#825 Results for --model UD-IQ4_XS

Description

Just want to say thank you for a quick update here are the results for --model UD-IQ4_XS

CPU 9950X3D
RAM 64GB
GPU RTX5080 16GB.
NVME
Windows 11


32k Token Context size, everything else default

Time Status Prompt Reused Output Tok/s VRAM hit rate Duration
01:59:27 PM	Done	61	0	23,947	26.6	72.7% +15.1% PCIe	902.9 s
01:58:02 PM	Done	18	0	8	23.7	38.6% +27.2% PCIe	1.2 s
Since Sun 01:57 PM: 2 requests · 79 prompt tokens read at 37 tok/s (0 reused) · 23,955 written at 26.6 tok/s



128k Token Context size, everything else default

Time	Status	Prompt	Reused	Output	Tok/s	VRAM hit rate	Duration
02:37:17 PM	Done	64,723	64,507	510	27.1	61.9% +20.8% PCIe	22.0 s
02:37:04 PM	Done	64,256	63,881	250	26.7	54.5% +23.7% PCIe	13.1 s
02:36:37 PM	Done	63,300	63,136	579	24.0	61.0% +21.1% PCIe	26.3 s
02:36:26 PM	Done	62,929	62,421	207	29.2	64.6% +19.5% PCIe	10.5 s
02:35:36 PM	Done	61,570	48,115	851	30.6	67.2% +18.0% PCIe	49.6 s
02:35:23 PM	Done	47,904	47,783	211	25.7	58.8% +22.1% PCIe	10.4 s
02:35:15 PM	Done	47,642	47,573	141	25.4	59.5% +21.3% PCIe	7.8 s
02:35:00 PM	Done	47,440	47,327	133	31.6	65.5% +19.0% PCIe	6.4 s
02:33:55 PM	Done	47,204	47,177	124	25.9	60.9% +21.0% PCIe	5.7 s
02:33:43 PM	Done	47,016	46,639	161	22.1	56.5% +23.0% PCIe	11.1 s
02:33:20 PM	Done	46,195	46,027	444	22.9	62.1% +20.1% PCIe	22.6 s
02:33:09 PM	Done	45,807	45,639	219	27.6	62.5% +20.5% PCIe	10.9 s
02:32:04 PM	Done	43,748	43,663	1,890	30.2	66.6% +18.4% PCIe	64.8 s
02:31:56 PM	Done	43,558	43,428	104	23.7	61.4% +21.0% PCIe	7.2 s
02:31:37 PM	Done	42,866	42,773	562	33.0	71.6% +16.3% PCIe	19.0 s
02:31:28 PM	Done	42,597	42,462	177	29.0	68.5% +17.1% PCIe	8.9 s
02:31:14 PM	Done	42,200	42,010	263	24.9	64.4% +19.2% PCIe	13.4 s
02:31:02 PM	Done	41,817	41,490	190	22.7	61.0% +20.4% PCIe	11.9 s
02:30:40 PM	Done	41,356	31,505	130	25.0	57.3% +21.7% PCIe	22.1 s
02:25:51 PM	Done	31,510	20,401	9,517	35.9	70.2% +17.4% PCIe	288.2 s
02:25:36 PM	Done	20,270	16,858	131	32.1	64.9% +20.7% PCIe	14.4 s
02:24:55 PM	Done	16,591	0	268	36.8	62.8% +21.9% PCIe	40.4 s