Upload
Login or register

gasur

Last status update:
-
Date Signed Up:1/24/2013
Last Login:12/16/2015
Stats
Content Thumbs: 5 total,  11 ,  6
Comment Thumbs: 200 total,  236 ,  36
Content Level Progress: 15.25% (9/59)
Level 0 Content: Untouched account → Level 1 Content: New Here
Comment Level Progress: 40% (2/5)
Level 110 Comments: Funny Junkie → Level 111 Comments: Funny Junkie
Subscribers:1
Content Views:1022
Times Content Favorited:1 times
Total Comments Made:96
FJ Points:167

  • Views: 1021
    Thumbs Up 11 Thumbs Down 6 Total: +5
    Comments: 0
    Favorites: 1
    Uploaded: 01/24/13
    Creepy Creepy

latest user's comments

#15 - Wondering too.  [+] (6 replies) 11/02/2015 on grafficks 0
#17 - anon (11/02/2015) [-]
For those wondering they released a "4gb" card which was actually 3.5gb ram and .5gb of "virtual ram" and this information was not released and only became public by people buying the card and seeing they were not pulling what they were expecting.
User avatar
#38 - nathanbiggs (11/02/2015) [-]
It wasn't "virtual ram", it was actual RAM that just didn't have the read/write speed of the 3.5 GB partition.
#39 - ragealicious (11/03/2015) [-]
It didnt even matter anyways tbh. People were trying to get WAY too much out of their 970's blowing them to the max when that card wasn't designed for it. Sure nvidia lied, and i understand their angst, but plenty of people were mad about it not pushing 4k(which is how most people found this problem) with a mid end graphics card. Next to no games would push you to ram limits unless you;re trying to go above and beyond what the card is capable of.

On a side note this problem doesn't exist in the 980ti, which is beatiful
#37 - anon (11/02/2015) [-]
Except that's wrong. It's 4 GB of physical memory on the card. However the 7th and 8th share the same port onto the crossbar. Typically a port is designed for maximum speed of one 0.5 GB segment, not two as the 970 has. What this means then is that at 100% utilization that shared port needs to twice the throughput capacity of the others, which it doesn't half. So instead it shares the available throughput by favoring the 7th 0.5 GB segment over the 8th. During full utilization however you'll see a case of memory thrashing where the memory needing to be accessed is being accessed much slower than the requests are coming in. Hence you'll generally see a pretty noticeable collapse in throughput on the last 1 GB of vram.

Nvidia got around this issue by segmenting the 4 GBs of actual memory into two pools, one with 3.5 GB and the other with the 0.5 GB. All memory usage will by default go into the reliably fast 3.5 pool, but the issue arises when you go over that. As you can imagine, the GPU will let you access the extra 0.5 GB pool but at a theoretically big performance loss. This would be somewhat analogous to a road system. Where you build a road originally to support, say 100 cars, and then diverting nearby traffic onto this 100 car capacity road, thereby doubling it to 200. You'll find that traffic jams become a rather notable issue.

Nvidia also incredibly stated the number of ROPs and L2 Cache memory was available. That is to say, it was stated as being the same as that of a 980. Which doesn't make sense to anyone who understands GPU manufactering, because many cards are actually based off a trimmed version of the more powerful card. In this case the 970 is actually a trimmed down version of the 980. This is pretty typical design for both AMD and Nvidia. Though I do not condone any of this behavior, as it is ultimately misleading.

So in summery, this anon has no clue what he's talking about.
#34 - akkere (11/02/2015) [-]
This video has a good summary of the whole thing.
>>#16,
User avatar
#35 - akkere (11/02/2015) [-]
#159 - It's just as fast (in theory) to QS, as it is to wait, yes? Bu…  [+] (2 replies) 10/06/2015 on counterstrike 0
User avatar
#160 - askafj (10/06/2015) [-]
Yes but you don't have to react to anything, you already know whats happening so you're just doing it on the correct frame.

Human response time isn't relevant here as it's all muscle memory.
#161 - gasur (10/06/2015) [-]
Even so, you can never be accurate at the exact time. It's still better to wait, if you know prehand that you're going to peek it again.
#153 - The point is, if you're holding a corner and wish to peek it a…  [+] (4 replies) 10/05/2015 on counterstrike 0
User avatar
#157 - askafj (10/05/2015) [-]
It isn't though

since you can scope at the same time as you normally would
#159 - gasur (10/06/2015) [-]
It's just as fast (in theory) to QS, as it is to wait, yes? But then you have to take in human interaction. We have delayed response time, so in reality it is indeed slower. And if you indeed know you are going to peek it again, you should wait. Else, switch to knife and get behind another position.
User avatar
#160 - askafj (10/06/2015) [-]
Yes but you don't have to react to anything, you already know whats happening so you're just doing it on the correct frame.

Human response time isn't relevant here as it's all muscle memory.
#161 - gasur (10/06/2015) [-]
Even so, you can never be accurate at the exact time. It's still better to wait, if you know prehand that you're going to peek it again.
#84 - Quick switching is NOT faster than waiting it out. It's actual…  [+] (6 replies) 10/05/2015 on counterstrike -1
User avatar
#90 - askafj (10/05/2015) [-]
if you're good enough you don't lose time
#153 - gasur (10/05/2015) [-]
The point is, if you're holding a corner and wish to peek it again, it's slower to quick switch and scope back in.
User avatar
#157 - askafj (10/05/2015) [-]
It isn't though

since you can scope at the same time as you normally would
#159 - gasur (10/06/2015) [-]
It's just as fast (in theory) to QS, as it is to wait, yes? But then you have to take in human interaction. We have delayed response time, so in reality it is indeed slower. And if you indeed know you are going to peek it again, you should wait. Else, switch to knife and get behind another position.
User avatar
#160 - askafj (10/06/2015) [-]
Yes but you don't have to react to anything, you already know whats happening so you're just doing it on the correct frame.

Human response time isn't relevant here as it's all muscle memory.
#161 - gasur (10/06/2015) [-]
Even so, you can never be accurate at the exact time. It's still better to wait, if you know prehand that you're going to peek it again.
#15 - It only increases your movement speed for a very short time, s…  [+] (9 replies) 10/04/2015 on counterstrike -7
User avatar
#89 - askafj (10/05/2015) [-]
you can do it without losing any time
User avatar
#22 - Faz (10/04/2015) [-]
Notice how every single pro who plays the game quick switches after shooting? That is all the evidence you need to know you're wrong on the subject, quick switching is more advantageous since it allows you a larger view and you don't have the recoil animation which sucks, plus now more than ever quick switching is the best way to awp since you get a huge speed nerf when moving while scoped in.
#84 - gasur (10/05/2015) [-]
Quick switching is NOT faster than waiting it out. It's actually slower, if you want to shoot once more, since you also need to scope in. You also have a larger view when you've shot, as you won't be locked into the scope, but the scope will only come back up whenever it's done reloading.

Also, most pros tend to equip the knife right away, to get to another position, not to peek again.
User avatar
#90 - askafj (10/05/2015) [-]
if you're good enough you don't lose time
#153 - gasur (10/05/2015) [-]
The point is, if you're holding a corner and wish to peek it again, it's slower to quick switch and scope back in.
User avatar
#157 - askafj (10/05/2015) [-]
It isn't though

since you can scope at the same time as you normally would
#159 - gasur (10/06/2015) [-]
It's just as fast (in theory) to QS, as it is to wait, yes? But then you have to take in human interaction. We have delayed response time, so in reality it is indeed slower. And if you indeed know you are going to peek it again, you should wait. Else, switch to knife and get behind another position.
User avatar
#160 - askafj (10/06/2015) [-]
Yes but you don't have to react to anything, you already know whats happening so you're just doing it on the correct frame.

Human response time isn't relevant here as it's all muscle memory.
#161 - gasur (10/06/2015) [-]
Even so, you can never be accurate at the exact time. It's still better to wait, if you know prehand that you're going to peek it again.
[ 96 Total ]