Unity Compute Node Specifications

Some details such as core count, memory, GPU presence, and shared/exclusive status for each compute node in Unity.

To see a more up to date version of this information you can run the following command on Unity:

	NODELIST	CPUS	MEMORY	PARTITION	GRES	AVAIL_FEATURES
	u001	24	515000	batch, bigmem, bond	(null)	10Gb,cpu,avx2,bigmem,broadwell,scom,rhel8
	u002	48	1540000	mib	(null)	10Gb,cpu,avx2,haswell,mib,rhel8
	u003	40	1030000	mib	(null)	10Gb,cpu,sse4_2,westmere,mib,rhel8
	u004	48	1030000	sullivan	(null)	10Gb,cpu,avx2,haswell,sullivan,rhel8
	u007	40	128000	white-1gpu	gpu:p100:1(S:0)	10Gb,avx2,broadwell,white,gpu,p100,p100-16,rhel8
	u009	40	128000	batch, batch-gpu	gpu:p100:1(S:0)	10Gb,avx2,broadwell,psy,sederberg,gpu,p100,p100-16,rhel8
	u010	44	256000	batch	(null)	10Gb,cpu,avx2,broadwell,asc,shared,rhel8
	u011	20	128000	batch, batch-gpu	gpu:p100:1(S:0)	10Gb,avx2,broadwell,asc,gpu,p100,p100-16,shared,rhel8
	u012-u015	16	257000	trivedi, trivedi-broadwell	(null)	10Gb,cpu,ib,avx2,broadwell,trivedi,rhel8
	u029	28	64214	batch, batch-gpu	gpu:p100:1(S:0)	10Gb,avx2,gpu,p100,p100-12,asc,shared,rhel8
	u030	16	95920	batch7	(null)	1Gb,cpu,ib,avx512,skylake,asc,shared,rhel7
	u031-u033	28	94000	batch	(null)	1Gb,cpu,ib,avx512,skylake,asc,shared,rhel8
	u034	56	94000	batch, batch-gpu	gpu:v100:1(S:0)	10Gb,ib,avx512,skylake,gpu,v100,v100-16,asc,shared,rhel8
	u042-u047	40	191000	stat, stat-skylake	(null)	10Gb,cpu,ib,avx512,skylake,c18,stat,rhel8
	u056-u059	16	94000	randeria	gpu:v100:1(S:0)	10Gb,gpu,avx512,skylake,v100,v100-16,mohit,randeria,rhel8
	u086-u089	40	191000	batch, stat, stat-cascade	(null)	10Gb,cpu,ib,avx512,cascade,stat,rhel8
	u096-u098	48	385000	white-2gpu	gpu:v100:2(S:0-1)	10Gb,gpu,avx512,cascade,v100,v100-32,white,rhel8
	u126	52	372000	moortgat, moortgat-gpu, moortgat-a40-gpu	gpu:a40:4(S:0-1)	10Gb,gpu,ib,avx512,cascade,a40,moortgat,rhel8
	u154-u158	96	500000	trivedi, trivedi-sapphire	(null)	10Gb,cpu,avx512,sapphire,trivedi,rhel8
	u159-u161	96	500000	batch, gomez, gomez-sapphire	(null)	10Gb,cpu,avx512,sapphire,gomez,rhel8
	u170	96	500000	moortgat, moortgat-gpu, moortgat-l40-gpu	gpu:l40:2(S:0-1)	10Gb,gpu,avx512,sapphire,moortgat,rhel8
	u181	32	380000	moortgat, moortgat-gpu, moortgat-l40s-gpu	gpu:l40s:8(S:0-1)	10Gb,gpu,avx512,epyc,moortgat,rhel8
	u184	96	500000	bradley-gpu	gpu:a2:4(S:0-1)	10Gb,gpu,avx512,sapphire,bradley,rhel8
	u185-u186	96	502000	trivedi, trivedi-granite	(null)	10Gb,cpu,avx512,granite,trivedi,rhel8
	u187	96	502000	whelan	(null)	10Gb,cpu,avx512,granite,whelan,rhel8
	u188	96	507000	trivedi	(null)	25Gb,cpu,avx512,granite,trivedi,rhel8
	u189	128	1023350	accad-gpu	gpu:h200:1(S:0)	25Gb,gpu,avx512,granite,asc,rhel8
	u190-u191	128	1023350	batch-gpu, batch-gpu-new	gpu:h200:1(S:0)	25Gb,gpu,avx512,granite,asc,rhel8

 *The Family field contains the widely-used code name for the model of Intel Xeon (or AMD Epyc) server chips found in each Unity node. These names and their features are described in the Wikipedia article "Xeon". 

Currently, the oldest Unity nodes are "Westmere" and the newest are "Granite Rapids". Newer nodes are generally faster and support more advanced instruction sets. In increasing order of vector performance: SSE_4.2 < AVX < AVX2 < AVX512. The features are upward-compatible, so a program compiled for "AVX" will run on an "AVX512" system, but not vice-versa.

Change Log

Reviewed or Updated   13 February 2026