Home
Compare GPUs
MI250 vs MI300X

AMD MI250 vs AMD MI300X

Two AMD data center GPUs, one CDNA2 and one CDNA 3: 128 GB HBM2e against 192 GB HBM3 and 66% more memory bandwidth on the MI300X.

At a glance

Memory

MI250: 128 GB
MI300X: 192 GB

1.5x more on MI300X

Memory bandwidth

MI250: 3.20 TB/s
MI300X: 5.30 TB/s

1.7x higher on MI300X

Cheapest on-demand

MI250: --
MI300X: $2.59 /GPU/hr

MI250 is not sold on demand

Price comparison

Median price per GPU per hour by billing type, with the cheapest listing under it.

Billing	MI250	MI300X
On-demand	No on-demand listings	$2.99 /GPU/hr from $2.59
Reserved	-- /GPU/hr from $1.30 (12 mo)	$3.39 /GPU/hr from $1.61 (12 mo)
Spot	No spot listings	-- /GPU/hr from $1.60

Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Medians are across providers, one vote each, so a large catalog does not outweigh a small one. No median for reserved on the MI250 and spot on the MI300X: we only show one when at least 3 providers list the GPU on that billing type. Verify before provisioning. More on how we price.

Price history

MI250 vs MI300X price history

Weekly median on-demand price per GPU per hour

Last 12 months

Aggregating historical prices...

Specs comparison

Spec	MI250	MI300X	Difference
Memory
Capacity	128 GB HBM2e	192 GB HBM3	1.5x more on MI300X
Bandwidth	3,200 GB/s	5,300 GB/s	1.7x higher on MI300X
Bus width	8,192-bit	8,192-bit	Identical
Compute · dense
FP8	Not supported	2,614.9 TFLOPS	MI300X only
FP16 / BF16	362.1 TFLOPS	1,307.4 TFLOPS	3.6x higher on MI300X
INT8	362.1 TOPS	2,614.9 TOPS	7.2x higher on MI300X
FP32	90.5 TFLOPS	163.4 TFLOPS	1.8x higher on MI300X
FP64	45.3 TFLOPS	81.7 TFLOPS	1.8x higher on MI300X
Platform
Interconnect	Infinity Fabric 400 GB/s	Infinity Fabric 896 GB/s	2.2x higher on MI300X
Board power	500 W	750 W	1.5x higher on MI300X
Released	Q4 2021	Q4 2023	2 years newer on MI300X
Precision
Supported formats	FP16BF16FP32FP64INT4INT8	FP8FP16BF16TF32FP32FP64INT8

Compute: Vendor peak figures, dense and per GPU. Not measured throughput.

Source: MI250 datasheet and MI300X datasheet.

Where to rent them

Providers listing these GPUs, those with both first, so you can switch without moving clouds.

Provider	MI250 /GPU/hr	MI300X /GPU/hr
Cirrascale	Reserved (12mo) $1.30	Reserved (12mo) $3.12
DigitalOcean	--	On-Demand $2.59
Cyfuture AI	--	On-Demand $2.78
Hot Aisle	--	On-Demand $2.99
Oracle Cloud	--	On-Demand $6.00
Azure	--	On-Demand $7.86
Crusoe	--	Reservation on request
Runpod	--	Reservation on request