#[DUAL GPU SYSTEMS] Performance Issues and Troubleshooting
1 messages · Page 5 of 1
yup, at least I don't think normal TIM can have those temps. Mine was running 61-64ºC with 74-77ºC hotspot
damn
better temps than me
this was at 355watts?
I was running at 300-320W (slightly undervolted it)
ah so stock of the 7900xt
yup, I wasn't enjoying seeing spikes of 400W on that card
specially with another card running LSFG
It's one of the reasons I sold it. I loved it so much, but the power consumption was too much for me
rip
I wouldn't worry about yours. I would consider that way better than average
I tend to undervolt my 7900xt to 270W and with slightly higher fps than stock
but that was stock
makes sense tbh
my 7900xt is tiny asf compared to other cards
model?
reference
very interesting, I wonder if you're affected by the vapor chamber issue that some reference models had
which made temps with it sitting "up" a lot better (vertical mount)
I doubt that because those cards were hitting 105c easy
yea nope
vertical was worse for me 😄
true, but weren't those xtx models and without ptm?
yea
interesting, but still I wouldn't worry about it
This brick was running 80ºC on the core and 107ºC hotspot with factory tim btw 😛
Now it runs at 71ºC on the core and 89ºC on the hotspot
much better
I think this is tiny atleast
I still think it's a bad cooler though, big, chunky, but bad cooler
yup, that looks great for the size though!
feels like it weighs 2kg though lol
even better temps than my previous 4060 ti ventus, which I had to run it at 80% tdp in order to not hear the fans kick in semi-loud
and it's a 160W card
the hell
💀
MSI makes amazing coolers, but not ventus (or mech) models. Those suck so bad, it's insane.
after the 7900xt I honestly prefer reference cards more, atleast it is compact and not loud
I've seen mech models everywhere
It's insane how massive the jump is from Ventus to "gaming x" models. It's like comparing an intel stock cooler to a noctua nh-d15
and suprim is like putting a 360mm AIO
lmao those are big boys
yup, and I don't know if at your place is normal, but here, most refurbished models are ventus models
It's like they are ticking time bombs
I've seen ventus a lot too
2nd hand market:
with undervolt 277w
and better perf than stock
exaaaactly
coolers on components are not the problem of today.
nearly all hi perf stuff has some vapour chambers and huge heatsink + whatever cheap fan (they skimp on fans SO MUCH)
it can be noisy but it will cool down said component with ease.
bad case design & air flow are real culprits. Mostly gpus and SSDs are affected.
Customers who buy cheaper GPU skews tend to look for other budget pc parts, and buy small cases with no airflow.
I agree with you besides the 5090. The FE card will literally burn itself to oblivion once summer hits. And no, I'm not talking about the connector shenanigans.
Look at the memory temperatures right now in winter... they're already hitting above 90ºC
On a open air bench...
Yeah but not on an air bench
oh and also not on a 40 dBa cooler 😛
tjmax is 100 celsius for gddr6~
everyone goes for the safe bet of 95C throttle
there are zero to non guidance for gddr7 temps
guidance "none" got me.
clown fiesta.
But one could imagine when the 5090 card gets 95C~ throttle at open case, what it would do in your typical setup: with 80C hot air nicely syphoned right from your CPU 😄
one of the reasons I want to do a dual gpu setup.. to use frame gen and not push my main card so much, as I will absolutely not buy 50xx series.
that's insane
damn am5 motherboards
trash af
am4 motherboards for sub $200 supported bifurcation but not $500 am5 ones
i got asus 650 tuf and I hope that I can believe this:
Dang TUF 7900xtx is 2.9 inch thick.. welp I dunno if I wanna go single GPU (XTX) or go with evga 3080 and 6600xt. 6600xt case hasn't open and I can send it back
riser possible ? mount somewhere else or vertically. PCIE 4.0 riser is 40$
If I want to use riser.. where can put it in? leave it on floor inside the case? I don't know lol
im a fan of DIY, so i would propably do that but (and gonna do exactly that), but secure the gpu somehow so it cannot move at all and still has some air intake.
underclock the core if not needed and OC the memory of needed as I heard the choke point for lsfg is memory
(gpu mem)
I've been googling all the riser dual gpu setups and find nothing of it
😄 ye ... deep web on som forums logged in maybe
wdym?
I used to do dual gpu with riser 😄
I'm just bit worried about my 3 inch thick XTX fitting in with my secondary 6600xt gpu.
hm... thunderbolt gpu setup ? the cases are 150$ 😄
I might have messed up. I've ordered the pc parts before I found out people using dual gpus with LSFG
pc case has no vertical slots?
When an b650's manual couldn't tell me straight if the second slot was pcie 3.0 x4 or not, that was when I knew something was wrong
its that bad huh
devolving from am4
Fractal Design North ATX mATX Mid Tower PC Case - Charcoal Black Chassis with Walnut Front and Mesh Side Panel
the only motherboard that has what I want is the proart mobos
rip
doesn't seem like it
mATX are not really made to fit even one big gpu inside, but its gonna be tight fit even with riser for 2
yeah I'm f up lol
my case isn't the greatest but man it has space
had vertical slots
felt like I hit the jackpot
oh well I can go single GPU with XTX and send 6600xt back
ww its not for you ?
what did they put there xD
Maybe I'll do this in like 5 years lol
"noooo look at the table nooooo don't ask straight away"
And the table was a convoluted mess
yea but IS your secondary pcie on that 650m?
PCIE1x1 ? 😄
I was looking at b650's to pick one for a friend since he wants to go for a ryzen 7 9800x3d
But wanted to choose one with potential for a second GPU if he ever wants to do it
and I kid you some motherboards I never found out what the second slot actually ran at
did you manage to find anything 😅
its like a wasteland there
proart mobos are the only ones and they barely exist
no
but my friend told me he would rather hit his buggy gtx 1650 with a sledgehammer than repurpose it for a second GPU
lmao
So I guess any b650 is fine
I would only run micro ATX case if I had a decent integrated graphics CPU or a micro ATX b550 and miraculously got an rx 6400
crazy
It arrived 🗿
how... how ... air ? 😄
No air, only hope
to pull hot air from below
What is that
Basically a way to install a row of fans on a PCI case bracket
Damn, based
I mean what GPUs are those
ah whoops, my bad. I was on mobile and didn't see the correct reply.
They are MSI RTX 3070 suprim (top) and MSI RTX 4080 Suprim X at the bottom
Aww, they are cuddling
MSI RTX(3070 + 4080)
Reminds me of my and my brother's GPUs
His MSI Mech RX 5700 looks very similar to my MSI Ventus 2x RTX 3060 12GB
well rip, it seems that less mm of clearance destroyed my temps
that's with 80% power limit btw
1mm of extra clearance reduced temps by 10ºC and fan speed to half
Are they already undervolted?
but with my motherboard and case config I can't push it higher. The 4080 has amazing temps though
only 80% power limit on the 3070
These are the 4080 temps
My RTX 3060 regularly reached 80°C with no undervolt
yeah but this is a suprim card, has amazing temps but I literally have <1mm clearance with the bottom card. It's being chocked
these are the temps when I gave it +1mm clearance
It's a miracle that it's not thermal throttling lol
4080 Suprim
Width
140 mm
142 mm
5.5 inches
5.6 inches
Do you have to use Full Tower case for dual gpus? I'm just learning your ways.
This is mine
I have a question. I put my pc together as dual gpu. I ran Spider-Man 2 without activating lsfg in this state. But sub gpu still works. Is this normal?
My personal computer specification
i7 6700k
gtx1070(main) + rx6400 (dp connected)
Interestingly, even though my pc is a low specification that doesn't even meet the game's FHD recommended specifications, the game runs smoothly with a very amazing QHD resolution + high preset.
intel i7 6700k
GTX 1070 + RX6400
DDR4 16g
1440p
High preset
fsr3.1.3 performance + fsr3.1.3 FG
lsfg off, only fsr fg
afmf2 off, fsr3.1.3 fg on
the 2nd gpu working without lsfg is due to the gpu passthrough
the frames that are rendered from the 1070 still get sent to the 6400
that shows up as the usage you see
GTX 1000 series achieved the goat status!
my setup when I was trying to get dual gpus working 💀
don't mind the cables and the dust
or the duct taped ssd
or mixed ram
yup
if my setup worked, that system would have been a beast
temps weren't bad
but my issue screwed it all up
I noticed you guys use fan cpu coolers, so the AIO block CPU cooler don't conflict it right?
if it doesn't take too much space, it should be good
awesome!
you're still planning to use dual gpus?
yeah I plan it someday in the future
nice
same lol
I wanna go to am5 for that pcie gen 4 but the motherboards are complete ass
no bifurication support unless I spend a crapton
So you're saying that sub gpu isn't affecting performance at all?
best I can do is gen 4 x4
it can if it gets too overloaded with frames but your 1070 is no where near that powerful for that
you're good
Yeah I noticed some mobo are expensive if I wanna get Bifurication like 8x 8x
I can get an import mobo that has it for like $600 aud
still expensive asf
and who knows what the warranty is like
I noticed some people here using pcie chipset slot instead of pcieCPU slot.. does it make it any difference?
My poop-like pc got weird after the dual gpu installation.
it might
that card is tiny wtf
yeah 1000 series is a small gpu
cpu = i7 6700k 4.0
gpu = gtx1070 + rx6400
ram = ddr4 2400 16g
power = 500w
1440p QHD
40 x 3 = 120fps
마블 스파이더맨 2
Marvel's Spider-Man 2
lsfg duel gpu lsfg
tbh I think it could be related to my issue
I think its around 1.5 in thick
but dranas got it fixed with pcie 4 x8 x8
so not sure
but 99% of people use chipset slot
could be related
I see
The big gpu is 1070.
The small gpu is rx6400
thats a badass 6400 gpu lol
My 1070 is powerful.
Yeah its a good card
My 1070 will break the head of 3070.
I turned on the fsr frame generation, and the vram usage of the sub gpu increased.
fsr from LSFG or the amd software?
only fsr
lol
The lsfg program itself has not been executed.
more frames = more usage
that applies for vram too I guess
oh yeah 6400 doing the FG can raise up the vram usage
yes
But the one that actually uses fg is 1070.
In fact, the 1070 increase in utilization when using fg.
^
if your 1070 is outputting more frames, the 6400 will have to work harder too
Do I need a vram for operation?
wdym?
its just what happens when you transfer frames between 2 gpus
lsfg will use even more vram but it should be enough for anything you are doing
You make it sound like you're talking about passthrough
When a higher framerate is being inputted into LS, it uses more GPU
I understand. Are you saying that rx6400 also needs vram to output additional frames to the screen?
Yes, because a GPU needs VRAM to function
LS doesn't use much VRAM though, and when you're offloading it to a secondary GPU, it has no impact on the render GPUs VRAM
so VRAM is mostly out of the picture when it comes to dual GPUs
I didn't use lsfg, but rx6400 is responsible for screen output, so I'm going to use vram more to process more frames?
Not VRAM, no, but PCIe bandwidth, because more data is being transferred
The data doesn't stay in the VRAM though, it just copies through it
But I didn't run lsfg and the vram increased. I only enabled fsrfg.
Oh. By how much did it increase?
100mb
What if you go into a game menu where you typically easily get hundreds of FPS
I can't try it now.
The main 1070 increased by about 300mb,
The rx6400 increased by about 100mb.
Only fsr fg
Makes sense, yeah Opbush was probably right
More frames copying means more VRAM usage ig
A negligible amount though. Not sure how much VRAM it's using on the render GPU
@rain hullUnfortunately my problem wasn't resolved by resetting NVCP to default values.
Here is a screenshot of my HWINFO while I'm running Fortnite, just sitting in the menu with 120 FPS. It seems weird to me that the 4070 Ti Super is under max load, but just pulling 25% of its TDP.
Anyone see anything odd that could possibly be the reason for the 4070 not bringing the performance, like it does in the top PCIE Slot?
Usually it would be around 70% Utilization there
I'm using a Fractal Design Torrent Solid which is a pretty big case yes
Thanks I'm exicted to see your dual gpu results!
I'm going to swap my 3070 for the 6800 I sold to a friend to see if the LSFG card can breathe, since it has 3mm more of clearance
If it doesn't work, I'll have to try running a PCI-E adapter with dual 120mm fans behind the card to pull hot air from the cooler, to see if it helps the card to not thermal throttle
When I removed the 3070 yesterday, it was so damn hot to the touch that it was almost burning my fingers, since it's a metal cooler
what motherboard are you running? and CPU?
and it seems you're CPU bottlenecked, look at CPU busy metric
That's actually pretty interesting, I never thought about checking that metric when I was having issues. It could really be a CPU bottleneck issue when running LSFG
@clear sundial here are my specs
Hm, but that's actually without any FG on
PCI-E 3.0 x4 you're completely bottlenecked by the slot
PCI-E 3.0 x4 is already low bandwidth, but running it as a chipset slot will completely tank performance. Your chipset is probably overloaded, giving you CPU spikes
Unfortunately, since we can't measure chipset usage with software, there's no way to be 100% sure, but considering I was having issues with PCI-E 4.0 x4 (chipset), that's probably it
'Intel® Z690 GAMING Motherboard with Direct 16+1+2 Phases Digital VRM Design, PCIe 5.0 Design, Fully Covered Thermal Design, 4 x PCIe 4.0 M.2 with Enlarged Thermal Guard, 2.5GbE LAN, 120dB Hi-Res Audio with WIMA, Rear USB 3.2 Gen2x2 TYPE-C®, RGB FUSION 2.0, Q-Flash PlusSupports Intel® Core™ 14th/ 13th /12th processorsDual Channel Non-ECC Unbuff...
You are running your main card (4070TIS) at the bottom slot, or at the top?
This actually makes sense lol. Are there MOBOs that have 2 PCIE slots directly running to the CPU?
At the bottom right now. While in the top slot, all games refused to use it as the render GPU
On am5 yes, so it's probably the same for intel chips, but they are expensive
Which I find weird in itself
The easiest way for you to take advantage of your current setup, imo, would be running the main GPU on the same slot as you're running and locking the FPS 60. Then use LSFG to interpolate frames using the closest multiplier to your monitor's refresh rate
PCI-E dictates the max FPS you're going to run, so your 4060 should be installed in the fastest slot of your system
You need to make sure your advanced graphics settings in Windows have the 4070 ti S as the "performance GPU"
like this
If, you need high base FPS, you're going to need to be running at high PCI-E speeds. If you only care about interpolated frames, then the LSFG GPU needs to be run on the fastest slot and the "main gpu" on the other slot
I did and tried it both on Windows 10 and 11.
Well then this is a deal breaker. The game I'm playing a lot right now (Kingdom Come Deliverance 2) crashes instantly with this configuration. And the performance drop means I'd have to resort to higher FG multiples, ruining input latency.
As it stands I'll stay with a single GPU setup, possibly upgrading that in the future. Brings me better performance all around. Quite unfortunate
Yeah understandable
Windows version: 23h2
Driver(s) version: latest nvidia
GPUs used: render RTX3080, lsfg RX6400
Platform information (CPU+Mobo): intel i9 11900k - ASUS PRIME Z590-P Intel
How monitors are connected and used: 1 single 4k monitor via displayport to the RX6400
Any anomalies:
Weird performance loss. I used to play with lsfg with the window mode check because it reduces drastically the latency (hardware indepentent flip), but now with the dual gpu setup, not only is a stuttery mess, but the performance is terrible and the fps cap doesnt work (at least it never show i capped to 30-40fps, it always fluctuate like uncapped fps). when i uncheck the window mode, it doesn't stutter but it doesn't reach the desired performance at any resolution, i tried even in 1080p with the resolution scale all the way down
for a reason the pcie is capable of 4.0 but its working on 3.0, in any case, it should be enough for 2k
i'm using DXGI since WGC is not wirking properly on 23h2
It's probably not getting nearly as much PCIe bandwidth in the bottom slot.
I think it is enough badwidth for that card, but its still sitting on the chipset rather than on cpu, could lead to some latency, but his issue is different
All monitors connected to lsfg gpu?
i just have 1, and its connected to the lsfg gpu
via DP
i'm just testing, it has to be a bandwidth problem, now it run "as expected" without window mode checked in lsfg but its not nearly as powerful as stated in the gpu chart
because if i want 40x3 ->120 i have to lower the resolution scale below 50%
and for me its really distracting all the UI artifacts at 50%
when i run lsfg in windowed mode it just doesnt cap my fps and its a stuttery mess
i don't know, my 3080 was capable of running lsfg flawlessly, my only concern was the performance hit in the VRAM, that some games required me to lower my resolution a lot, but thats it
and with the dual setup at minimum resolution and resolution scale still induces more latency than when i was using my 3080 alone
What is the bandwidth for the 6400?
That should be enough tbh
No, PCIe 3.0 x4 simply cannot transfer that much data
128gbps
at least is what it shows in my pc
3.0 x4 has 4gb/s of bandwidth
the thing is, my pcie is 4.0, but its working in 3.0 for some reason
is there a way to change that?
Chances are your motherboard is wired that way
I'd check the manual to see what the expansion slots run at
i'm not even capable of running 30x2 at 50% scale resolution xD what a shame
I also read somewhere that everything using the PCIE Chipset is sharing the bandwidth. Maybe that also plays into it
meh, i'm disappointed, guides misleaded me, they say with a 3.0 could be enough for 2k and i can't even run at 1080, and also the rx6400 was enough too, i'm not sure what i'm doing wrong, or if im not wrong and things are like this
I didn't mislead anyone

Specifications:
-B550 Asus Plus AC HES
-Ryzen 5 5500
-RTX 3060 12GB (Main Graphics Card)
-RX 6600M (LSFG Graphics Card)
-40 GB DDR4 3200 Mhz RAM
Tested game settings:
-Main GPU (RTX 3060 12GB) was running at PCIE 3.0 X16
-LSFG GPU (RX 6600M) was running at PCIE 3.0 X4
-1440p 165hz monitor
-82 FPS cap in RTSS to ensure smoothness and avoid teari...
My set-up runs my rx 6600m on pcie 3.0 x4
And perfectly at that
I'm playing Kingdom Come right now and there's a dual GPU issue. Even if I don't apply lsfg, the game won't start in a dual GPU configuration. Here, the 3070 is the master GPU and the second GPU is the 1070Ti. Also, the DP monitor output is output to the 1070Ti, and in this situation, if you select 3070 as the default high-performance GPU for Windows 11, the game will not start. However, if I connect the monitor cable to the 3070, the game will start. Conversely, if I select the 1070Ti as the default high-performance GPU, it will run. There seems to be a dual GPU issue with CryEngine. Does anyone know how to fix this?
4090 + RTX A4000, i cant get A4000 working right in this mobo. Probably get a different mobo this weekend
Should I send back my 4060 and get a 7600XT instead for ls? They're about the same price here. Is the 7600XT better than the 4060?
I'm assuming it wouldn't be better at handling graphics than the 4070 I already have
I also don't see that card in the dual GPU chart, depending on if you guys think I should get it, I could test and add it
When I used 4.0 x16 + 3.0 x4 (secondary), it limited base FPS passthrough to around 220fps. Sylux's might be lower than mine for that reason, but their motherboard isn't going to be passing through 1440p 360fps any time soon.
You could, though 7600s are known to have massive stability issues when running LS (as in, sudden full system shutdown).
The rx7000 series also seems to not have improved much from the 6000 series for LSFG at all.
He should get a 6650 xt instead
That one is better than the 4060 for ls?
The 6650 xt did better at LSFG than a 7600, but I don't know if the 7600 was better than a 4060
I guess I was only thinking about vram
So.. im.not sure id call this a performance issue but on my 5080ti (pcie 5 x16) and 4070 (pcie 3.0 x4) i get 190ish base fps on Cyberpunk benchmark with no LSFG but when lsfg is active I only get 140ish base.
Granted I am running 4k with low settings to get the base rate that high.
Is this the cpu overhead from lsfg? I assume I am cou bottlnecked here even thought I am only at about 60% usage.
Ah nevermind, as I was testing again with rt overdrive as i was typing that, I saw only 1-2fps drop. So I guess it is cpu.
whats your cpu
You mean pcie 4.0 x4?
Nope
Intel z690 the second port in an pcie 3.0 x16 in length but x4 in bandwith
The 3rd pcie is pcie 3.0 x16 size but only 1x in bandwith.
Dang, it's a miracle that that's transferring 190fps at 4k
When I tested 3.0 x4 it hard capped FPS to 85
I find it hard to believe that that's actually 3.0 x4
Will show you a gpuz when I get back home.
Actually I have screenshot of manual.
Msi tomahawk z690 ddr4
this tomahawk is doing 155 base 555FG with your GPU setup? Damn I should have kept my x670e Tomahawk
Somethings off i think. I just did the maths 4k at 32bit pp work out at about 33MB per frame raw at fps it works out at about 4.9GB/s.. PCIE 3.p x4 should top out at 4GB/s.. unless there is something sort of compression in use 😕
yea weird
I do know of a compression that amd can use but its for vram only I think
but it could also help pcie maybe
Perhaps.
Perhaps
indeed.
Fun fact about my stutters in Helldivers 2, the base framerate is just that bad. Played without lsfg and my fps goes down from 46 to 25 out of nowhere. It HAS to be the game and their new update. Not ruling out any other issues than could be the culprit, but HD2's case, its that bad
every time I hear about helldivers 2, they get worse and worse each update
The performance just tanks no matter the system, it's the main reason I stopped playing. What made me come back to it was discovering lossless scaling
Over 1000fps on timespy extreme
4k with lsfg 😕 - that just doesn't sound possible, but its what I am getting.
Bro got the PCIe slot of infinite power and bandwidth
Its looking that way 😱
Maybe chipset lottery is a thing now
maybe cpu lane lottery is a thing too considering my situation 😄
Maybe I got your lanes 🤷♀️
@clear sundial was pcie 4 x4 not enough bandwidth for your setup? or you haven't tried it before
I don't own a 4k monitor, but used Virtual Super Resolution on my AMD RX 6600M, my chipset pcie 3.0 x4 slot allowed a maximum of 82 base fps
Im gonna test more thoroughly tomorrow because the math dont math.
indeed
my goofy setup loses 20fps just by having a secondary card 😭
and here he is with casual 1k
Not 1000fps base obviously but lsfg generated.. i just cranked it up to max multiplier and set the scale to 25%
what was the base you could get?
Cant remember the base frames but it was sub 100.
hmm
Im in bed now so can't check.
Or maybe im talking sh*te.. i honestly cant remember I ran so many different benches.
But itbserves no purpose at all in the real world.
Just a shame none of this works for vr.. fs 2024 in vr is my thing at the moment.
it felt similar to PCI 2.0 x8 when I was testing it. But I do have a lot of USB devices plugged in, so... not great
damn, looks like rng with motherboards ngl
I'm gonna retest the 6800 I've used before on that config (PCI-E 4.0 x4) which I think was the huge bottleneck, since I can't use the 3070 on my system without thermal throttling
pcie 4 x4 being a bottleneck is crazy
yes but it isn't "true PCI-E 4.0 x4". It's PCI-E 4.0 x4 shared with every other chipset device
every USB and sata device shares those PCI-E lanes
And on AM4 platforms, the chipset is bottlenecked by PCI 3.0 x4 connection to the CPU https://www.overclockers.ua/motherboard/asus-prime-x370-pro/01-asus-prime-x370-pro.png
this is why chipset PCI-E performance is very inconsistent
I see, so cpu lanes are the only true pcie performance
yea most people seem fine with it though
I bet most people aren't playing at 4K
true
I do tho 
that performance hit of yours 🤨
Which is why the dev recommended 50% res scale at 4K. It reduces the data traveling massively, even without big image loss. At 80% res scale, even my 4060 ti didn't drop fps a single time.
So AM5 chipset should be better with handling another GPU?
technically yes, but you have a lot more ways to spend that bandwidth with faster devices. So if you have fast USB devices plugged in, those will steal a lot of bandwidth
honestly my performance hit acts like a pcie bottleneck ngl
the safest way is to have a CPU PCI-E slot reserved to the LSFG GPU
$$$ mobo
All my usb slots are basically filled 
I had to buy a powered USB hub, so that explains why I can't have the same experience as ravenger 😛
how much usb wtf
I tend to use overkill amounts of usb too but thats just something else
It's more of ease of access. When you have a 20KG case and you have to shuffle USB pen drives, it becomes conbersome
And another thing is that AM5 motherboards, unlike AM4, skimp out on USB slots a lot, unless you pay more than double the price.
I also have like 8 USB + Sata devices plugged in lol
maybe should try another time without all those plugged in
I looked into techpowerup database and it told me it was 3mm leaner compared to the 3070 suprim, but that's not true since the 3070 suprim fits, while this one doesn't
Sue them
My last hope will be this:
Hey, do you by any chance have that glorious 3D print blueprint for the fan mounted on top of the GPU?
Does this look fine for you guys?
I got this when my fans failed. A couple of decent fans ats its better than a stock cooler
Bought it from this guy https://www.etsy.com/listing/1417013236/active-gpu-backplate-cooling-120mm-fan?ref=yr_purchases
Thanks ❤️
3090 Ti + 4060, can achieve 100% resolution on FG, 4060 is in a PCIe4 @ 4x slot.
What's the resolution and framerate?
4K @ 60, framlimiter with RTSS (Reflex)
2X FG = 120 at 100% resolution
Game is stuck forever at 120, it doesn't even lower 1 FPS on an entire match, I'm so happy.
But why are you using 55GB RAM?
He's a legend
hahah
I have Primo Cache with 32GB of RAM Cache for the storage with 60 seconds delay write, obviously I have an UPS.
That program can literally turn your SSD into a speed demon
I have 128GB of RAM btw. You absolutely need tons of RAM for that program
The software is miles better than the standard "write cache" that Windows does.
https://www.romexsoftware.com/en-us/primo-cache/index.html
Home page of PrimoCache product which is a supplementary software caching scheme to improve the system performance.
Works on Windows 10, 11, LTSC, you name it. It's an amazing piece of software.
Damn, you have made me curious now
what cpu do you have ? I struggle to run helldivers 2 past 40 to 50 fps. I'm curious how you got to stable 60
after testing what 1050ti can do, frame latency etc with 4090 as a render gpu,
im srsly considering buying 6650xt. It seems to work.
I just need to find how gsync/freesync works in dual gpu setup configurations.
Depends on the card that is connected to your monitors.
Gsync for Nvidia and FreeSync for AMD
I'm really interested i using this? I got 64 GB of ram. Can you use it for everything or is it just for games ?
yeye, but when you are not using lsfg, and your render gpu is nvidia but your output one is AMD, that is a problem no ?
No, you just use Freesync
oh, great. ty
FreeSync and Gsync are not game dependant so it doesn't matter if you actively use the card or not
For the first time, I see these numbers
😄 I've seen 1.5tb ram usage at our prod server node...
but try to mod fallout 4. I overfilled 24gb VRAM & 64gm RAM with that.
I too run primocache but I use l2 cache instead
Got 2 128gb ssds, put software raid on them and made it the l2 cache in primocache
So far works great
What ssd would you recommend? Is 1tb for caching too much for just games?
1tb is overkill imo
Just use any drive your willing to sacrifice
I use spare 128gb because i have no use for them
I see
Should I prioritize write speed of the cache SSD instead of the actual storage on said drive
Like 256gb but with better read and write speed than a 1tb
Since I have 64gb of RAM, I could allocate like 20 for caching?
if you want to, ram cache is obviously faster than l2 cache
really depends on what you want from it
right now I use 300mb of ram just as write cache
it takes load off my hdd
and does extremely fast writes for those 300mb
Faster video game cache and audio files that I move a lot between SSDs, I also have lots of games of on both my main SSDs. If the work of writing could be off loaded to another drive and spare RAM, that's pretty much what I'm looking for
hmm so you want more writing cache
And for that usage, I guess 256gb is more than enough? Like every games' cache plus regular files
its a bit risky imo without a ups
What's a ups
yea should be good
its like a backup power supply incase the electricity goes out
or a unexpected crash happens
write cache can lose data in those situations
Ah I see, I've never had any of those so I'm not super worried
well better safe than sorry imo
but its up to you
how big are you gonna go with write cache?
1gb?
That's the part I'm not sure on, or what it exactly means
so write cache basically writes everything to ram and slowly writes it on the target drive
1gb = 1gb of data in ram
but you have instant access to it as if its already on that drive
And RAM does it in the background
yea with primocache
Cause I can spare lots of ram
Not yet but very soon, I'll buy a new drive for it
okay then
you will understand most of this stuff its pretty simple in the gui
also don't worry about block size
just make it 64kb or 128kb
up to you
hdd with primocache
hdd without
basically got my sata ssd speeds on it
even my 300MB of write cache did something
But the file has to be transferred at least once for it to be written to cache? Those speed are only possible because the cache SSD knows that file already if I'm right?
yea for read cache it does
which is why games get huge boosts from it
repetitive loading
Bought a new SSD, will buy primocache when I get it
theres a free trial
30 days
you can try it right now with that
just use ram and see what it does for you
I don't think I ever saw genshin running this smoothly before without needing to resort to a fps cap
primocache is like magic
SSD, but I remember showing to @rain hull how inconsistent the frametime graph was before
oh damn
Did you give it 1GB?
I remember when relo was playing the first decendent and it was eating his ssd
then with primocache it fixed it
I have no idea what a reasonable number is so I gave it half of my RAM (20 GB)
Genshin is really bad with frame drops, so if that helps I might just try it out
I still get the dip from first-time load
helps me tremendously in 7 days to die
But then it becomes extremely smooth
you could probably even do like 8gb
I get dips like every few seconds when exploring. It's so annoying
How much at least does it need to help with that?
4gb minimum
This good?
Looks good to me
yes
Ok then
Possible
Actually seems a lot smoother now, even though it's already installed on an Nvme SSD
PCs are really something
Yeah all of my hoyoverse games run on a SSD
I tried it on Black Ops 6's Warzone too
Less stuttering
Graph is literally flat now
I mean
I'm too spoiled now with LS.
An inconsistent 100-120 FPS feels so bad now 💀
Problem is I'm probably gonna upgrade my motherboard in a few months, so I'd probably need to buy primocache again after
Yes am5
oh nice
And now I'm even tempted to go for 64gb ram 
True
I'll stay wth AM4 for at the very least another 2 years
im back to 16gb of ram so I'm crying rn
primocache solved one of my main gripes with staying on AM4
My 3800X is already struggling with something like HZD.
It sometimes goes as high as 80% usage on all 16 Threads
which was stuttering
I just want dual gpu support and huge ram with good speeds
$260 aud for 64gb of ddr5 ram is something
cheapest I could find tbh
Ram prices are so low compared to other pc components it's wild
yup mobo is gonna cost me so much
to get bifurication support its a minimum of $600
insanity
A 16gb DDR4 3200mhz stick costs me 20 bucks
Usd tho
20 usd
Why is my frame higher than the what the games set to in lossless?
Magic
Base frames i mean, games at 80, lossless says 89
Turned off hwinfo fixed it
Now im back to dropping frames on my 2nd gpu
Lol . Turning off msi aftrburner overlay fixed it
💀
Doesnt do that on my amd build
I conducted the test on the 4070ti super + rx6400 system yesterday.
This pc's motherboard is b660m and supports pcie4.0.
The vram of the second gpu (6400) has not changed at all, regardless of whether the frame is increased or not, and always stays at 60mb.
Wouldn't frame increase in dual gpu have no effect on dual gpu's sub vram?
Both the gpu and ram should be utilised if dual gpu is in use.
When I play Spider-Man 2 with my pc, my stutter appears despite being a very stable base fps.
For example, if you set 60fps x 2 in that game, you don't get 120fps and it's 116 to 118.
Other games didn't have this problem. Base fps is very stable and overall usage is low, but it does cause problems.
Does lsfg performance fluctuate depending on the game?
If overlay causes problems, it's can be related to MPO support.
Your GPU & Monitor & OS have to support it.
Check in special-k if your card suports this.
One of the few thing I have to check regarding dual GPU setup, this HAS to behave correctly otherwise its a no go for me.
Using WCG in LSFG should completely eliminate the need for MPO support but.. i find the delay unbearable, maybe its placebo but something is off about WCG+VRR.
recently downloaded this. even though both my os and games are downloaded on a pretty fast m.2, it fixed a lot of frametime inconsistencies, my graph is practically flat.
godsend my man
didnt affect 1%/.1% lows but capping framerate obviously fixes that
thanks for the info, i dont think the RTX A4000 supports MPO
Hi all- just set up a 4090 as primary and 4060 as secondary. I have set Windows to use the 4090 as the preferred GPU, but when I open Fortnite and click in, I see both GPUs spike in usage- I doubt this is normal and I doubt this is a game issue. Anyone run into something similar with other setups/games?
Go in nvidia app, search for Fortnite and check the cuda cores setting at the bottom, uncheck the 4060 from it and see if that happens again
Copying over the video and frames takes GPU usage
But if it's just that the game for some reason is running on the 4060, try what Pyro suggested
Depending on resolution and PCIe bandwidth, it's normal for the secondary GPU to hover around 15-40% usage due to video copying to it in a game.
1440p QHD
high preset
FSR 3.1.0 upscale (not FSR FG)
PC spec
cpu = i7 6700k 4.5
gpu = GTX1070(main) + RX6400(lsfg)
ram = ddr4 16g
Perfectly solved the stutter in the game.
Changing the cpu priority of lsfg to 'high' greatly affected it.
gtx1070 + rx600 + intel i7-6700k
Huh
cpu = i7 6700k 4.0
gpu = gtx1070 + rx6400
ram = ddr4 2400 16g
power = 500w
1440p QHD
40 x 3 = 120fps
마블 스파이더맨 2
Marvel's Spider-Man 2
lsfg duel gpu lsfg
Game = cpu priority 'high'
lsfg = cpu priority 'normal'
game = cpu priority 'high'
lsfg = cpu priority 'high'
lsfg to cpu priority 'high', Setting the cpu priority for the game to 'normal' was the most stable.
I just noticed that when my performance "degrades" while in-game it seems the cause of it is that at some point my main GPU just stops going over ~81-83% usage in GPU bound cases
That's korean 😭 @brave mesa
Everything. It sits almost at the kernel level.
Yesterday I used it and got caught off guard when an area change in HSR happened instantly
Everyone that doesn't utilize all of their RAM should use this
Yes, L2 SSD as buffer for HDD is great.
But I already have NVMe (no HDD) so for me, RAM > NVMe
We are talking about 24GB/s vs 6.5GB/s. And you can really feel that SSD is near instant. I use a huge 32GB cache for my NVMe with a 70 second write delay with an UPS that can hold up to 10 minutes. Check this crazy score, this is an SN850X btw.
Near 5GB/s on 4k-64Thrd is major craziness!
People don't realize just how much of an impact a good Caching system has on the storage. I bet at least half of the stuttering people experience in games is due to poor reading/writing patterns in games.. things that PrimoCache can help minimize by softening the storage use.
I run it daily
Ever since I tried it
I'm tight on money so I'm using the free trial
Does it forgive me if I uninstall it and install it again once the trial expires? It genuinely is worth the money but I can't spare anything right now
It will not. Your PC generates an internal key
Which is sent to their servers. This key will check the time you have.
It doesn't matter if you uninstall it, when you activate it, your key is having the time counted on their servers.
Oh well
My suggestion: buy it.
They do not charge for updates and once you buy it, it's yours forever. It's one of those programs that cost much less than what it actually gives you.
Until you change motherboards at least 
Which I'm planning to do this year
I had issues with activation when I change my PC (3 times now) and sending a support ticket they invalidated my previous license and resent it my way.
@cedar mango https://amzn.eu/d/2w2M2Vq
I got this ssd to setup primocache with
Les précieuses performances Gen4 sont là. Voici le Crucial® P3 Plus Gen4 NVMe™ SSD, qui offre une vitesse impressionnante avec des lectures/écritures séquentielles allant jusqu'à 5000/4200MB/s. Incluant une accélération dynamique de l'écriture et une protection thermique adaptative pour garder vo...
So no, that is no problem, they are super kind and nice.
Ok,thats good then
Yes, they are not the greedy bastards kind.
You simply explain that you sold your previous motherboard and you changed it to something new and they will reset your license.
You should be good, provided the storage you want to speed up is slower than 4.5Gb/s
Yes, assign the entire storage as L2 and enjoy performance for cheap!!
I am telling you, these guys have a much much better caching algorithm than the Windows kernel itself has.
They are the "Loseless Scaling" of Storage 😂
One friend I recommended it to did say that
He is absolutely right!
In this age of "badly optimized software", there are still ABSOLUTELY GOOD pieces of software to be had.
🔥
I'm almost a system engineer. Been working with hardware all my life. I've never seen a program handle memory as good as Primo Cache and Primo Ramdrive
You know what I do?. I also have 8GB for Primo Ramdrive and send all the browser temp storage to that.
Not only I can browse the internet much faster, I can simply restart the PC and have my history cleared without effort as the Ramdrive is set to forget contents upon restart.
They are both modular and built to work alongside, so you should have no problems using both at the same time.
You can set the Ramdrive for all the stupid programs such as Spotify that loves to write tons of TEMP files... so they are "off" your SSD. This also improves system responsiveness.
I don't have that much ram 💀
my poor 16gb
What does the Cache hit rate mean btw?
its when the cache that your ram or l2 cache gathered up actually does something
Damn I'm at 84% rn
Does this matter? The inactive part
It cleared my ram, so I don't think it is lmao
rip 😭
I think it's because my ram usage is too high
I deleted the task, created a new one and it gives me an error that I don't have enough resources
Although I'd still have 6GB overhead
I am using 16gb atm and it didn't let me create a 12gb task, but 8gb is fine
Standard 32
hmm
Yeah, I was just trynna upgrade it a little since I usually have enough free ram
I have no idea how block sizes impact the performance so I just left it at default
Yeah they all are, does that mean that I should allocate it 4.5 GB or RAM? Or more?
its just how well it handles allocation
each block is 32kb which is more accurate than a 64kb block
you don't really notice much in performance so a higher value isn't bad
with 32kb your getting like a near 2gb ram overhead
on 64kb its half that
and 128kb is half of that again
it has no ties with the speed
allocate how much you want
its only the cache size
Ah okay
Did you partition your E drive?
Block size does not impact performance much when in L2, but it does for RAM. The more you use, the faster (64-128k being ideal) but it will consume more RAM if it's bigger.
No, D is a Samsung 960 Pro (SSD) and E is an SN850X 4TB NVMe
Is there a way to know what block size corresponds to how much RAM needed?
Yes. Before pressing the button on create task, it will tell you how much will "waste" depending on the block size you use.
Ok cool
Look at the "Memory Overhead", that's basically how much you're going to waste depending on that blocksize.
look at the lower left corner.
That one
I'll get my ssd on Wednesday and I'll get the free trial then
Overhead means = memory wasted
OK thank you
There's no real minimum needed.
You could just give it 1GB RAM if you wanted to
You can defer write to an L2-SSD, you cannot defer write to RAM unless you have an UPS.
Exactly. Problem doesn't care what you allocate but obviously it will be faster and more effective the more you allocate.
I use defer write with ram 😅
its only a bit though
300MB
Would it be beneficial to use some on the Boot drive?
Just to have an idea, with my 32GB allocation I can update a game on Steam and it will instantly validate upon downloading, like instantly, then I can see after 50 seconds a burst write of 6Gb/s on the disk.
Yes it will.
It will reduce the latency of the entire kernel.
I just allocated it all to my game drive
And if you're on 11, it's not "just a tiny more"
When I upgrade platform I'm probably just gonna get 64gb ram so I have 32 or more to spare for primocache
if I had more ram I'd definitely use some on my boot drive
On my system, just using 1GB for boot drive improved performance quite a bit
Yes, this one program is the sole reason why I've got 128 🙂
128gb damn
Also, do not assign L2 to boot drive, just RAM, only read cache, not write.
pretty sure am5 suffers in speed if you use 4 sticks 😔
C drive should have the safest caching possible to avoid corruption.
wut
How do I do read only?
yup, not sure how bad it is but it exists
Yes mine also suffers but you know what. Comparing 4-dimm slower overall performance decrease with 100% storage performance increase, I can assure you, you end up winning.
Man I was gonna go 4x16
very bad idea
This is basic engineering, Storage is the slowest part of the entire system.
just get 2x32
get 4 x 32 🙂
also that
Once you start using big caches you will never go back
You won't care about the 4-dimm slower access time anymore
tbf it matters for cpu bottleneck situations
but it depends on if you care about it or not
The level of "snappiness" you can achieve by separating tasks (C drive to read only cache) and game drive for read/write is unheard of.
you could have a x3d cpu and it won't matter much
@cedar mango
just don't enable defer write?
Oh ok
Is defer write safe on the game drive without a UPS?
I use 300MB without ups but it isn't recommended
Kk
Get money first
hahha yes of course
am5 = $$$$$
Alrighty then, 8 and 2GB
I'm gonna aim for 64gb of ram
One invalid shutdown can completely corrupt the game you're playing.
yup lol
What do you mean with 8 and 2?
playing with fire but its always worked fine
GB Ram allocated
Also, if you're doing L2, use RAM as well, don't use L2 alone
I'm only using L1
"mama who turned off the power"
hahahah
Both my drives are M.2 Nvme so no need for L2
Or best way ever to screw the kid... wait till he starts playing the favorite game and BAM... there it goes the power haha
just make sure to never use defer write on boot drive without a ups 💀
For NVMe never use L2, just RAM
At least the 9800X3D is slowly reaching MSRP lmao
Yes, that's what I'm doing
8 and 2GB respectively
When you have more RAM, test 16GB cache onwards... and watch your disk become god.
I'm telling you... it's out of this world performance.
I tried 12 an hour ago and it gave me a 'not enough resources' error
Watching Steam sucking all 32 cores while validating a game and seeing the RAM cache getting filled and watching the disk doing nothing is short of extacy.
But I'm deffo going for 64GB on AM5 to give them both 16GB each
even with my 2 striped ssds, I get huge decrease in stutters and insane loading speeds
Also, there is another thing that this program does to your SSD. It severely reduces the fragmentation when defer writes are enabled.
This is because the multiple read/writes are all packed into the RAM, then the program writes a single burst of 6-7GB/s on the disk and this is done "sequentially" instead of randomized as it could be done originally. This reduces fragmentation and improves performance of the storage system.
The defer-writes severely changes how the storage subsystem behaves, so the more RAM you have, the more performance you get from it. It's like literally maximizing read/write performance by using RAM as buffer.
This is absolutely correct. All the messy read/writes are re-organized in RAM, this is why it reduces stuttering.
Yup, also noticed it in games that were BAD with loading stutters every few seconds
Smooth as butter now
Yes, absolutely correct. This piece of software is among the best you could ever have in your windows, along with Loseless Scaling. The other beautiful piece of software that must be there installed in your system is Process Lasso.
You can do wonders with that software as well.
I can literally send lower important background apps to E-cores and take them out of my P-cores, also change the default hetero-policy to redirect most high performance applications to the P-cores in my 13900K.
For sure it does a much better job than the lousy "thread director" in the W11 Kernel.
Tbf AMD doesn't really have P and E cores
Yes, AMD handle this much much better.
It's still useful.
For example, the stupid Epic Games Launcher forces the Overlay Launcher every time it detects a new game. Sometimes even outside the Epic Games Launcher (steam games).
There is no way to block the App without breaking the Launcher... but, here comes Proces Lasso you simply wait till the Overlay turns on, and you "disallow" the process.. bingo. No more worries and no breaking.
It's amazing how good you can handle every process in the system with this program. I bought the Pro because in my system, it simply replaced the Task Manager for good.
I have it mapped in "Win+C" with a free tool called "HotkeyP" and I can invoke it whenever I want. Have total control of the system 🙂
ddr4 is dirt cheap now... disk drives (all.. hdd & ssd) should adopt the concept of using volatile storage for caching if needed. 1gb-2gb of ddr4 speed right next to disk drive would be insane, no sw required.
but for gaming, it's snake oil. You are limited by so many translation layers, API's, drivers and so on, that once you get your main drive to be NVMe, your gaming drive could be the crappiest of sata ssd's and for 2017~ < games even HDD and it would not touch your 0.01 percentiles a bit. There are exceptions, but frame time health is not affected by disk specs much. At least for not of the absolute newest of games that rely on texture streaming compression/decompression.
Want to get rid of stutters ? win11 + mpo + gsync / vrr + specialk is the way.
Also limiting base framerate to 60 and using lsfg is one of ways of getting just insane framepacing with rock solid 0.01 perceltile lows.
That being said... loose files modding could cause stutters depending on the implementation in engine. Heavily modded Skyrim and Fallout 3/4 should see some direct performance benefit by using ramdisk
Windows 11 hates Lossless Scaling
I agree. Too many factors to create a good experience. Its great if you know how the program works on technical level and what you need to have and do to actually use the damn thing on win 11 😄
For many confirugrations, win 11 will not be pleasant experience. It's expensive to have panel with MPO support. At least 900$, and it's mostly TV only.
Your argument is very interesting. Will it improve performance on my pc?
My pc is i7-6700k and gtx1070+rx6400+dr4 16g.
In this situation, would connect the extra 120g ssd and apply it as an L2 cache increase the fps in the game?
Only if your game is on a HDD if you want to use L2 cache
It also would just decrease loading times and stutters, not actually give more performance
Wouldn't it improve 0.1% fps, 1% fps in the game?
Does less lag in the game mean an improvement of 0.1 fps or 1% fps?
well yea
less stutters improve that
Very interesting. I just happen to have two spare ssd. Should I apply the raid option and designate it as an l2 cache?
its up to you
I would definitely do it
thats pretty much my setup
but instead of raid I just striped them in the windows partition manager
You mean Windows Disk Manager made two ssd into one?
I'm going to try it today. I'm wondering if there will be any changes in dual lsfg.
yes
Very interesting conversation about primo cache 🤔 I currently have 64GB of RAM and I am very interesting in trying out L1 read only cache, do you guys recommend a sweet spot number to test? I'm running a NVME 4.0 boot drive and another NVME 4.0 game drive.
64gb is a great amount
how much are you willing to use for cache?
if I had that amount, I'd atleast use 20gb
How much do you have (free)
good
I don't know.

So you have 16gb ram
Is my PC set up now?
ok
Did you give the cache to a hard drive?
yes my c:(nvme ssd) and d:(sata ssd)
So not hard drives
Then what you're doing doesn't really make sense
You're using an SSD as Cache for other SSDs.
Meaning you won't really get any speed improvements
It's either RAM for SSD or SSD for HDD
its better in some ways imo
That makes most sense
but ram is so much better for a system like that
But with 16gb ram not much to work with tbh
I'll test it in this situation.
And I'll install the 32gb RAM in the warehouse and try again.
Something's weird lol
Stuttering is the same but I think base fps has changed very stable. And I had a little pause when I needed to load during the game and this issue is gone too.
well I didn't expect it to work that well considering its already a sata and nvme drive
but if it works better then great
compare them
wow
your numbers seem close to my cache speeds
yea it doesn't seem to do anything for your case
if you want
remove the l2 cache and instead make a 2gb ram cache instead
see the speeds then
you will see how fast it can go
l2 cache doesn't seem to cut it for you
alright
I will shut down the power of my computer lol
📈 🔜
I installed the ram. What should I do now??
try 32 or 64
64 sweet spot I think
Yea
ok
I think the game's loading is more than twice as fast.

Oh, now that I see it, cpu overclocking isn't applied. I need to re-apply it and come back. Geez..
It's so funny. The fps have increased.
Stuttering isn't gone, but the base frame has increased. And it's stable.
I wanted to solve the stutter, but the base frame increased.
I haven't looked at the framerate tbh, but stutter hell now turned to Smooth as butter in some games for me
oopsie 
^ monitor with RX 6400 connected, fixes itself after I turn off and on the monitor
avg dual gpu experience
card is on warranty til September anyway
This happened after my Windows reinstall, got a BSoD and hasn't happened since lmao 
for me it doesn't even impact performance let alone stability
Yeah, mine also just started artifacting this way and after a few minutes I got a really corrupt looking blue screen 💀
Since then nothing has happened
I'm using Radeon for monitor output. What additional settings do I need in the adrenaline app?
How do I set up G-Sync in lsfg settings?
In LS? Just turn on the toggle
hows this?
making sure i do things right, i followed a video on youtube and he said to do those settings for faster read and write
Why 512 block size?
And I wouldn't activate defer write when using L1 cache, dangerous thing
okay, i put it to 256 and no defer write. About the l2 cache options, can i put read/write in the middle? so it can do both? not just reading or writing
Defer write is the write option
But it can be dangerous when you have a power outage
so even if i put the l2 cache to write only, its the same thing as defer write even tho its not enabled
Idk much about L2 cache tbh, I only use L1.
But write with defer write disabled should be fine
i'll do 50/50 for reading and writing since i'm pretty sure i'll do both often
should i use keep l1 cache on fast shutdown? it seems useful
Uhh, I don't even know where that is lmao
so i set it up and transferred a file from my hdd to the desktop and it didnt speed anything up, it was at 120mb/s. then I tried the same thing again after and tha time it took 2 seconds lmao.
but now i did the same thing from my desktop to my hdd and it doesn't work, it stays at 120mb/s even if i just copied it? did I miss something somewhere ?
lmao
im assuming ill have to precompile caches again for all games ?
Is this primocache program supposed to help lossless?
Not specifically, just general load times and stutters
No
yep
at least it did it in stalker2 and cod for me
or it wasn't cod
def in s2
I should delete all the games' caches manually or is it just gonna do it automatically?
Its a cache do it should populate itself as you use your machine.
First time you use a file/data it will be normal speed and copied to cache. Then next time it will read from cache and be much faster.
Its sucking every last drop out of my 10GB 💀
What happened when I was asleep 😭
World war 3
Please do not do this
😭
Why
I have RAM to spare anyway
Or should i do read only?
So 80% for reading and 20% for writing instead of 50/50. Should I do that for l1 and also l2 cache?
this for l1 cache
and this for l2 cache
Why not just leave it at default with the tick off?
It's off by default 
Also I've been playing a game for 3 hrs and had 450GB read and 10GB write
so that way its just automatic
Yes
do you really need write cache?
I transfer lots of files between lots of folders in different drives. and use them in DAWs fetching from different drives, I thought it'd help but I guess it can just be on automatic?
hmm
I'm not sure how safe it is without a ups
but if you are gonna do that, read cache should be 80% write 20%
im not using defer write option since it seems dangerous without a ups
then why are you splitting them?
yes, should i do that for l1 and the l2 cache then?
l1 should be like that
l2 should be even less
but you aren't using defer write
no point
so just default then
yep
Btw prefetch yes or no?
dumb question but when i create a cache task, i don't need to select the drive i use for the caching? would it be caching itself?
yeah that makes sense
The volume isn't even there.
You'd only select the 16mb local disk
This forums make no sense anymore 😂