WEBVTT

00:00:00.060 --> 00:00:04.440
have you ever taken your shiny new graphics card home only to be

00:00:03.120 --> 00:00:10.679
disappointed at how you can't turn the settings up as far as you want in some hyped up AAA game it turns out that in

00:00:08.220 --> 00:00:14.340
many cases this isn't because the GPU is underpowered instead you might not have

00:00:12.599 --> 00:00:19.320
enough video memory or the amount you do have might just be too slow but the next

00:00:16.320 --> 00:00:21.060
generation of vram gddr7 promises to

00:00:19.320 --> 00:00:25.140
help alleviate some of these issues with never before seen bandwidth this is a

00:00:23.699 --> 00:00:29.039
big deal because of the rapidly increasing demands being put on gaming

00:00:27.119 --> 00:00:33.180
PCs where the high resolution textures and complex visual effects of big name

00:00:31.080 --> 00:00:36.960
titles can overload all but the highest End Hardware meaning you need enough

00:00:34.800 --> 00:00:40.620
high-speed vram to hold all that data that the GPU is being asked to process

00:00:38.520 --> 00:00:45.000
in fact one of the more common gripes from folks who bought the RTX 3070 on

00:00:42.899 --> 00:00:48.480
upper tier GPU was that it only came with eight gigabytes of vram resulting

00:00:47.219 --> 00:00:52.920
in performance issues and graphical glitches and games the GPU itself could

00:00:50.460 --> 00:00:56.579
probably have handled enter gddr7 which is expected to bring us a huge bump in

00:00:54.960 --> 00:00:59.760
memory bandwidth without sucking down more power which is great news for those

00:00:58.620 --> 00:01:03.719
who just want to turn up the settings but are already fighting the rising

00:01:01.800 --> 00:01:09.900
power draws of neurogen graphics cards we're talking 36 gigabits per second per

00:01:07.560 --> 00:01:16.500
pin once you translate this into a card with a 256-bit wide memory bus that is a

00:01:12.840 --> 00:01:18.060
full 1.15 terabytes per second of

00:01:16.500 --> 00:01:22.200
throughput to put that into perspective the RTX 4080 only has around 700

00:01:20.400 --> 00:01:25.500
gigabytes per second of every bandwidth so all that extra space in the pipe

00:01:23.820 --> 00:01:30.780
could help demanding titles even without adding more actual vram but what exactly

00:01:28.200 --> 00:01:33.720
is the secret sauce that allows gddr7 to reach these new heights we'll tell you

00:01:32.400 --> 00:01:39.000
right after we think of videocom struggling to make your presentations engaging video com can help with that

00:01:37.320 --> 00:01:43.619
with a bunch of features to spruce up your projects such as per slide webcam

00:01:41.100 --> 00:01:47.460
embedding automatic Cloud saving and team collaboration tools making a

00:01:45.659 --> 00:01:51.360
presentation has never been more easy and intuitive turn that drab PowerPoint

00:01:49.619 --> 00:01:54.479
presentation into an experience that will be sure to leave an impression with

00:01:52.799 --> 00:01:59.340
your audience check out Videocon at the link below for free and use code Linus

00:01:56.159 --> 00:02:02.040
for 50 off any paid subscription gdr7

00:01:59.340 --> 00:02:05.520
uses a little trick called Pam three and that's not the stuff you spray on your

00:02:03.479 --> 00:02:10.500
pan so the tater tots don't stick it's pulse amplitude modulation in Pam 3 each

00:02:08.640 --> 00:02:15.239
clock cycle can have three voltage States positive one negative one or zero

00:02:13.020 --> 00:02:19.500
each sequence of two cycles encodes three bits of data for example of

00:02:17.700 --> 00:02:25.319
negative one followed by a positive one corresponds to the bits 0 1 0. in this

00:02:22.200 --> 00:02:27.480
way each clock cycle encodes 1.5 bits of

00:02:25.319 --> 00:02:30.780
data instead of just a single bit but you're a Savvy viewer and maybe you

00:02:29.220 --> 00:02:36.900
already know that the current top end video memory gddr6x actually supports

00:02:33.480 --> 00:02:39.720
pam4 I mean isn't that better four is a

00:02:36.900 --> 00:02:45.540
bigger number than three I think maybe well it's only better to a point Pam 4's

00:02:43.500 --> 00:02:50.340
encoding scheme actually delivers two bits per cycle making it faster than Pam

00:02:47.940 --> 00:02:54.540
3 on paper but there are some drawbacks four possible States per cycle means

00:02:52.560 --> 00:02:58.140
that there's a greater chance for errors in data transmission and to compensate

00:02:56.340 --> 00:03:02.160
for this you need more error correction functionality as well as a new physics

00:02:59.879 --> 00:03:06.300
physical design and different memory controllers all of these things increase

00:03:03.900 --> 00:03:10.800
both the cost and the power consumption not to mention that the gddr6x can only

00:03:08.760 --> 00:03:15.840
operate in pan4 mode when data is being sent in eight byte bursts instead of the

00:03:12.959 --> 00:03:19.800
more standard 16 bytes meaning gdpr6x can actually have less bandwidth overall

00:03:17.819 --> 00:03:23.580
in real world scenarios put all this together and gddr7 is supposed to give

00:03:21.959 --> 00:03:27.780
us more bandwidth while being more energy efficient than gdr6x not to

00:03:26.040 --> 00:03:32.340
mention it can switch off Pam 3 when it's not needed and instead use the one

00:03:29.640 --> 00:03:35.940
bit per cycle nrz scheme but what are we actually going to see in our graphics

00:03:33.480 --> 00:03:39.480
cards the answer is probably late 2024 as that's when we're due for another

00:03:37.260 --> 00:03:43.739
major GPU refresh from both AMD and NVIDIA and reports are already out that

00:03:41.220 --> 00:03:49.500
NVIDIA may be planning to use gddr7 for its upcoming RTX 5000 GPU series which I

00:03:46.980 --> 00:03:52.860
will definitely be buying as long as I can sell a couple of my organs first

00:03:51.060 --> 00:03:56.580
thanks for watching guys like dislike check out some of our other videos account with video suggestions down

00:03:55.379 --> 00:03:58.700
below and don't forget to subscribe and follow