WEBVTT

00:00:00.280 --> 00:00:08.080
today I'll be using a synthetic Benchmark in this case 3D Mark 11 to

00:00:04.839 --> 00:00:11.719
evaluate something that I have received

00:00:08.080 --> 00:00:14.440
a lot of questions about over the years

00:00:11.719 --> 00:00:20.320
so what you may or may not notice is that I have changed my testbench I have

00:00:17.160 --> 00:00:23.080
gone with an Intel x58 platform for this

00:00:20.320 --> 00:00:26.920
particular test I'm using the Gigabyte G1 gorilla

00:00:25.119 --> 00:00:31.920
motherboard's get you guys better look at that G1 gorilla motherboard with an

00:00:29.240 --> 00:00:38.640
Intel Cor i7 960 so that is a quad core processor

00:00:34.960 --> 00:00:41.920
I've got that chip clocked at 4 GHz in

00:00:38.640 --> 00:00:48.199
order to reduce the any CPU bottlenecks

00:00:41.920 --> 00:00:50.000
that may occur in my test and now okay

00:00:48.199 --> 00:00:54.320
no do not always ask before opening this file so what I'm going to be looking at

00:00:51.480 --> 00:01:00.600
today is the effect of PCI Express bandwidth on SLI performance so you can

00:00:58.719 --> 00:01:05.040
see I have not used the top slot and there's a very good reason for that if I

00:01:02.399 --> 00:01:08.840
populate the top slot then it will run at 60x bandwidth and then if I populate

00:01:07.360 --> 00:01:15.880
the second slot I will have two cards running at 16x bandwidth if I populate

00:01:11.720 --> 00:01:20.640
these two slots I will have two cards

00:01:15.880 --> 00:01:22.320
running at here we are PCI e16 at adex

00:01:20.640 --> 00:01:30.200
bandwidth so I'm just going to switch to the other card and show you guys here so

00:01:25.200 --> 00:01:31.360
there you are at 8X bandwidth so that is

00:01:30.200 --> 00:01:35.720
what we're going to test I'm going to see what the 3D Mark score is running in

00:01:33.680 --> 00:01:40.079
these two slots versus these two slots so you guys can see how much of a

00:01:38.040 --> 00:01:44.479
difference there is in performance okay so for my first run through we got X

00:01:42.680 --> 00:01:51.680
3386 3386 3D marks in 3D Mark 11 so now let's

00:01:49.320 --> 00:01:56.439
run them at 60 next and see what kind of performance boost we get from giving

00:01:53.680 --> 00:02:00.439
them the full available bandwidth our two card configuration running the same

00:01:59.200 --> 00:02:06.560
test achieves

00:02:02.560 --> 00:02:09.800
3,389 3D marks so for

00:02:06.560 --> 00:02:12.040
comparison at 8X and 8x bandwidth we

00:02:09.800 --> 00:02:17.000
achieved 3,386 3D

00:02:14.080 --> 00:02:23.400
marks so we oh oh hold on hold on no no we should put in the other one

00:02:19.120 --> 00:02:25.599
3389 okay divided by

00:02:23.400 --> 00:02:29.599
3386 to find out the performance Improvement .1

00:02:32.680 --> 00:02:35.800
or rather that's a

00:02:36.720 --> 00:02:49.560
.01% performance Improvement by going from 60 uh 8X 8X to 16x 16x PCIe 2.0

00:02:46.239 --> 00:02:53.400
slots so um to anyone who wants to run

00:02:49.560 --> 00:02:55.800
SLI or Crossfire on the Intel p67

00:02:53.400 --> 00:03:00.959
platform I guess this means you're pretty safe from performance degradation

00:02:58.599 --> 00:03:04.720
even with high performance gpus like the GTX 580 thank you for checking out this

00:03:03.120 --> 00:03:11.120
little episode of Linus Tech tips don't forget to subscribe for more unboxings

00:03:06.200 --> 00:03:11.120
reviews and other computer videos