- 第 1 页 -
UAD-1 -2 效能對比
在UAD-1 論壇看到一個用戶做簡單比較, 大家可參考:-
[size=200%]la2a limiting amplifier (v5.0.0 08-14-2008)
single instance - uad-1: dsp 7.7, mem 0.0
single instance - uad-2: dsp 4.7, mem 0.0, pgm 4.5
8 instances - uad-1: dsp 47.4, mem 0.2
8 instances - uad-2: dsp 37.7, mem 0.0, pgm 6.9
conclusion:
on uad-1 you can't just do the maths based on one single plugin instance. ie. 8 instances won't require the same power as 8x one single instance. on uad-2 however it seems that 8 times the single instance equals the required processing power for 8 instances. i'll be making rough projections based on the 8-instance results from now on. "dsp loadlock" doesn't affect the la2a notably.
uad-1: ~17 instances
uad-2 solo: ~21 instances, ~1.25x uad-1
uad-2 duo: ~42 instances, ~2.5x uad-1
uad-2 quad: ~85 instances, ~5x uad-1
[size=200%]plate 140 reverb (v5.0.0 08-14-2008)
putting the eq section in/out or alterning any of the other controls on the plate 140 did not result in any higher or lower dsp load.
the max. instance count per uad-2 core was 5. even though dsp load lock was set to 100% and the dsp1 usage was at 82.4% i couldn't add another instance (~16.5%) which would have resulted in 98.9%. instead the uad-2 added the sixth plugin to dsp2. clearly a load balancing issue (even if the limit turned off). similar applies to the uad-1, but far worse. that means i'll be comparing similar dsp usages, ie. 4 plugins on uad-1, 5 plugins on uad-2.
single instance - uad-1: dsp 25.7, mem 11.6
single instance - uad-2: dsp 16.5, mem 1.2, pgm 22.9
4 instances - uad-1: dsp 89.8, mem 46.1
5 instances - uad-2: dsp 82.8, mem 6.0, pgm 50.3
conclusion:
once again on uad-1 you can't just do the maths based on one single plugin instance. on uad-2 you can. definitely by now. "dsp loadlock" doesn't affect the plate140 notably.
plate140 on uad-1: ~4 instances
plate140 on uad-2 solo: ~5 instances, ~1.35x uad-1
plate140 on uad-2 duo: ~10 instances, ~2.7x uad-1
plate140 on uad-2 quad: ~20 instances, ~5.4x uad-1
[size=200%]neve 88rs channel strip (v5.0.0 08-14-2008)
observations on uad-1:
with all modules and the power button turned off the plugin uses 3.4% of dsp power. with the power button turned on and the rest still off it goes up to 5.4%. the gate/exp adds 2.6%, the limit/comp adds 3.2%, the eq adds 2.2% (eq buttons don't affect the usage), the cut filters as well as the other controls add nothing. the 88rs requires between 5.4% (all off) and 13.2% (all features on) on an uad-1 pci.
dsp loadlock feature:
with the power switch on and all modules off the plugin uses 5.5% of one core. max. that is, since loadlock is enabled. thus turning the power switch off doesn't lower the dsp load. also turning on any of the modules (ie. gate/exp, limit/comp, or the eq) does not add any more dsp usage either. meaning the plugin always uses 5.5% dsp, whatever you switch on or off.
![]()
snap0001.jpg (34.14 KB) Viewed 24 times
observations on uad-2 with "dsp loadlock" disabled:
with all modules and the power button turned off the plugin uses 1.8% of dsp power. with the power button turned on and the rest still off it goes up to 2.6%. the gate/exp adds 1.4%, the limit/comp adds 1.7%, the eq adds 0.9% (eq buttons don't affect the usage here as well), the cut filters as well as the other controls add nothing. however turning all features on requires only 5.5%, you can't just add the percentages. concludes: the 88rs requires between 2.6% (all off) and 5.5% (all features on) on one uad-2 core.
hardware usage (full fledged):
single instance - uad-1: dsp 13.2, mem 0.1
single instance - uad-2: dsp 5.5, mem 0.0, pgm 15.9
8 instances - uad-1: dsp 94.8, mem 0.4
8 instances - uad-2: dsp 43.7, mem 0.0, pgm 15.9 (shared)
conclusion:
interesting results. and wow, 8 instances of a full fledged 88rs on one single uad-1. not bad. however, i had to up the dsp limit to 100% on the uad-1 in order to enable the 8th instance. with a 95% limit the load balancer went nuts, even tho we're still only at 94.8%. the driver always had those well known problems with dsp usages peaking above the dsp limit, even if it's just for a moment. if often happens when you have a plugin chain (eg. multiple inserts stuffed with uad plugs) that exactly fits below the dsp limit. when you disable and reenable the whole chain at once you frequently get the "plugins have been disabled" message and usually one of the plugs has been disabled internally. that's what i mean. i also had to redo this test with "dsp loadlock" off in order to get proper results.
uad-1: ~8 instances (full-fledged)
uad-1: ~29 instances with gate/exp only
uad-1: ~27 instances with limit/comp only
uad-1: ~32 instances with equalizer only
uad-1: ~22 instances with equalizer and limit/comp only
uad-2 solo: ~18 instances (full-fledged), ~2.1x uad-1
uad-2 solo: ~19 instances with equalizer and limit/comp only
uad-2 duo: ~36 instances (full-fledged), ~4.3x uad-1
uad-2 duo: ~39 instances with equalizer and limit/comp only
uad-2 quad: ~73 instances (full-fledged), ~8.6x uad-1
uad-2 quad: ~78 instances with equalizer and limit/comp only
[size=200%]la2a limiting amplifier (v5.0.0 08-14-2008)
single instance - uad-1: dsp 7.7, mem 0.0
single instance - uad-2: dsp 4.7, mem 0.0, pgm 4.5
8 instances - uad-1: dsp 47.4, mem 0.2
8 instances - uad-2: dsp 37.7, mem 0.0, pgm 6.9
conclusion:
on uad-1 you can't just do the maths based on one single plugin instance. ie. 8 instances won't require the same power as 8x one single instance. on uad-2 however it seems that 8 times the single instance equals the required processing power for 8 instances. i'll be making rough projections based on the 8-instance results from now on. "dsp loadlock" doesn't affect the la2a notably.
uad-1: ~17 instances
uad-2 solo: ~21 instances, ~1.25x uad-1
uad-2 duo: ~42 instances, ~2.5x uad-1
uad-2 quad: ~85 instances, ~5x uad-1
[size=200%]plate 140 reverb (v5.0.0 08-14-2008)
putting the eq section in/out or alterning any of the other controls on the plate 140 did not result in any higher or lower dsp load.
the max. instance count per uad-2 core was 5. even though dsp load lock was set to 100% and the dsp1 usage was at 82.4% i couldn't add another instance (~16.5%) which would have resulted in 98.9%. instead the uad-2 added the sixth plugin to dsp2. clearly a load balancing issue (even if the limit turned off). similar applies to the uad-1, but far worse. that means i'll be comparing similar dsp usages, ie. 4 plugins on uad-1, 5 plugins on uad-2.
single instance - uad-1: dsp 25.7, mem 11.6
single instance - uad-2: dsp 16.5, mem 1.2, pgm 22.9
4 instances - uad-1: dsp 89.8, mem 46.1
5 instances - uad-2: dsp 82.8, mem 6.0, pgm 50.3
conclusion:
once again on uad-1 you can't just do the maths based on one single plugin instance. on uad-2 you can. definitely by now. "dsp loadlock" doesn't affect the plate140 notably.
plate140 on uad-1: ~4 instances
plate140 on uad-2 solo: ~5 instances, ~1.35x uad-1
plate140 on uad-2 duo: ~10 instances, ~2.7x uad-1
plate140 on uad-2 quad: ~20 instances, ~5.4x uad-1
[size=200%]neve 88rs channel strip (v5.0.0 08-14-2008)
observations on uad-1:
with all modules and the power button turned off the plugin uses 3.4% of dsp power. with the power button turned on and the rest still off it goes up to 5.4%. the gate/exp adds 2.6%, the limit/comp adds 3.2%, the eq adds 2.2% (eq buttons don't affect the usage), the cut filters as well as the other controls add nothing. the 88rs requires between 5.4% (all off) and 13.2% (all features on) on an uad-1 pci.
dsp loadlock feature:
with the power switch on and all modules off the plugin uses 5.5% of one core. max. that is, since loadlock is enabled. thus turning the power switch off doesn't lower the dsp load. also turning on any of the modules (ie. gate/exp, limit/comp, or the eq) does not add any more dsp usage either. meaning the plugin always uses 5.5% dsp, whatever you switch on or off.
observations on uad-2 with "dsp loadlock" disabled:
with all modules and the power button turned off the plugin uses 1.8% of dsp power. with the power button turned on and the rest still off it goes up to 2.6%. the gate/exp adds 1.4%, the limit/comp adds 1.7%, the eq adds 0.9% (eq buttons don't affect the usage here as well), the cut filters as well as the other controls add nothing. however turning all features on requires only 5.5%, you can't just add the percentages. concludes: the 88rs requires between 2.6% (all off) and 5.5% (all features on) on one uad-2 core.
hardware usage (full fledged):
single instance - uad-1: dsp 13.2, mem 0.1
single instance - uad-2: dsp 5.5, mem 0.0, pgm 15.9
8 instances - uad-1: dsp 94.8, mem 0.4
8 instances - uad-2: dsp 43.7, mem 0.0, pgm 15.9 (shared)
conclusion:
interesting results. and wow, 8 instances of a full fledged 88rs on one single uad-1. not bad. however, i had to up the dsp limit to 100% on the uad-1 in order to enable the 8th instance. with a 95% limit the load balancer went nuts, even tho we're still only at 94.8%. the driver always had those well known problems with dsp usages peaking above the dsp limit, even if it's just for a moment. if often happens when you have a plugin chain (eg. multiple inserts stuffed with uad plugs) that exactly fits below the dsp limit. when you disable and reenable the whole chain at once you frequently get the "plugins have been disabled" message and usually one of the plugs has been disabled internally. that's what i mean. i also had to redo this test with "dsp loadlock" off in order to get proper results.
uad-1: ~8 instances (full-fledged)
uad-1: ~29 instances with gate/exp only
uad-1: ~27 instances with limit/comp only
uad-1: ~32 instances with equalizer only
uad-1: ~22 instances with equalizer and limit/comp only
uad-2 solo: ~18 instances (full-fledged), ~2.1x uad-1
uad-2 solo: ~19 instances with equalizer and limit/comp only
uad-2 duo: ~36 instances (full-fledged), ~4.3x uad-1
uad-2 duo: ~39 instances with equalizer and limit/comp only
uad-2 quad: ~73 instances (full-fledged), ~8.6x uad-1
uad-2 quad: ~78 instances with equalizer and limit/comp only
回复 ker 的帖子
比UAD-1 强劲很多啊~
我對比了 ＵＡＤ 的 所有 ＥＱ，如果用在 ＢＡＳＳ和 ＫＩＣＫ， 我毫不猶豫會選用 １９７３或 vibe-eq
ＬＩＭＩＴＥＲ 方面， stillwell ＥＶＥＮＴＰＬＵＳ 跟 ＷＡＶＥＳ ＡＰＩ２５００ 有一拼．．．．
ＳＰＬ ＴＲＡＮＳＩＥＮＴ ＤＥＳＩＧＮＥＲ 比 ＳＴＩＬＬＷＥＬＬ ＭＯＮＳＴＥＲ 又好很多
uad 的 ＮＥＶＥ３３６０９ 跟 ＰＯＣＯ 的 ＴＵＢＥＴＥＣＨ ＣＬ１Ｂ 也高。。。。
[ 本帖最后由 himhui 于 08-9-23 21:05 编辑 ]
可能不够时尚