {"category": "Core i9-13900H 6p", "group": "2-FMA f32", "value": 726}, {"category": "Core i9-13900H 6p", "group": "3-DPA2 i32/i16", "value": 1453}, {"category": "Core i9-13900H 6p", "group": "4-DPA4 i32/i8", "value": 2906}, {"category": "Core i9-13900H 8e", "group": "1-FMA f64", "value": 172}, {"category": "Core i9-13900H 8e", "group": "2-FMA f32", "value": 344}, {"category": "Core i9-13900H 8e", "group": "3-DPA2 i32/i16", "value": 344}, {"category": "Core i9-13900H 8e", "group": "4-DPA4 i32/i8", "value": 688}, {"category":"Ultra 9 185H 6p", "group": "1-FMA f64", "value": 431}, {"category":"Ultra 9 185H 6p", "group": "2-FMA f32", "value": 860}, {"category":"Ultra 9 185H 6p", "group": "3-DPA2 i32/i16", "value": 1724}, {"category":"Ultra 9 185H 6p", "group": "4-DPA4 i32/i8", "value": 3452}, {"category":"Ultra 9 185H 8e", "group": "1-FMA f64", "value": 210}, {"category":"Ultra 9 185H 8e", "group": "2-FMA f32", "value": 421}, {"category":"Ultra 9 185H 8e", "group": "3-DPA2 i32/i16", "value": 836}, {"category":"Ultra 9 185H 8e", "group": "4-DPA4 i32/i8", "value": 1682}, {"category":"Ultra 9 185H 2LPe", "group": "1-FMA f64", "value": 37}, {"category":"Ultra 9 185H 2LPe", "group": "2-FMA f32", "value": 75}, {"category":"Ultra 9 185H 2LPe", "group": "3-DPA2 i32/i16", "value": 149}, {"category":"Ultra 9 185H 2LPe", "group": "4-DPA4 i32/i8", "value": 298}, {"category":"AI 9 HX 370 4p", "group": "1-FMA f64", "value": 279}, {"category":"AI 9 HX 370 4p", "group": "2-FMA f32", "value": 559}, {"category":"AI 9 HX 370 4p", "group": "3-DPA2 i32/i16", "value": 1112}, {"category":"AI 9 HX 370 4p", "group": "4-DPA4 i32/i8", "value": 2236}, {"category":"AI 9 HX 370 8e", "group": "1-FMA f64", "value": 422}, {"category":"AI 9 HX 370 8e", "group": "2-FMA f32", "value": 844}, {"category":"AI 9 HX 370 8e", "group": "3-DPA2 i32/i16", "value": 1688}, {"category":"AI 9 HX 370 8e", "group": "4-DPA4 i32/i8", "value": 3376} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Performance (Gop/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "Operation"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

CPU multi-core peak performance depending on the type of operation (higher is better).

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "Peak performance on CPU (multi-core) depending on the type of operation.", "width": 1000, "height": 300, "data": { "values": [ {"category":"Ryzen 9 7945HX 16c", "group": "1-FMA f64", "value": 1281}, {"category":"Ryzen 9 7945HX 16c", "group": "2-FMA f32", "value": 2574}, {"category":"Ryzen 9 7945HX 16c", "group": "3-DPA2 i32/i16", "value": 5176}, {"category":"Ryzen 9 7945HX 16c", "group": "4-DPA4 i32/i8", "value": 10350}, {"category": "Core i9-13900H 6p+8e", "group": "1-FMA f64", "value": 535}, {"category": "Core i9-13900H 6p+8e", "group": "2-FMA f32", "value": 1070}, {"category": "Core i9-13900H 6p+8e", "group": "3-DPA2 i32/i16", "value": 1797}, {"category": "Core i9-13900H 6p+8e", "group": "4-DPA4 i32/i8", "value": 3594}, {"category":"Ultra 9 185H 6p+8e+2LPe", "group": "1-FMA f64", "value": 678}, {"category":"Ultra 9 185H 6p+8e+2LPe", "group": "2-FMA f32", "value": 1356}, {"category":"Ultra 9 185H 6p+8e+2LPe", "group": "3-DPA2 i32/i16", "value": 2709}, {"category":"Ultra 9 185H 6p+8e+2LPe", "group": "4-DPA4 i32/i8", "value": 5432}, {"category":"AI 9 HX 370 4p+8e", "group": "1-FMA f64", "value": 701}, {"category":"AI 9 HX 370 4p+8e", "group": "2-FMA f32", "value": 1403}, {"category":"AI 9 HX 370 4p+8e", "group": "3-DPA2 i32/i16", "value": 2800}, {"category":"AI 9 HX 370 4p+8e", "group": "4-DPA4 i32/i8", "value": 5612} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "CPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Performance (Gop/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "Operation"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

CPU multi-core peak performance depending on the type of operation (higher is better).

Raw Data¶

Benchmark	AMD Ryzen 9 7945HX	Intel Core i9-13900H - p-cores	Intel Core i9-13900H - e-cores	Intel Core Ultra 9 185H - p-cores	Intel Core Ultra 9 185H - e-cores	Intel Core Ultra 9 185H - LP e-cores
`cpufp`	download	download	download	download	download	download

GPU Benchmarks¶

front node: Intel Iris Xe iGPU
az4-n4090 partition: AMD Radeon 610M iGPU and Nvidia GeForce RTX 4090 dGPU
az4-a7900 partition: AMD Radeon 610M iGPU and AMD Radeon RX 7900 XTX dGPU
iml-ia770 partition: Intel Arc Mobile iGPU and Intel Arc 770 eGPU
az5-a890m partition: AMD Radeon 890M iGPU

Memory Throughput¶

Measurement of the memory throughput between the GPU and its global memory (VRAM) with the clpeak benchmark. When it is a iGPU, its global memory is the RAM and it is shared with the CPU.

class="vegalite">{ hema/vega-lite/v5.json", .", "group": "1-float32x1", "value": 38}, "group": "2-float32x2", "value": 48}, "group": "3-float32x4", "value": 49}, "group": "4-float32x8", "value": 43}, "group": "5-float32x16", "value": 36}, "group": "1-float32x1", "value": 72}, "group": "2-float32x2", "value": 74}, "group": "3-float32x4", "value": 75}, "group": "4-float32x8", "value": 74}, "group": "5-float32x16", "value": 72}, "group": "1-float32x1", "value": 65}, "group": "2-float32x2", "value": 71}, "group": "3-float32x4", "value": 69}, "group": "4-float32x8", "value": 72}, "group": "5-float32x16", "value": 56}, "group": "1-float32x1", "value": 89}, "group": "2-float32x2", "value": 95}, "group": "3-float32x4", "value": 96}, "group": "4-float32x8", "value": 93}, "group": "5-float32x16", "value": 96}, "group": "1-float32x1", "value": 399}, "group": "2-float32x2", "value": 404}, "group": "3-float32x4", "value": 407}, "group": "4-float32x8", "value": 411}, "group": "5-float32x16", "value": 417}, "group": "1-float32x1", "value": 708}, "group": "2-float32x2", "value": 751}, "group": "3-float32x4", "value": 757}, "group": "4-float32x8", "value": 811}, "group": "5-float32x16", "value": 857}, "group": "1-float32x1", "value": 895}, "group": "2-float32x2", "value": 924}, "group": "3-float32x4", "value": 941}, "group": "4-float32x8", "value": 952}, "group": "5-float32x16", "value": 963}, 4090 (PoCL)", "group": "1-float32x1", "value": 894}, 4090 (PoCL)", "group": "2-float32x2", "value": 922}, 4090 (PoCL)", "group": "3-float32x4", "value": 941}, 4090 (PoCL)", "group": "4-float32x8", "value": 951}, 4090 (PoCL)", "group": "5-float32x16", "value": 961} "type": "nominal", "sort": "none", "axis": {"title": "GPU", "labelAngle": -0}}, "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "sort": "none"}, "title": "Datatype"} "type": "quantitative", "sort": "none"}

GPU memory throughput (higher is better).

Peak Performance¶

Measurement of the GPU peak performance. The clpeak benchmark is used. It is an OpenCL benchmark that executes a compute intensive program to estimate peak performance.

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "GPU peak performance (32-bit float).", "width": 1000, "height": 300, "data": { "values": [ {"category": "Radeon 610M (iGPU)", "group": "1-float16", "value": 1.1}, {"category": "Radeon 610M (iGPU)", "group": "2-float32", "value": 0.5}, {"category": "Radeon 610M (iGPU)", "group": "3-float64", "value": 0.3}, {"category": "Radeon 610M (iGPU)", "group": "4-int8", "value": 0.5}, {"category": "Radeon 610M (iGPU)", "group": "5-int16", "value": 0.5}, {"category": "Radeon 610M (iGPU)", "group": "6-int32", "value": 0.1}, {"category": "Iris Xe (iGPU)", "group": "1-float16", "value": 4.4}, {"category": "Iris Xe (iGPU)", "group": "2-float32", "value": 2.2}, {"category": "Iris Xe (iGPU)", "group": "3-float64", "value": null}, {"category": "Iris Xe (iGPU)", "group": "4-int8", "value": 1.5}, {"category": "Iris Xe (iGPU)", "group": "5-int16", "value": 4.2}, {"category": "Iris Xe (iGPU)", "group": "6-int32", "value": 0.8}, {"category": "Arc Mobile (iGPU)", "group": "1-float16", "value": 9.8}, {"category": "Arc Mobile (iGPU)", "group": "2-float32", "value": 4.8}, {"category": "Arc Mobile (iGPU)", "group": "3-float64", "value": 0.1}, {"category": "Arc Mobile (iGPU)", "group": "4-int8", "value": 2.9}, {"category": "Arc Mobile (iGPU)", "group": "5-int16", "value": 7.5}, {"category": "Arc Mobile (iGPU)", "group": "6-int32", "value": 1.3}, {"category": "Radeon 890M (iGPU)", "group": "1-float16", "value": 7.3}, {"category": "Radeon 890M (iGPU)", "group": "2-float32", "value": 4.3}, {"category": "Radeon 890M (iGPU)", "group": "3-float64", "value": 0.2}, {"category": "Radeon 890M (iGPU)", "group": "4-int8", "value": 3.2}, {"category": "Radeon 890M (iGPU)", "group": "5-int16", "value": 3.4}, {"category": "Radeon 890M (iGPU)", "group": "6-int32", "value": 1.0}, {"category": "Arc 770 (eGPU)", "group": "1-float16", "value": 19.5}, {"category": "Arc 770 (eGPU)", "group": "2-float32", "value": 13.0}, {"category": "Arc 770 (eGPU)", "group": "3-float64", "value": null}, {"category": "Arc 770 (eGPU)", "group": "4-int8", "value": 11.4}, {"category": "Arc 770 (eGPU)", "group": "5-int16", "value": 18.0}, {"category": "Arc 770 (eGPU)", "group": "6-int32", "value": 5.5}, {"category": "Radeon RX 7900 XTX", "group": "1-float16", "value": 63.9}, {"category": "Radeon RX 7900 XTX", "group": "2-float32", "value": 33.8}, {"category": "Radeon RX 7900 XTX", "group": "3-float64", "value": 1.1}, {"category": "Radeon RX 7900 XTX", "group": "4-int8", "value": 31.1}, {"category": "Radeon RX 7900 XTX", "group": "5-int16", "value": 31.1}, {"category": "Radeon RX 7900 XTX", "group": "6-int32", "value": 6.7}, {"category": "GeForce RTX 4090", "group": "1-float16", "value": null}, {"category": "GeForce RTX 4090", "group": "2-float32", "value": 81.4}, {"category": "GeForce RTX 4090", "group": "3-float64", "value": 1.4}, {"category": "GeForce RTX 4090", "group": "4-int8", "value": 38.9}, {"category": "GeForce RTX 4090", "group": "5-int16", "value": 37.8}, {"category": "GeForce RTX 4090", "group": "6-int32", "value": 44.9}, {"category": "GeForce RTX 4090 (PoCL)", "group": "1-float16", "value": 89.4}, {"category": "GeForce RTX 4090 (PoCL)", "group": "2-float32", "value": 80.8}, {"category": "GeForce RTX 4090 (PoCL)", "group": "3-float64", "value": 1.4}, {"category": "GeForce RTX 4090 (PoCL)", "group": "4-int8", "value": 21.2}, {"category": "GeForce RTX 4090 (PoCL)", "group": "5-int16", "value": 22.1}, {"category": "GeForce RTX 4090 (PoCL)", "group": "6-int32", "value": 30.8} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "GPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Peak Performance (Top/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "Datatype"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

GPU peak performance (higher is better).

Warning

The Nvidia GeForce RTX 4090 GPU supports float16 format in hardware but this is not supported by the OpenCL Nvidia driver. This is why the PoCL driver has also been installed.

Kernel Launch Latency¶

The kernel launch latency is the duration between the time of the order to execute a kernel from the CPU user-space and the time of its beginning of execution on the GPU (excluding data buffers memory transfers). The clpeak benchmark is used.

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "Kernel launch latency.", "width": 1000, "height": 300, "data": { "values": [ {"category": "Radeon 610M (iGPU)", "group": "OpenCL", "value": null}, {"category": "Iris Xe (iGPU)", "group": "OpenCL", "value": 34.4}, {"category": "Arc Mobile (iGPU)", "group": "OpenCL", "value": 39.5}, {"category": "Radeon 890M (iGPU)", "group": "OpenCL", "value": 5.8}, {"category": "Arc 770 (eGPU)", "group": "OpenCL", "value": 4.4}, {"category": "Radeon RX 7900 XTX", "group": "OpenCL", "value": null}, {"category": "GeForce RTX 4090", "group": "OpenCL", "value": 4.7} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "GPU", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Kernel latency (us)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "API"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

GPU kernel launch latency (lower is better).

In the previous histogram we did not reported the latency of the PoCL GeForce RTX 4090 because it is abnormally super high (3.6 ms).

Warning

Some OpenCL implementations do not report the correct kernel launch latency and this is why the Radeon 610M and the Radeon RX 7900 XTX do no appear in the previous histogram.

Raw Data¶

Benchmark	AMD Radeon 610M	Intel Iris Xe	Intel Arc Mobile	AMD Radeon 890M	Intel Arc 770	AMD Radeon RX 7900 XTX	Nvidia GeForce RTX 4090
`clpeak`	download	download	download [Xe][i915]	download	download [Xe][i915]	download	download

SSD Benchmarks¶

front node: 3x Samsung 990 PRO 4 TB SSD
az4-n4090 partition: 1x Samsung 990 PRO 4 TB SSD
az4-a7900 partition: 1x Samsung 990 PRO 2 TB SSD
iml-ia770 partition: 1x Kingston OM8PGP41024Q-A0 1 TB SSD
az5-a890m partition: 1x Crucial P3 Plus CT1000P3PSSD8 1 TB SSD

{ "$schema": "https://vega.github.io/schema/vega-lite/v5.json", "description": "SSD throughput.", "width": 1000, "height": 300, "data": { "values": [ {"category": "Samsung 990 PRO 4 TB (front)", "group": "1-seq-read", "value": 3.5}, {"category": "Samsung 990 PRO 4 TB (front)", "group": "2-seq-write", "value": 2.5}, {"category": "Samsung 990 PRO 4 TB (front)", "group": "3-rnd-read", "value": 1.2}, {"category": "Samsung 990 PRO 4 TB (front)", "group": "4-rnd-write", "value": 0.9}, {"category": "Samsung 990 PRO 4 TB (az4-n4090)", "group": "1-seq-read", "value": 3.5}, {"category": "Samsung 990 PRO 4 TB (az4-n4090)", "group": "2-seq-write", "value": 1.9}, {"category": "Samsung 990 PRO 4 TB (az4-n4090)", "group": "3-rnd-read", "value": 1.2}, {"category": "Samsung 990 PRO 4 TB (az4-n4090)", "group": "4-rnd-write", "value": 0.7}, {"category": "Samsung 990 PRO 2 TB", "group": "1-seq-read", "value": 3.5}, {"category": "Samsung 990 PRO 2 TB", "group": "2-seq-write", "value": 1.8}, {"category": "Samsung 990 PRO 2 TB", "group": "3-rnd-read", "value": 1.2}, {"category": "Samsung 990 PRO 2 TB", "group": "4-rnd-write", "value": 0.7}, {"category": "Kingston OM8PGP41024Q-A0 1 TB", "group": "1-seq-read", "value": 3.7}, {"category": "Kingston OM8PGP41024Q-A0 1 TB", "group": "2-seq-write", "value": 3.4}, {"category": "Kingston OM8PGP41024Q-A0 1 TB", "group": "3-rnd-read", "value": 1.2}, {"category": "Kingston OM8PGP41024Q-A0 1 TB", "group": "4-rnd-write", "value": 0.8}, {"category": "Crucial P3 Plus CT1000P3PSSD8 1 TB", "group": "1-seq-read", "value": 4.8}, {"category": "Crucial P3 Plus CT1000P3PSSD8 1 TB", "group": "2-seq-write", "value": 2.1}, {"category": "Crucial P3 Plus CT1000P3PSSD8 1 TB", "group": "3-rnd-read", "value": 1.3}, {"category": "Crucial P3 Plus CT1000P3PSSD8 1 TB", "group": "4-rnd-write", "value": 0.7} ] }, "mark": "bar", "encoding": { "x": {"field": "category", "type": "nominal", "sort": "none", "axis": {"title": "SSD", "labelAngle": -0}}, "y": {"field": "value", "type": "quantitative", "axis": {"title": "Throughput (GB/s)"}, "scale": {"type": "linear"}}, "xOffset": {"field": "group", "sort": "none"}, "color": {"field": "group", "title": "Operation"} }, "layer": [{ "mark": "bar" }, { "mark": { "type": "text", "align": "center", "baseline": "middle", "dy": -10 }, "encoding": { "text": {"field": "value", "type": "quantitative", "sort": "none"} } }] }

SSD throughput (higher is better).

Raw Data¶

Benchmark	Samsung 990 PRO 4 TB (`front`)	Samsung 990 PRO 4 TB (`az4-n4090`)	Samsung 990 PRO 2 TB	Kingston OM8PGP41024Q-A0 1 TB	Crucial P3 Plus CT1000P3PSSD8 1 TB
`dd`	download	download	download	download	download
`iozone`	download	download	download	download	download