Original subtitles

[Music]

today we're reviewing an amd

apu that doesn't come out officially for

another two months but it's been

floating around in the oem pre-built

market for a little while now and that

is the 5600g so it's an r5 apu

that means it has vega graphics

integrated it's not rdna

and it also means that it's minimally a

stop gap if you don't have a gpu

to potentially a reasonable solution for

certain types of gaming

so we're going to be reviewing that of

course we'll review the time machine we

use to get this later

but we'll start with the amd 5600g and

the 260 dollar price bracket for an apu

before that this video is brought to you

by keoxia said to be the original

inventors of nand flash memory and

formerly known as toshiba memory

kiouksia's xg6

m.2 ssd is a high-end nvme ssd claiming

up to 3180 megabytes per second read and

2960 megabytes per second write speeds

sequentially

and it's ideally used in high

performance gaming pcs

or power conscious laptops the xg6 is

available up to one terabyte

and the company also has its bg4 ssd

in a much smaller m.2 2230 form factor

for up to one terabyte of ultra compact

storage learn more about

keoxia at the links in the description

below including

about their enterprise storage products

so some basics for these these are

actually available

they are being sold in pre-built and if

you check amd's website for the 5600g

it actually lists the official launch

date as march 13th or so

of this year but the official launch

date to the

diy market as opposed to just the launch

date to the market in general

is august 5th so it's a little ways away

because this is a retail available

product we're under no obligations from

amd in terms of embargoes or anything

like that

and it also means that it's a finalized

product to some extent but

with a couple of months between when

these started shipping out to

actual users who bought pre-built

systems at retail

and the time at which the diy systems or

diy parts come out separately

it's possible that there are iterative

changes in drivers

or in the silicone quality or something

like that so we'll revisit these as

we get in the uh more official diy part

review samples

we don't suspect there to be any major

changes because again this is something

that you can actually buy it's not like

we went through back channels to get it

you can just buy it on the market openly

if you especially if you go through like

hp or someone who

carries it in larger quantities couple

of notes here if you go to try and buy

one of these standalone right now as

opposed to through

a pre-built system you're probably going

to pay scalper prices

and cpus are not being as heavily scaled

as gpus

we've been able at least in the u.s

market where we check to find 5 800 x's

5600 x's or some variant of a 10 series

intel cpu

in stock close to msrp or below it in

intel's cases a lot of time

pretty regularly now for a few months so

don't pay scalp

prices if you can avoid it for something

like an apu

because there are other parts out there

and some of them even do have igps in

them

so it's not quite as bad as the gpu

market uh we're just trying to make sure

people don't

rip themselves off by accident because

they see this review

they see one of these on ebay for 500

and they buy it you shouldn't do that

so these are supposed to be 260 parts

when they come out for the 5600g

a little more expensive for the 5700g

but there are normal cpus in stock as

well

again through retailers so the specs

page from amd says it's a six core 12

thread part that matches for the 56

naming like the 5600

x it says it has seven gpu cores or

compute units

cus by amd's radeon terminology it's a

65 watt tdp

it has an adjustable tdp down to 45

watts and it runs a 3.9 to 4.4 gigahertz

base or

boost for limited core boosting there's

a 1900 megahertz

gpu spec and we'll check on that in

benchmarking official memory support

like the rest of zen three and this

is zen three by the way it's seven

nanometers is 3200 megahertz you can

overclock as desired

we tested with our normal c14 3200

megahertz with tuning

uh trident zrgb kit as a reminder we've

found that this

setup that we use for standardized

testing on everything

runs faster than a lot of the 3600 kits

so this is all standardized that means

you can compare everything to each other

from one igp from intel to amd's apus

or between the cpu and dgpu combinations

if there's enough interest

and please let us know below we might

look into

a separate memory piece with the apu

specifically the reason that's been

interesting in the past

is pretty simple it's because there's no

onboard memory because it's not a video

card

and so that means when the cus in this

need to get something out of memory

they're traveling to system memory if

they can't find it

in a closer cache and traveling to

system memory is

it's a further journey and uh the memory

is

not particularly fast when you compare

against say

gddr6x on a pcb that contains

the substrate for the gpu in a dgpu

solution in speaking with amd though

previously the

memory impact on these apus it sounds

like shouldn't be as severe as we've

seen

sometimes in the past in fact amd's

charts that it presented during its

keynote whenever it announced them were

using 3200 megahertz memory but as a

reminder i was just tuned the most maybe

obvious competition with this would be

a cheap cpu and dgpu that's normally the

case the market is screwy now

so it throws the pricing off but for the

last week while we've been working on

this we've been consistently able to

find an intel i3 10 100

in stock for about 125 to 140 depending

and a gt 1030 gddr5 version for about

120 bucks

so that puts us in the range of 240 to

maybe 260

for the cpu plus dgpu cheap combination

versus the 260 apu that would be the

best comparison because

in the past what we've been looking at

in a normal market is

very simple it's if you have a really

strict budget

do you buy an apu for something like 120

140 bucks

or do you buy a cheap cpu and d gpu that

might be like a pentium

g 3568 or whatever it's called

or it might be something like a cheap r3

cpu

and some low end gpu often it would be a

710 or a 1030 something in that class

don't buy a ddr5 1030 by the way now

there's a slightly different angle where

aps are also a stop gap solution

if you need something to run anything

while you wait for

a more permanent gpu and so with the

5600g

you end up with a cpu that's relatively

powerful so that it doesn't become

useless when you do eventually upgrade

it to a dedicated gpu you still have a

six core 12 thread

zen three ryzen part in there even if

the igp becomes worthless at that point

let's get into some of the testing as a

reminder

this is an official product that means

it is all working and it's something you

can buy

through oams but it's also not

officially out through diy

very strange situation so we will be

revisiting this as things iterate

and move towards diy launch let's start

with some validation testing

before we get started with the testing

some notes on the setup

first off not every board will support

these apus you'll obviously need

on-board display out of some kind

usually hdmi in order to use the igp

but other than that it's pretty standard

you'll also need a bios update to

support the 5600g and 5700g our gigabyte

x570 master for example was unable to

boot with the 5600g

so we had to switch over to a different

board it's the gigabyte b550 aorus pro

this isn't a big deal

just if you're planning to buy a

pre-built system and rip the apu out of

it

check the cpu support list on the board

you're intending to use before you

commit to buying both of those things

even if you intend on using a normal

video card and never using integrated

graphics

you'll need to make sure the board is

compatible bios is mostly the same it's

the usual amd overclocking options

and gigabyte bios combination with the

key difference being

an amd overclocking sub page for

integrated graphics overclocking

this menu isn't very fleshed out it has

two entries

and that's it but as this is ahead of

the official diy launch

it's possible that this expands with

other boards in the future

the options are for graphics frequency

and graphics core voltage

both of which can be entered manually

graphics frequency needs to be adjusted

with the voltage for stability of course

and gigabytes bios notes that graphics p

states remain active when doing this

bioalso also notes that the graphics

voltage is derived from the soc

or the uncore voltage rails we may try

overclocking for a separate video but

for now we'll be testing stock since

that's how most people use these

installing the amd driver package

includes the same software suite

as what you'd get with an amd gpu it

looks the same it's just got different

options in it

currently the drivers show the gpu

frequency the temperature

some cpu stats some specs fps and frame

times

but there's no options for tuning in our

driver install at least and there aren't

really

any options for input in general you

have to do all that in bios if you do

any of it this chart plots the average

all core frequency of the 5600x and the

5600g

useful for establishing how useful the

5600g

is as just a cpu the r55600g ends up

plotting an average all core frequency

during

blender of about 4 300 megahertz with

the 5600

x at 4 400 megahertz the 5600x will

outperform the 5600g when ignoring the

presence of the igp

and you can see the 5600x finishes the

work first as well that's why the line

ends sooner

so this emphasizes that point in single

core testing with cinebench we monitored

the r5 to 600g is at maximum single core

frequency per interval

at a fairly flat 4440 megahertz compared

to the 4650 megahertz on the r5 5600x

the 5600x unsurprisingly is faster when

used only as a cpu counter-strike global

offensive starts us off for the game

benchmarks in our new igp test suite

we're testing at 1080p medium settings

with a pretty wide mix of parts here

and remember that some of these

combinations are designed more for

academic understanding

of where the bottlenecks are rather than

for practical use

the r55600g does absolutely here we can

refer to our pairing of the gt 1030 with

an i9 10900k which obviously doesn't

make sense but

it's helpful for understanding the

bottleneck it's the gpu

the gt1030 simply is not as powerful as

the r55600g

in this instance despite running an

older vega architecture

on amd's part the 260 dollar apu

leads the similarly priced 10 100 plus

gt 1030

g5 pairing by about 12 that's impressive

performance for andy's igp we don't

normally see them

match the same price to dgpu and cpu

pairing as well as they're doing here

and of course the gpu shortage isn't

doing the dgpu option any favors

the iris xe g7 combination in the intel

nok

ran at 81 fps average allowing amd elite

of 47

intel's uhd graphics 750 in the 11600 k

ran at 60 fps average

showing a gpu bind and allowed a

significant lead in the 5600g

as for the gt 1030 ddr4 you still

shouldn't buy it

just like a few years ago but we can see

it is better in the one percent low

consistency

than intel's uhd igps fortnite is next

run at 1080p low for the widest support

for comparisons

the amd apu again leads this particular

set of data

of course a higher-end dgpu like a gtx

1650

would leave the radeon 7 apu but

we're then outside of a direct price

match and it's not really the same type

of comparison

the apu is roughly equal to the gt 1030

gddr5 card

which is bottlenecking the 10900k and

11600k

cpus the 10100 illustrates this dropping

only a few frames from the higher end to

cpu pairings

intel's integrated uhd 750 and the 630

igps

aren't even close to the gt 1030 gddr5

or the 5600g

you wouldn't get a good experience on

those uhdi gps

but for the gt 1030 g5 or the 5600g

they're completely playable and they

even allow a little bit of extra room

for higher settings

fortnite's performance mode is even

easier to run and expectedly

it looks even worse but that's the

exchange for a higher frame rate

this tells us something interesting that

in this task we've become

cpu bound rather than gpu bound as we

saw in the first

three charts the r55600g therefore falls

below the gt 1030 performance when

combined with the higher performing cpus

illustrating a potential weak spot for

the apu the intel i3 10 100 and the gt

1030 also maintain an edge gear

at a 7.8 percent advantage over the

r55600g for average fps

this tells us that the r55600g won't

always lead the i3 and 10 30

g5 pairing that is although the ddr4

version

you should ignore in rocket league the

gt 1030 gddr5 and the i3 10 100

outperform the r55600g by about

60 68 here forming a massive rift

that we didn't see in some of the other

games we'll need to analyze the hardware

activity to better understand

why this happens we retested and found

the same results

so the next thing we did was pair a gt

1030 with the r5 5600g and then disable

its igp

to find the root cause of this

difference it was the gpu

the gt 1030 and the r55600g which is

a pairing that you shouldn't do it

doesn't really make sense financially

but

that propelled to 136 fps average that's

a spike of 65

from the vega gpu only in this instance

the amd apu is weaker than a cheap cpu

and d gpu assuming you can get them both

but better than the 11th gen igps by a

wide margin running hardware info

logging during

the same rocket league gameplay helps

explain what's going on you can see the

cpu utilization is only sitting around

ten percent here

a lot of people think this is a good

thing because they think it means they

have more cpu threat availability for

other tasks

in fact utilization this low just means

we have wasted resources incapable of

serving a benefit

for their primary current task the

radeon

gpu is fully loaded here and is clearly

the bottleneck

something has to be a bottleneck in

every system so that's normal

but it's balancing the two parts as

close together as possible

that helps get that better performance

and in this instance the gpu is so

heavily bound

that we're not getting a lot of benefit

out of the cpu minecraft isn't much of a

benchmark but

it and games like it are a great fit for

these types of builds

the i3 10 100 and gt 1030 do end up

ahead running at 115 fps average so the

r5 3600g is 100 fps average

for a total of 14.6 advantaged

the iris xe g7 from the intel nook the

new one is next

with the uhd 750 on the 11 600k

down at 75 fps average that's definitely

playable

if you want to stop gap while you're

waiting for a gpu and

you already want this cpu but if you're

not planning to

use this thing other than for a

temporary solution amd's igp holds a

significant 53

advantage and obviously going with the

dgpu kind of makes both of these

irrelevant

but that's a fantasy land at this point

because they're hard to get

the gt 1030 is actually struggling to

fight with amd's apu here

and that's the part that's a little bit

different from some of the other ones

we've looked

at in the past transitioning briefly to

cpu only production benchmarks will be

our next

set of charts in blender all core

rendering the amd r55600g required 25

minutes to complete

which has it about on par with an amdr

six core 12 thread part the i5 11 600k

outperformed the 5600g by a few minutes

with the r5 3600x requiring 11 percent

less time to render

than the 5600g adobe premiere aggregate

scoring has the r5 5600g at 695 points

allowing the 5600x an

overall performance lead of about 6.2

percent and cpu bound tasks like

rendering and scrubbing

keep in mind that this is tested with

the same rtx 3080 in all configurations

on the same ram so we're looking at

cpu differences here the 5600g is more

similar to the 3600 or the 3700x

than to the modern 5600x in performance

intel's 10 700k

leads at 740 points roughly tied with

the 5600x

with its 11 600k closer to the 5600g

the 5600g is reasonably close to 5600x

if you can't find one in stock

but then the 3600 isn't much different

either the cheaper of these two wins if

you only need a cpu in chromium code

compile the r55600g required 147 minutes

an hour benchmark

versus the 5600x is 131 minutes

that's back to the 11 reduction time

required that we saw in blender

the 3600 takes about six and a half

percent less time

with the 11 600k leading more

significantly at a 16

reduction the 5600g leads the 11400 by

a similar 16 reduction we have other

production benchmarks complete but

they're all about the same as these

that includes photoshop 7-zip handbrake

and others

but the pattern persists moving back to

games we'll look at how the 5600g does

when

all of the cpus are tested and paired

with the same hardware

including the gpu so no igp's allowed

the 5600g

really isn't meant to be used this way

but if you were to buy one now as maybe

a stop gap and add a gpu later

this gives you a better idea for the

maximum that the cpu

will allow out of the gpu in rainbow six

siege at 1080p

the r55600g matched with the rtx 3080

ends up at 393 fps average with lows

slightly out of proportion versus the

immediate

r7 3700x and i5 11400 neighbors but

overall they're similar

the r55600x leads the 5600g by 16

percent in this one

the i5 11 400 and 10 600k are both

roughly six percent ahead the r5 3600

allows the 5600g a slight lead but

if you can find a 3600 for significantly

cheaper than a 5600g

and you don't need its igp you'd get

mostly the same performance the 3600

being older and without an igp at all

does have different disadvantages though

at 1440p the amd cpu slam into a wall of

about 340 fps average with the nvidia

gpu

where we see everything from the 5600x

to the 5950x with an average fps error

due to

a bottleneck the r55600g is able to

catch up here

where the gpu bind reduces the 5600x's

advantage to 5

that's the nature of a gpu bind the cpu

obviously doesn't matter as much

cyberpunk is of course more relevant

than ever thanks to tesla's use of it on

stage

although we don't have a test the model

s plaid listed in this benchmark charge

we can at least show the 5600g the 5600g

ran at 132fps average with those

proportional to that

that places it ahead of the 3700x by

about seven and a half percent

with the 11400 leading the 5600g by 6.7

percent

and the 5600x leading the 5600 ge by

16.5

that's similar to what we saw in rainbow

six at 1080p in red that too the 5600g

ran at 147 fps average with lows at 79

and 68 fps

that allows the 5600x a lead in all

three metrics

and about 11 on average that has the

5600g outdone by the 11400

but it's out doing the r5 3600 and the

r7 3700x marginally

f1 2020 runs a comically high frame rate

when cpu bound much like in rainbow six

and has the 5600g at 254 fps average

with lows around 100 fps at the worst

for the 0.1 percent

this is obviously completely playable in

terms of a relative comparison the 5600x

benefits from a lead of 26

more noteworthy than some of these other

games while the 10600k

ends up closer to parity with the 5600g

the 11400 just for reference of

something kind of on the cheaper end for

intel

is also about equal to the 5600g right

now although it's a little long on the

tooth gta 5

is useful for understanding how cpus

treat an older style build of game

especially dx11

with limited use beyond four threads gta

5

is definitely that game the r55600g

ran at 115 fps average establishing

a 17.8 percent advantage for the 5600

x and about the same for the 11 600k

the 115 fps average result is also near

the intel i5 10 600k

stock result power consumption is up now

this is tested at the eps 12 volts

cables so you're looking at the numbers

before the vrm efficiency losses

but after the wall that means we don't

have any measurement of the gpu power

consumption for this set of tests

but we have a special set of power

consumption tests we'll look at next

this is with the rtx 3080 running so

it's entirely cpu bound

in a blender cpu only workload the

r55600g pulled about 72 watts in the

b550 pro board

that's close to the 67 watts we observed

in the x570 master since the motherboard

had to change by necessity to support

the apu

we're not fully like for like between

the 5600x and the jeep

but we can call these functionally the

same the 11400

68 watts for comparison and the 3600 was

closer to 80 watts

remember that the igp again isn't doing

much of anything here in cinebench r20

tested within tau for intel

so showing higher power consumption

where applicable we see the 5600g at 77

watts

for the uh comparison of 66 with the

5600

x this was only a 5 watt or so increase

over blender for the 5600g

and it was repeatable but it's also a

different application and it's not that

crazy of a difference so this all makes

sense for single core rendering and

cinebench r20

the 5600g pulled about 16 watts while

the 5600 x was closer to 22 to 23 watts

power consumption in a fully cpu

constrained single core scenario was

exceptionally low for the 5600g here we

mixed in a new one-off test for the apus

as well

in this one we're measuring mixed

workload power consumption with gaming

this is rather than using the cpu

targeted power draw with a d gpu

previously

the r55600g pulled between 32 and 40

watts at the eps 12 volt cables

that's when in a workload combining both

cpu and gpu utilization

on the same chip of course and in the

more gpu bound conditions like rocket

league or gta

at 1440 we saw lower or slightly lower

power consumption than in the more cpu

bound conditions

because none of these are 100 workloads

on the cpu

given that the igp is doing work here we

don't see power consumption as high as

the previous cpu only charts

the gpu appears to be the more limiting

of the two components

and again something has to limit you

eventually

it just happens to be the igp in this

coupling of the two

just as some final validation we'll look

at the igp frequency to check whether

the

software reads it accurately and whether

everything is working as it should

and this 5600g spec list the gpu clock

at 1900 megahertz

and looking at our testing during some

rocket league gameplay we landed at

exactly 1900.

it didn't deviate at all at least

according to the software and it stayed

at flat 1900 whenever under load so

that's it for the rfi 5600g ap

review if you're looking at buying this

thing either in the future

first of all check back but either in

the future is a standalone solution or

an oem pre-built right now again you can

find them on

hp for example pretty easily at least we

were able to then

the 5600g is a relatively capable cpu

but

it's not quite as powerful as the 5600x

and a lot of the cpu only tasks like in

things such as blender

premiere where we're testing for cpu

performance

and ignoring the presence of the apu the

igp in those tasks

we end up with a cpu that's roughly in

the range of say an r5 3600

sometimes the 10 600k depending on what

application you're looking at

so it is definitely not a drop in

replacement for a 5600x

runs a little bit slower maybe that'll

change in the future but

as it exists now that's where it lands

and the cpu

isn't really a thing that's going to

change with drivers zen 3's

fairly mature at this point so it's the

bios and the frequencies are running at

amd spec so we really don't suspect the

cpu performance will move much at all

and so you'll end up where if you just

want a cpu and you want a six core 12

thread one and you can find a 5600x

you should just probably just buy that

instead but if you need the

igp component that's where it gets more

interesting so as

a combined unit you end up with

performance roughly

in the range in some games of a 10 100

plus a gt 1030 gddr5

not ddr4 version uh or sometimes behind

it and it does depend on the game

the gpu as we found in the testing today

the

cus in here from vega are definitely the

weakest point of it

the cpu does well and it's the igp that

ends up bottlenecking if you bottleneck

in gaming there are some breakouts from

that and extremely cpu bound games for

the most part

it's an igp limitation so then the apus

end up being something that

you can actually use these days to play

games at reasonable settings

it's once you start getting into higher

end stuff

we call it aaa typically where they'll

start to struggle

and you're going to have a hard time

running graphic settings that are

good looking in games like say red dead

2 or

cyberpunk 2077 i'll be everything

struggles with that game

but you'll have a little more trouble

running the the higher quality games

with higher quality graphics

and that's the nature of something that

is both a cpu and a gpu

in this package size can't expect the

world from it

but these are far better than they used

to be and it's something you could

actually use to play games now

whether or not it makes sense it's going

to depend on the price we buy it with an

oem system well that's

we can't really judge that because those

parts are all completely variable and

there's

at least seven eight other components in

there so you'll need to evaluate that

system by system

as a standalone part it's not worth the

scalper prices do not pay that money for

it you'd be better off getting

an intel or an amd cpu standalone or an

intel one if you need the igp as

a stop gap but the igp is much worse

than this or

you'd be better off waiting for this to

launch the diy or buying the pre-built

that's another option too so that's it

for this one thanks for watching as

always

subscribe for more check back for

follow-ups on this there's a lot of

interesting stuff we can do with apus

we don't have a really firm conclusion

on this as a diy part right now just

because you

can't officially buy it that way so

we'll reevaluate

as amd perhaps for the diy launch but as

always check back for more you can go to

store.gamersnexus.net if you'd like to

grab

some of our shirts like these pc

shortage t-shirts

or you can go to patreon.comgamersnexus

to get behind the scenes videos thanks

for watching we'll see you all next time