NVIDIA Grace

EvoBench v3.3 · 24 Livermore kernels · Waverian Evolution benchmark

Version	3.3.0
Date	08:32 02/05/2026
Compiler	gcc (Debian/Linaro 4.9.2-10+deb8u1) 4.9.2
Logical cores	72
CPU name	NVIDIA Grace
Comment	1 x NVIDIA Grace Neoverse V2 / 72 cores / 72 threads @ 3.1 GHz
Clockrate	3,350 MHz
TDP	200 W
Normalized	ST/GHz: 1,273 · MT/Watt: 1,555

3,991

Overall score

604 - Singlecore	43,607 MP ratio x72.2 Multicore
2,414 MP ratio x4.0 Quadcore	44,150 MP ratio x73.1 Workstation

Non optimized	Single core	Multi core	Quad core	Workstation
Maximum	2,120	154,226	8,463	155,195
Average	693	50,067	2,769	50,671
Geometric (= official score)	604	43,607	2,414	44,150
Harmonic	536	38,703	2,144	39,184
Minimum	275	19,801	1,098	20,103

#	Kernel name
1.	Hydro fragment	1,126	81,438	4,503	82,452
2.	ICCG excerpt (Incomplete Cholesky Conjugate Gradient)	859	62,206	3,429	62,909
3.	Inner product	517	37,262	2,065	37,746
4.	Banded linear equations	631	45,686	2,525	46,101
5.	Tri-diagonal elimination, below diagonal	343	24,649	1,368	24,903
6.	General linear recurrence equations	439	31,533	1,750	31,938
7.	Equation of state fragment	1,252	90,197	5,000	91,502
8.	ADI integration	821	59,066	3,272	59,918
9.	Integrate predictors	988	71,351	3,945	72,286
10.	Difference predictors	368	26,486	1,469	26,887
11.	First sum	297	21,371	1,186	21,657
12.	First difference	445	32,834	1,817	33,286
13.	2-D PIC (Particle In Cell)	275	19,801	1,098	20,103
14.	1-D PIC (Particle In Cell)	404	29,112	1,613	29,548
15.	Casual Fortran. Development version	878	63,219	3,506	64,225
16.	Monte Carlo search loop	642	47,018	2,601	47,642
17.	Implicit, conditional computation	607	43,835	2,431	44,515
18.	2-D explicit hydrodynamics fragment	590	42,517	2,356	43,156
19.	General linear recurrence equations	288	20,702	1,149	20,812
20.	Discrete ordinates transport, conditional recurrence on xx	1,121	80,532	4,469	81,526
21.	Matrix*matrix product	511	36,903	2,036	37,241
22.	Planckian distribution	2,120	154,226	8,463	155,195
23.	2-D implicit hydrodynamics fragment	624	45,157	2,492	45,662
24.	Find location of first minimum in array	479	34,505	1,904	34,893

28,353

Optimized ratio x7.1

Optimized score

4,265 - Singlecore	311,082 MP ratio x72.9 Multicore
17,179 MP ratio x4.0 Quadcore	311,802 MP ratio x73.1 Workstation

Optimized	Single core	Multi core	Quad core	Workstation
Maximum	18,892	1,357,001	75,927	1,352,887
Average	5,692	415,267	22,988	415,821
Geometric (= official score)	4,265	311,082	17,179	311,802
Harmonic	3,359	245,188	13,522	246,000
Minimum	1,323	95,831	5,313	96,514

#	Kernel name
1.	Hydro fragment	18,892	1,357,001	75,927	1,352,887
2.	ICCG excerpt (Incomplete Cholesky Conjugate Gradient)	4,864	354,187	19,716	353,688
3.	Inner product	4,432	313,324	17,415	316,033
4.	Banded linear equations	4,445	317,071	17,715	316,936
5.	Tri-diagonal elimination, below diagonal	1,730	126,080	6,949	126,849
6.	General linear recurrence equations	2,498	181,981	10,051	183,278
7.	Equation of state fragment	16,127	1,176,009	64,785	1,177,271
8.	ADI integration	12,706	920,007	51,038	917,746
9.	Integrate predictors	11,940	916,013	51,277	924,945
10.	Difference predictors	3,686	269,813	14,788	269,258
11.	First sum	2,031	147,031	8,101	147,316
12.	First difference	4,773	345,750	19,164	347,086
13.	2-D PIC (Particle In Cell)	2,086	152,422	8,379	152,606
14.	1-D PIC (Particle In Cell)	2,715	198,684	10,911	198,811
15.	Casual Fortran. Development version	6,172	451,191	24,786	451,720
16.	Monte Carlo search loop	5,332	384,027	21,214	384,503
17.	Implicit, conditional computation	2,995	219,209	12,029	219,707
18.	2-D explicit hydrodynamics fragment	11,454	837,840	46,003	837,080
19.	General linear recurrence equations	1,602	117,062	6,435	117,730
20.	Discrete ordinates transport, conditional recurrence on xx	1,323	95,831	5,313	96,514
21.	Matrix*matrix product	5,353	389,736	21,482	390,392
22.	Planckian distribution	3,873	283,096	15,557	283,096
23.	2-D implicit hydrodynamics fragment	3,755	274,933	15,083	275,570
24.	Find location of first minimum in array	1,818	138,112	7,589	138,679