---
title: 'ULTIMATE Local AI Quad 3090 Build'
source: 'https://youtube.com/watch?v=So7tqRSZ0s8'
video_id: 'So7tqRSZ0s8'
date: 2026-06-15
duration_sec: 0
---

# ULTIMATE Local AI Quad 3090 Build

> Source: [ULTIMATE Local AI Quad 3090 Build](https://youtube.com/watch?v=So7tqRSZ0s8)

## Summary

This guide details building a cost-effective quad RTX 3090 system for local AI inference, focusing on maximizing VRAM per dollar. It covers parts selection, assembly, power considerations, and benchmarks comparing Ollama and Llama.cpp on an AM5 platform, with an alternative AM4 build for savings.

### Key Points

- **Build Goal** [00:00] — Quad 3090 setup for LLM inference using cost-effective consumer desktop parts, optimizing spend on GPUs.
- **Motherboard Choice** [00:46] — Gigabyte B650 Eagle AX has four full-width PCIe slots (first at x16, rest at x1), ideal for inference at ~$150.
- **CPU and RAM** [01:24] — Ryzen 5 9600X ($190) for DDR5 entry; 64GB GSkill Trident Z5 DDR5-6000 with AMD Expo for easy tuning.
- **Power Delivery** [06:12] — Top slot delivers 75W, others less; use PCIe powered risers or set power limit to 175W for lower slots to avoid issues.
- **Software Compatibility** [07:59] — Ollama, LM Studio, Llama.cpp spread workload evenly; vLLM may need server-grade components for best performance.
- **Benchmark Results** [12:32] — Llama.cpp outperforms Ollama on prompt processing and text generation; e.g., GPT-OSS 12B: 1785 vs 1022 t/s prompt, 102 vs 125 gen t/s.
- **Power Consumption** [15:07] — Idle ~150W (higher than server due to CPU cooler); water cooling recommended for noise reduction.
- **AM4 Alternative** [18:36] — B550 board with five full-width slots ($99) allows up to 5 GPUs, saving on RAM/CPU if upgrading from AM4/DDR4.
- **Cost per VRAM Analysis** [22:17] — Quad 3090: $3,650 total, $38.02/GB VRAM (96GB). Quad 3060 12GB: $1,550, $32.29/GB. Always optimize for total VRAM.
- **5-GPU Option** [28:54] — Five 3060 12GB: $1,775, $29.58/GB (60GB VRAM). Cheaper than server builds due to high DDR4 ECC prices.

### Conclusion

This quad 3090 build offers excellent VRAM per dollar for local LLM inference, with Llama.cpp providing better performance than Ollama. The AM4 alternative further reduces costs, making it a compelling option compared to server setups.

## Transcript

Looking to build a local AI powerhouse
but don't know where to start? Then this
is going to be a great guide for you as
we put together a quad GPU setup today
that really does have some excellent
performance and is using some great
cost-effective consumer desktop parts.
Especially if you're looking at running
multiple GPUs, you want to optimize what
you're spending so you spend the most
money on GPUs and not the rest of the
system. If you're looking for LLM
inference, this is going to be a great
build for you. We're going to be able to
run four 3090s on our AM5 platform. I'm
going to show you some benchmarks. We're
going to talk about the parts and
components, and I'm also going to
present you an alternative if you
already have an AM4 system and some DDR4
laying around that can save you a ton of
money. Let's get started. The system
we're putting together is going to be
based off of a B650 Eagle AX. And if you
are a owner of a AM4 platform, I'm also
going to show you the B550, a really
good option to save some cost.
Additionally, of course, we're going to
be using our Quad 3090s. These still are
for the performance you get some of the
best bang for the buck that you can get
out there. We're going to look at all
this on a cost per gigabyte sheet, and
that will help illustrate that pretty
well. As far as the rest of the
components, I picked up during a Prime
special, a Samsung EVO Plus 990. That's
a 1 TB NVME we'll be throwing in here.
We've also got our AMD 9600X.
One of the cheapest ways to get into the
DDR5 class systems. The RAM that we're
going to be using is some GSkill Trident
Z5. All of this is going to be going
into our GPU rig frame. The first thing
that we're going to do is check out the
motherboard. This is the Gigabyte B650
Eagle AX. And the reason this
motherboard is one to consider strongly
is the fact that it has
four PCIe full width slots. That's
pretty cool and very unusual in a AMD
desktop system. Now only the first one
is going to operate at the full x16. The
rest are going to operate at X1 for
inference work. That's fine. And this
also allows you to have a lead GPU. And
these GPUs here would not be useful for
something like doing video generation
where you need full bandwidth, but this
one could still perform very well.
You've got pretty much your standard
everything else on it. It's not really a
super fancy motherboard, but in the
150ish price range, it is also
rather affordable. And we're going to be
populating this with our Ryzen 5 9600X
for about $190 and $195 does allow us to
get all the benefits that you're going
to get with your 9000 series and one of
the cheapest prices that you can
Next, we're going to go ahead and pull
out our Samsung 990 EVO Plus.
Now, this will negotiate 5 at 2 or 4x4.
It is a cheap, but for our purposes,
it's going to be just fine in VME.
So, the RAM that I've got here is some
Gskill. This is 6000 speed, DDR5, and
it's 64 GB of it. Now,
I don't think that it's wildly important
to get the fastest RAM out there. I
would recommend not spending money
trying to do that. Instead, go for
volume, always, and your RAM. Make sure
it does have AMD Expo, though, on it so
that you can just really quickly click
click and have everything tuned and set
so you can get optimal performance. And
I've got links to all this stuff on the
website with the written article with
highresolution photography of all of
these parts and components which should
help you if you are putting this
together. And so we want to populate
this in A2 and B2 for the slots.
And we're going to put just a little
P-s size drop right in the center. So,
this is a good piece of information for
you. If you're putting a desktop
motherboard, which has the CPU kind of
centered a little bit further this
direction into this frame, it's going to
absolutely be different than if you're
using a gigantic server motherboard like
this, like we had used for the quad rig
in the past, which had all the PCIe
separated much further down this way and
the CPU situated up this way. So, if you
put this in, you can see that yeah, a
tower CPU like an SP3 cooler could go
here and blow out that direction. But
for this, we needed a low profile
cooler. I've got this one temporarily in
here. It's going to be fine for now.
It's a 65 W CPU, so it's not really
going to generate a ton of heat. And
I've got a ordered one that I'll be
replacing this one with. You can find
the link to that in the description
below. All right.
And there we go. And it's good to have a
little extra power switch here.
Since we're not in a real case, this
will make it easy to turn it on and off.
And both of those I'll gather up and put
underneath there like that. When you're
looking at your PCIe power delivery,
this is an important consideration when
you're using a desktop component. This
motherboard delivers the full 75 watts
to this top slot, but it does not
deliver the full 75 watts to the
remaining three slots. So, this can
cause a problem for the power delivery
to really powerful GPUs like 3090s.
There are some ways around this. One of
those ways is to use PCIe powered
risers. Now, this is an option that if
you do end up going that route, you
probably want to have high-owered GPUs
in the first place to necessitate it.
And you also want to get ones that have
six pin or moax power. And you do not
want to get the ones that have the SATA
power to them as well. You can also
overcome this by just using some
traditional methods that are going to
shard the model for the LLM across the
GPUs and distribute the workload evenly
amongst the GPUs. This if you are
looking at a quad GPU setup like this
effectively means that you get about a
quarter of the full utilization. So
you're talking about 350 W GPUs. Each
one of those only able to run at a
quarter its performance. This does not
hold true in certain LLM runners like
VLM which are built to maximize every
last little ounce of performance out of
a system with great diminishing returns
as you get approaching the edges. So if
you are thinking of using VLM, you
definitely would probably want to
consider you're going to want a lot of
RAM to augment your system and you're
also going to definitely probably want
to consider going towards a server grade
component. If you are however only
looking at O Lama, LM Studio and Llama
C++, this is a great option for you.
They will actually spread out the
workload very evenly and this can be
overcome with a generous power level
applied. So for our GPUs here, if we
apply a power limit of about 175 on the
remaining three slots that are the lower
ones, we can overcome any power outage
potential. This also does indicate that
if you are looking for a really good
cost performance ratio, finding GPUs
that are in the 175 watt category,
things like a 5060Ti could be a great
option as well. So for inference
workloads, you won't be pushing your
GPUs at 100% power utilization. Whatever
the number of GPUs you have is, it'll
actually usually be divided by that
number. So in this instance, divided by
four. And if you look at the 350 watt
TDP that these typically have, it's
going to be if you set it to 225 for a
power limit or 200, perfectly fine to
run off of a single cable. I would
definitely say if you want to run
multiple GPUs that are high power and
you want to do things like image
generation, training, or other tasks on
them, make sure you have two independent
connectors going to each one of these.
But if you're just doing inference like
we're going to set this up for, you
don't necessarily have to worry about
that because like I mentioned, the power
level will be divided by however many
GPUs you've got. Let's get these wired
up.
So, I've got really high quality risers
here. These would work for most of the
use cases that you have out there. They
are gen fourspec. There are of course
gen 5-spec ones out there as well, but I
would urge you to if you are looking at
PCIe 3x1, consider that you could go
with much lesser risers in that scenario
and be okay. So, for these it'll be
fine. I'm just using them because I've
got them here handy and they're going to
demonstrate really well. But definitely
check the description below for some
links to some ideas for different ones
that are significantly less expensive
than these. These are pretty expensive
still. So, let's get these separated
out. But I got two long ones and two
short ones. And so we're going to have
short one, short one, long one, long one
as far as the arrangement. So we'll get
that first GPU plugged up here.
looking nice.
Very nice. Very nice. And you do have
two M.2s. Now, sometimes these are
shared with some of the PCIe slots, so
I'm not sure if these are going to stay
functional. Really, all that we've got
left to do is power it up and get the
system installed, do some benchmarking,
and then we'll go through some of the
components. Like I mentioned, if you do
have AM4 and DDR4 laying around, you can
go substantially cheaper. We're also
going to get wattage on this while we're
running it so that we can have a pretty
good idea about what the operating
expenses look like.
And power on.
Okay. So, go ahead, hit yes. So, you can
see we've got our 64 gigs of RAM. We've
got our R5 9600X. Got our 4800. It's
reading as and that's because our X expo
is not set. With the Expo set here, we
should be okay. You can see here we've
got PCIe 4 at 16, at 1, at 1, and at 1.
If you haven't taken a chance yet,
follow along with the guides on getting
set up with Open Web UI, Olama, and
Llama C++. Let's jump in and start doing
some benchmarks. We have our Llama C++
benchmarks in and we're going to run our
Olama benchmarks on the system and this
will give us a really good idea of on
the same machine the difference in
performance between using O Lama and
Llama C++. And I urge you to follow
along with some of the guides on the
website to get yourself up and running
with at least Llama C++.
And we'll run that against GPTOSS 12B.
And my script here runs three
iterations. So you can get a pretty good
reading of that.
And we got 1785 prompt processing eval
speed tokens per second and 102.20
generation tokens per second. So, I've
already got these things filled in, and
I wanted to talk about the results of
this so that we could see really what
this looks like as far as a head-to-head
comparison on the same machine between
Olama and Llama C++. And I've got some
charts that I think will help illustrate
that pretty well. Click that really
quick. Sorry for the uh insane
brightness of this. So, our prop
processing up here, our text generation
down here, and our Olama Ryzen system
versus the Llama C++ on the same Ryzen
system. You can see that the prompt
processing is faster across the board.
Yet, there are some that are going to be
significantly faster. Uh, definitely if
you start looking at Quinn 3, it's a
little bit tighter. Also, Gemma 3, a
little bit tighter. But you can see that
definitely when you get to GPTOSS, there
are big gaps. So if you are interested
in running GPTOSS 20B or 12B, Llama C++
really will extra give you some extra
tokens per second. Now on text
generation, you can see the Quinn 3 over
here, the A3B 32B. Pretty good results
that we got there. 120 on the Ryzen rig
uh for Olama, but really good 132 on
Llama C++. Now I would say Gemma 3 very
dense, very hard to process. So going
fast with Gemma 3 is just very difficult
to do with most of the runtimes out
there and 26 tokens versus 26.5 tokens
on the token generation side. So, not a
huge difference there, not statistically
significant, but once again, you get
over to the GPT OSS's 136 to 178, like a
massive difference. And also 100 to 125.
Huge difference. So, if you're
interested in running the script that
I've got here, drop a comment down
below. And I'm going to put this up. I
vibe coded this. So, if it blows up, not
on me. Uh, but definitely I'm going to
release this somewhere so you guys can
run this also open source it. Open vibe
it. That should be what we call it. I
think there's a pretty good reason to go
with llama C++. I think I've made the
case here. Now, let's compare just one
other thing. So, comparing the Epic rig
with a Lama to Llama C++, which is of
course faster on the Ryzen rig. Now, you
can see these are huge spreads on the
prompt processing. And I would say don't
look at this one and infer anything from
it. This is actually I think not the
right amount of prompt processing
tokens. So I wanted to be consistent
between the two and I'll need to go back
and rerun this test uh with the tools
that I have created here. But definitely
on the text generation side I think you
actually have a very good result that is
accurate and 104 versus 132. Definitely
there we see the Ryzen rig beating out
the epic rig. Now, Llama C++ needs to be
head-to-head with Llama C++ on the
different rig. And as soon as I take
this apart, I will definitely be putting
that together. And I've got a huge rig
that I'm building that is like, this is
going to be crazy. Definitely make sure
you like and subscribe. 25.4 to 26.5.
Again, Gemma 3 27B. I use it all the
time. Definitely a very hard model to
run and process. GPT OSS 20B50ish.
50. Yeah. Yeah. This is huge. So
definitely again the Ryzen rig kills
because of that single thread
performance and that does give you a
distinct advantage over something like a
7702 and on the Lama Epic rig 92.8
versus 125. So, while you can get lots
of cheaper RAM and run bigger models,
which is actually a benefit in and of
itself that is not to be trivial, uh,
you know, it it's worth it while do it.
And also, you get all those great power
delivery features that you have with
wonderful, very well-gineered
motherboards versus desktop class. Uh,
you also at the same time don't get the
single thread performance that you get
with something like a Ryzen. So, it's
not really surprising to me to see this
system be quite a performer. And I
expected that because of the fact that
it has a very good single thread speed,
better than the 7702. And as we saw in
prior testing, it definitely has an
impact. So, the power utilization on
this system in a very untuned state
here, I've done literally nothing to try
to tame this is about 150 watts idle.
That's actually higher than the server
setup. That's interesting, right?
Especially since you've got a BMC and a
lot more sticks of RAM on there. If
you're looking at this fan, this uh CPU
cooler is going to be replaced. A much
better option than this is going to be
on the way. And it should lower the
noise. Hopefully lowers the temperatures
also so the fan doesn't have to run at
an audible level the entire time.
Certainly having a fan wall along the
back of it that is blowing air into it
so that the motherboard does stay nice
and chill isn't a bad option to do
either. I've got that set up with those
little Silent X orange kind of well
yellow fans. And so definitely this
isn't as quiet as when I had the water
cooler on just the server motherboard.
So I would say if you did have a water
cooler, the prior mounting guide that
I've got linked on this page definitely
should be something that you check out.
I've got some modifications you can make
that allow you to fit extra-lar GPUs and
also fit in a water cooler at the top.
Pretty cool, very silent, and it was
definitely a bit overkill, but keeping
the noise down is a big thing to me. So,
if it's a big thing to you, also, you
might want to consider that. If you're
looking at whether or not you would be
able to fit five cards onto here, maybe
you've asked yourself that. Let me show
you something that is a little bit
crazy. This is the B550. Now, I
mentioned this kind of early on. This is
the B650. This is a newer build. This is
DDR5. If you have DDR4, if you've got an
AM4 CPU, I've got an AM4 CPU that I'm
going to be putting onto this, and it's
got DDR4, and it's sitting over there
right now, but this is such a cool
board. I don't know how I missed this
board up until now, but this, as a
consumer board, might be one of the
coolest boards I've ever seen. Why?
because it has five freaking full wide
slots. Freaking awesome. Now, I'm going
to put a 5950X on it and it's going to
go into a server case. One of the really
cool things about this also is this
allows you to have the opportunity to
get really high performance networking
on a desktop class system. Something
that's actually pretty hard to do. Would
that take and necessitate one of the
slots, probably the top slot, if you
really wanted to go fast? Well, it
would. But even if you look at the speed
you could attain with a dedicated
network card on any one of these other
slots that are going to operate at a 1x,
you'd be able to easily run 10 gigabit
networking. Pretty cool for a $99
motherboard. So if you've got an AM4
system, this is probably a good way to
consider. It also saves a lot of cost on
new RAM and of course new CPU, but it's
also a cheaper motherboard of course at
$99 versus about $150. Not that huge of
a difference, but five slots, that
really is actually pretty cool. So, I
would say that one is definitely a
strong consideration if you have an AM4
system. Right now, if you had to go out
and buy DDR4, it's almost the same price
as DDR5. Like, the prices on RAM right
now are kind of crazy. So, that is kind
of me helping you. And as this video
ages, it probably isn't going to stay
that way more than like six months is my
guess. But probably the next six months
is going to be a tight period of time
for buying system RAM. So pretty cool
consideration, good alternatives. I like
to try to save people money. You can
find links to that on the digital
spaceport.com website. Links to that
article in the description below and
pinned comment. So, while I've got the
3090s in here, certainly this
motherboard, in my opinion, has kind of
a good use case for having a strong lead
GPU that would be good for doing things
like image generation, video generation.
Most of that's going to run best on just
a single GPU. So, having a 24 GB GPU, if
you can find one, is going to yield the
best that you can get. It will also give
you better quality because you need
really 24 GB to get the best out of what
is available to run. Honestly, you need
80. But getting a GPU with 80 GB of
VRAMm in it is insanely expensive, like
way more than all of this. Now, the
remaining three GPUs in a 4GPU kind of
setup, which again, that's how many GPUs
I've got cuz I've got a freakish amount
of GPUs, but you could definitely start
with just one or two in a very similar
system to this and grow it throughout
time. The AMD 960 XT16 GB and also the
5060Ti, a pretty decent option. We're
going to look at these in the comparison
sheet and what the prices would look
like for doing a quad rig setup of that
versus 3090s. And also I'm going to put
the 3060 12 GB in surprising one around
that. So I think you definitely want to
stay tuned. We're going to jump into
that sheet here and take a look at some
of the performance that you get for the
dollars that you spend. We'll start off
looking at the AM5 build. That is the
one that we put together here. And you
can see definitely the DDR 64 GB of DDR5
not cheap as well. You've got a bunch of
other components. And like I mentioned,
check out that AM4 build. Uh we'll get
to that sheet here after this. So, as of
the time of me putting this together,
we're looking at right around $1,156
just for the base system components.
Now, this gets interesting if you look
at a dual GPU and quad GPU option setup
for this rig. So, we'll start with the
3060s. And with two of those at about
225 each, you're going to have 24 GB of
VRAM. Total system cost of $166.
And your price per GB of VRAM is $66.92.
Might not seem bad, but just keep
watching. Now, as you move on to your 16
GB cards here, your third 960 XT 16 GB
and your 5060 Ti 16 GB, you do bring in
a bit more for the Nvidia. So, that's
$1,896
and $2,16 respectively, but the price
per gigabyte of VRAM is actually sub $60
here for the 9060 XT at $59.25.
The Nvidia is $63.
The surpriser for a lot of people might
be the cost per gigabyte of VRAM for 24
gigabyte GPUs. And the reason why is
their cost per gigabyte per GPU is low.
And that is because you see $750, that's
about $1,500 for two of them. 48 total
gigabytes of VRAM. And there is a lot of
additional benefits to having more and
more VRAM. So definitely consider this.
The system cost overall does go up to
26.56, but the price per gigabyte of
VRAM goes down to 5533.
And when you add additional GPUs on each
incremental add addition uh kind of
helps defay some of the cost. It's the
most optimal to get as many GPUs writing
on a single system as possible. So if we
go in the same order again, you'll see
that the totals for the VRAMm double. So
we're at 48,64 and 96 as well. The costs
go up and that is 256
all the way up to 4156.
However, the price per gigabyte of VRAM
drops quite a bit. So all the way down
to 4238,
4119,
44.97, and 4329.
So there is a lot of very good
considerations that you want to make if
you are looking the number of GPUs and
the price per gigabyte because the
gigabytes of VRAM is what you should
always optimize for. I get this question
definitely the most frequent question on
the channel. And yes, you always want to
go for more gigabytes of VRAM regardless
if you go for a single GPU. Always
optimize for as many gigabytes as
possible and also never go under uh 12.
really you don't want an 8 GB GPU. It's
it's not going to be a good performer
for you for the most part. Now, let's
move on to the I think kind of more
exciting option, which is you happen to
have an DDR4 and AM5 CPU laying around.
I think this is cool because it lowers
that cost. And I mean, you know, there's
a lot of these other components you
probably could have laying around also
is my guess. You're looking at like
$650.
That's pretty affordable, honestly. So,
if you especially if you just need the
motherboard, if you look at the cost per
gigabyte of VRAM, we'll look at dual
quad and since it actually has five
slots, we'll actually take a look at
five also. You you're looking at $45.83
for the 3060 all the way down to $4479
for the 3090. The 5060Ti
is 4719 and the 960 XT 16 GB setup with
two of those comes out to 4344.
So you edge out a bit on the cost per
gigabyte of VRAM by going with the 9060
XT. However, I think it's very close and
a very compelling argument to also
consider a 3090. The interesting one is
the 5060 is probably one of the weaker
card recommendations that you would be
looking at in most of these setups if
you wanted to scale out. I would also
say you're definitely with 442 GB of
performance limiting system bandwidth,
which that's always going to be the big
impactor. That's, you know, same as with
the DGX Spark, same with framework. I
mean, it the system bandwidth is always
going to be your dictator of where
you're going to run into slowdown. And I
mean the 5060 Ti 442 gigabytes per
second, that's actually pretty
respectable compared to what you get
with the 9060 XT, which I believe is 322
GB per second. Quite a big difference.
So as you add GPUs, your speed will not
increase the more GPUs you add unless
you are doing a specific type of
parallelism. But most of the time you
want to run the bigger models. So it's
going to shard the model across. And as
it does that, it's going to
unfortunately be limited to the
performance of the GPU that is the
slowest. So if you tossed a bunch of
3090s into a rig and then you tossed one
3060, you're going to be going pretty,
you know, you're not going to like
yourself because that 3060 is going to
be your performance dictator. So, I
would urge you to also consider that if
you get into weird mixtures of GPUs that
are too weird, sometimes it can have
negative impacts. Looking at the quad
GPU setup, you have total system costs
ranging from $1,550
for the four 3060s all the way up to
3650 for four 3090s. So, 96 GB of total
system VRAM, 3650. That's really I think
pretty compellingly. I mean that's
that's good. And your price per gigabyte
of VRAM there is 3229
3328
3703
and all the way up to $382.
If you're looking at your 5GPU option, I
mean, you know, why not? you definitely
can see that you, you know, run into 120
potentially gigabytes of total system
VRAM if you have 24 GB GPUs for $4,400
of cost. And that comes out to $36.67
price per gigabyte. And I think probably
an interesting one here also is the 3060
12 GB gets you all the way up to 60 GB
of total system VRAM. And at a cost of
$1,775,
that comes down to sub $30. That's
$29.58.
So a 4 + one good GPU setup could be
compelling. Uh those kind of things I
think are pretty interesting. And I
think this is cheaper by a pretty good
margin than what you can get for servers
right now. mainly because of the
insanity that is DDR4 ECC pricing which
has gone through the roof recently.
Yeah, definitely not going to get that
setup for 3650 looking closer to about
5200 525. Uh that's getting lucky with
some RAM also. So, when you're
considering alternatives, I would
definitely say this is a fairly decent
one to consider, especially given that
you have a variety of price points based
upon how much VRAM and which type of
GPUs you're putting in. You can also go
all the way back to Pascal. Technically,
you can go back to Maxwell, but both of
these card generations are at the cusp
or being retired from the driver
support. This isn't actually as bad as
it might sound, but definitely moving
forward, new features that come out may
not be supported on those cards. Those
GPUs also sometimes don't get very good
performance for the watts that they're
using. The Volulta generation cards are
just very hard to find. So, I don't
think you're going to have a great luck
finding any of those. They just
seemingly are not out there in very
significant numbers. Starting with the
Amper lineup is pretty much where I
think most people have the best it makes
sense to get in. So, that's my take on
it. And certainly do consider you don't
have to go 3090s. you could put a
different GPU in there just as easily.
And so, we've covered a lot in this
absolutely insane build, setup,
benchmark, and valuation deep dive into
building for newbies. So, I hope you've
enjoyed this. And I know it's a lot of
information. Don't feel like you can't
go back and hit rewatch. And thanks for
all the shares, thanks for all the
likes, and a huge shout out and very
much grateful to everybody who is a
channel member. Also, the people that
buy me a coffee. You guys really do make
all of this possible. Everybody have a
great day. Let me know what you think
and I will check you out next time.