Ross Scott

I think that would only be true if the GPU couldn't keep up with the rate they were being recorded at, which isn't the case. So if the GPU is rendering at sub-64 fps, then yeah, the GPU is the bottleneck. Otherwise it doesn't "remove" it so much as it combines it all in one process. Here's how I see it: Step 1: Rendering the frames (this will cause things to be in slow motion, but in every recording I've ever done, this is ALWAYS faster than RECORDING the frames. You can tell by noticing the speedup when you bring up the console during recording, which halts recording) Step 2: Recording the frames (regular HL2 takes up a LOT of space doing this, that's where the hard drive bottleneck came in using my method. SrcDemo2 keeps this in memory, then Step 3: Process the motion blur, then record it to disk. Srcdemo2 does the motion blur processing on the frames right in memeory, recording to the drive at a fraction of the rate that my 180fps method does. I believe the motion blur processing is a single-threaded CPU-dependent process. Since it does the blur right there, like you said, this puts the bottleneck on the CPU instead of the GPU. You are right in that if the GPU can't stay ahead of the number of frames needed to process by the CPU, then BOTH become the bottleneck, but in this case, as long as the GPU can render 64 frames faster than it takes to condense 64 frames into one for motion blur, the CPU is still the bottleneck. Tell you what, I'll do a test later to get some hard data. I'll do one at higher quality settings (SGSSAA always takes a bite out of speed), but make sure it's still rendering ABOVE 64fps average (when not recording), then another one where I minimize EVERYTHING and see if there's any time difference between the two. If there is, I'm missing something, if not, it's as I said above.

Well I just tested it myself, it came to 8 minutes 15 seconds on my system. I may try the others later. I don't get it, that's much longer than my time, but your CPU should be a little faster than mine. Vsync was off and srcdemo2 was operating? This COULD be limiting things actually. I remember using this as a backup card when my 8800 GTS died and man, it ran some things extremely slow.

Yeah for the record, this is a totally vanilla copy of HL2 I was using, I only installed it again about a week ago.

Well here's the test to anyone who's interested: 1. Get SrcDemo2 working, and point it to an empty directory inside your Half-Life 2 folder. I'm using "blurtest" 2. Download demos I made here and add them to your HL2 directory. 3. Set your video settings to 1280x720 and turn off vsync. I don't THINK the other in-game video settings matter because the CPU will be the bottleneck. To be safe however, I suggest putting all the settings to the default. I would turn off AA and "motion blur" especially. 4. Have your settings for srcdemo2 match mine exactly (besides picking your own directories): Yes, I really want the framerate that high 5. Click "activate" then launch HL2 (don't forget to use an EMPTY subdirectory) 6. Type the following: host_framerate 1920 startmovie blurtest\nameofmovie playdemo short 7. After the demo is over, type "endmovie" (no quotes) 8. Go to the directory where you told scrdemo2 to output the image sequence. Look at the properties of the time the first frame was created and the last frame and calculate the difference. This is how long it took to process the demo. 8. Post your time and CPU specs here or else email me. Doing this may take your computer hostage and alt-tabbing will screw it up. This may NOT occur in windowed mode however, not certain. If you feel ambitious you can provide additional benchmarks: 1. Try out the "long" demo (about a minute in length realtime) 2. Try the same demos at 1920x1080 resolution instead. Thanks a bunch to anyone who tries this.

Hey everyone, I wanted to see if people were willing to help me do some benchmark tests. As it's been mentioned since my last video, the program "srcdemo2" is a game changer in terms of motion blur. It has pros and cons compared to my method, but unloads a lot of the work onto the CPU instead of the hard drive. If you're interested in helping me out, here's what you need: 1. srcdemo2 installed (requires Java + Dokan library) 2. A faster CPU than mine (Phenom II X4 955) 3. A copy of Half-Life 2 4. Ability to follow directions (if you have any experience recording demos or using console commands on the source engine, that's a big plus). Anyway, give me a reply here and I can provide some instructions later. If you can test out srcdemo2 and get it working yourself, that will save a lot of time.

Sure, I'll take those, I've been meaning to set up a wallet soon for that (and accept bitcoin donations). Another user contacted me about Dogecoin already. It's definitely slower than realtime, also single-threaded. Actually what your describing is what srcdemo2 does, although their method is more primitive than mine, but it DOES remove the load on the hard drive and put it on the CPU instead. I'm still experimenting with that. Also the speed may be not much of an issue as Source will wait on rendering the next frame until the first one is recorded, so it will wait on something external to process it if coded right, so no lost frames. Also the avisynth method I'm using is CPU based, doesn't touch the GPU. Check out srcdemo2, it's open source, maybe you can take something from it and combine it with my avisynth method for the best of both worlds. Yes, but sometimes I make stealth splices of footage when I screw up one section, then continue elsewhere (EP54 had a LOT of those). These are imperceptible and not the level loads that you see. It really would be a bit of a small nightmare trying to recreate it, sorry. Plus I use an emulator for HL Source since 1. I HATE auto updates that break things and 2. The updated Steam copy will flat out crash when trying to playback ANY demos. I don't mind sharing the emulation files and demos I have on hand, but again, it's nothing but agony ahead. Thanks a bunch, this is definitely going into "long-term Ross using money wisely to get sustained for life" fund. I actually own about a dozen souls from years back, I set up a booth with a friend where we bought people's souls in exchange for candy. I still have the contracts somewhere. No, I'll take either of those! I definitely have enough money for parts now, but the excess is going towards making sure I permantly stay afloat financially. Prior to now I've been eating into my savings to the point where it's basically exhausted. I AM earning more money over time, but it's been slow, so the additional funds offset all that. Besides developing my channel, which is slowly growing, I'm working on multiple projects that stand a chance of earning more income over time, just not much at the moment. This influx of income buys me a lot of time to stabilize everything. Additionally, I'm doing research into stocks (don't panic, I'm approaching this VERY slowly and I'm not a gambling person) that will likely pan out over time. That's on the "to do" list for sure. All I can say is that's going to happen AFTER Gorilla Gong gets some major upgrades. The guy helping me with that is kind of overworked. Guys, this is nothing, NOTHING compared to Civil Protection. Yes, but there is a waiting period (same with reply emails).

Well the first one is just SSD v. SSD, the other ones are single hard drive v. multiple drives, and the last one is only v. 40/80GB SSDs (SSDs tend to scale in performance along with size, so a 256GB will outdo almost all 128GB, and a 512GB will outperform most 256GB ones). I would like to see more apples to apples comparisons if you find any. I might not be able to find the data I'm looking for, but maybe I could do it manually if anyone's done a "worn down" SSD v. some of the raid 0 HDD benchmarks you posted. I will admit, I still have the perception of SSDs being less stable than hard drives, even in RAID 0. While one bad disk can compromise the array, your odds of grabbing data off it on the way down are much higher compared to a SSD where it's good one day, dead the next. This is more or less guaranteed under most scenarios. I'm mostly still in shock, I've never received this sum of money at once before, my brain doesn't have a lot of reference. That and I'm going to be scrambling to catch up with everything over the next few days, so getting an email reply might take longer than I anticipated, though I promise they're coming.

He's getting pretty low where he is in the game currently. It balances out, most enemies take less shots to kill, but I don't smash open every crate the way you do in HL normally.

It's hard for my brain to get around this concept, I was really on the fence about doing this video, the resolution was a bit of a smokescreen to see if I could get more income (though I meant everything I said in it). The reason I have reservations about this is the way I see it, people watch videos are FANS, they can always just go elsewhere and stop watching. Asking them for money feels like it cheapens the whole experience, when good will from people is the main thing I Want to preserve. Besides, I'm keeping my fingers crossed, but I plan to get a LOT of mileage out of this money. I think if I'm smart about it, I can get myself to a position where I never need money again. Hey, if you're a donator, just email me and I'll record it. I can at LEAST do that much looks pretty subdued to me. Tests of a rocket flying near someone or a rapid moving object can be some of the best (I almost had the "slicer" from HL1 for fundraiser video). No, that's against their policies. Plus even if it wasn't, I would feel more conflicted about that as I would be crowding out smaller people on their network. At least with this video, it's people coming straight to my channel.

Sure, in fact, I was going to add an option like that to the donate page, but haven't gotten to it yet. I'll try and dig up the bitcoin wallet info I have. It all helps, I live at a level where I not only think in dollars, but also in zlotys, which are about a third of the cost. Food costs around 1/2 to 1/3 of the cost of the USA. I can get a leg of chicken with the thigh from a deli for about $1.33 here. It's complicated! I've had a chance to test it out now. Mine absolutely looks better for the same amount of frames, but Srcdemo2 can scale to a high degree and move the load off the hard drive and onto the CPU. What's going on here is srcdemo2 does the "dumb" frame blending, which I showed in the ghosting video, but can do it any framerate. In theory, this means you need an infinite framerate in order to have perfect motion blur. My method works off the data, then takes a guess at the motion vectors for the final step, taking a bit of a leap of faith. The less frames it has, the more soupy it looks, but you don't get that double-image you see in the srcdemo2 shots. In practical terms, src2demo might still be worth it, but it will take a VERY high framerate to compete with my method (I think mine at 180 looks better than theirs at 900fps). I plan to push it higher before I make any final decisions. At least it supports .PNG compression, that helps me no matter what. Ha, well you could always mail it, though that depends on trust with the Postal service. I could also take a check and have my parents deposit it in the USA actually. I thought I would make the $500 mark within 24 hours and was expecting it to dwindle past that, but kept my fingers crossed anyway. It's far exceeded that, this is going to save me. I may NOT have a free slot actually, I already have a sound card and another RAID card and my videocard takes two slots. I wish I could trade in my PCI slots for more PCI express. As for the RAID 0 route, I really am reconsidering it, but I'd like to see more "real world" testing. Where hard drives completely fail me is when they get hung up something and things stutter. Even making this video, parts of it stuttered so horribly in editing, some shots I really couldn't see until after rendering. It's not just write time, but also random access as well. Anyway, if you know of any comprehensive testing with RAID 0, let me know. I'm still kind of in shock over it, I hate putting burden on fans like this, but it is saving me. It's in limbo, I have had zero luck contacting Valve. I may post on that later.

Wow, today has been sensory overload for me. I'm the last person to criticize people for NOT donating. In fact my favorite fans are ones that keep checking back after a while and will be looking forward to videos I'm making years from now, whether they pay a cent or not. I only care about money to the degree of keeping myself afloat, beyond that I just want to make more videos. Guys, I'm happy for anything, it all counts to me. That's actually MORE true of hard drives as I understand it, but I will be flushing the files pretty regularly after recording. Also I read some brands are better about keeping speeds up near max capacity than others. I plan to treat the SSD like it could fail at any time and have backup plans for anything that's CRUCIAL to keep (like demo files, voice recordings, editing projects, etc.). If I have the demo I can always re-record the video. I'm intentionally keeping it open as I still haven't decided. It's more like I have 50 things I want to do, but only time for a couple. I promise to get to the end of HL regardless however. I may do it this year anyway, but I PROMISE to do it if I hit 12k. It absolutely will degrade, but degraded, it's still likely to be faster than hard drives. It could last me a decade even with heavy use. Also writing NON-STOP to a SSD will NOT kill it in a few months. Many SSDs will just randomly die out of the blue because of the controller card. In other words, SSDs can die out of nowhere (warranties are important), but it didn't die BECAUSE you wrote to a lot in a few months, that would take years. As for 4 drives in RAID 0, jesus, I might need a specialty motherboard for that. Many people informed me of this. All I can say is if this works as advertised, this would have made the entire thing irrelevant, I never heard of it before. I'll definitely look into it. I don't have after effects, so no, I haven't. My experience was a framerate lower than 180 led to "soupy" results. Total fucking silence. I almost made a news post about this wondering about suggestions about what I can try next. I've sent multiple emails to Valve with no response and put posts in relevant forums with no answer either.

Well here's a video people may not have been expecting. It's part a send-me-money video, part behind-the-scenes of Freeman's Mind. Either way, I tried to make it entertaining. If you've ever been curious about some of what's involved in making an episode, you might enjoy watching this. The short version is this explains some of the obstacles to making the series HD and why more money will overcome them: DONATE HERE! Now the bad news is the next FM episode is NOT ready yet, however it is partially done. I'll re-record it in HD and get it out ASAP if this fundraiser goes well. In all honesty, I probably would have been done with the episode by now if I never made this video, but the "why isn't the resolution higher?" comments finally got to me, so by delaying the next episode and making this instead, it could be increasing the video quality on all future episodes. On a personal note, ever since Machinima.com prevented me from releasing videos for months at the same time my pay dropped over 80%, this past year or so has been like a marathon for me trying to get my income stable again. I'm still crawling my way back towards sustainability and have been involved with a LOT of different projects to earn more income. I'm doing everything I possibly can to try and stabilize things and hope to do so sometime this year, but any extra money you donate will be GREATLY appreciated. A big part of the reason I haven't bought this part myself is I've never been able to justify it; all donation money over the past year has gone directly to food and rent. By making this video, it FORCES me to buy the part and increase the resolution to the show. Anyway, because a lot of my time is divided up these days, I can say that for now, more money likely WILL increase the release rate of both Freeman's Mind and RGD since it will free up time I would be spending working on other things. As an extra "thank you", I've also updated all the downloadable copies for Freeman's Mind, and I'm going to try my best to reply to all the emails I've received over the past months, something I've been neglecting a lot (again). If you haven't gotten a reply from me, it's very likely you will over the next week. And if you don't donate, I'm still glad to have you all as fans and will keep making more videos no matter what. I throw out a lot of information in this video, so I can only imagine the comments I'll get from people telling me I'm wrong about everything, but here are some questions / comments I'm anticipating: Q: Why aren't you using Kickstarter or Indiegogo? A: Those fundraisers typically require many prize levels that would take a lot of time to fulfill. It's not that I don't want to give people prizes, but I'd rather not spend weeks mailing things when I could be making more videos instead. Q: Are there any prizes? A: Maybe! I will send an email of thanks to everyone who donates. Also I'm willing to give any prizes that don't require me to act like a business and take tons of time to deliver. I am willing to record a line for anyone who's donated $10 or more, just email me about it. Also, I was considering releasing all the isolated voice files for all the Freeman's Mind episodes if people are interested in that. If you have ideas about something else you'd like to see, let me know. Q: You don't need a SSD! You need a capture card or use Shadowplay then play back the demos in slow motion! A: For adding in motion blur, that wouldn't actually help save me any time, plus it would still be susceptible to potential dropped frames if there was some sort of hard drive stutter. Q: You don't need a SSD! Your CPU is too slow! A: While a faster CPU would speed up the motion blur processing some, none of the cores ever get maxed during recording, there's plenty of overhead. The hard drive speed is the bottleneck. Q: Will you be updating all the old episodes to HD? A: Unfortunately, no. I no longer have the old demos and project files necessary to do so. I never would have gotten rid of them if I expected FM to be as popular as it became. Q: What about the Game Dungeon? A: Due to voting results, I plan to focus mostly on FM for the near future, but I have a LOT of RGD episodes planned. This part will remove all barriers for recording in RGD also, so I'll be able to cover potentially any game after this, not just older ones (though I still plan to keep doing those too). Q: Do you eat only beans? A: No, but beans + onions have a good cost-to-flavor / nutrition ratio. - - - ADHD version: Send money to Ross if you want to see Freeman's Mind in HD! Or send money because you can, Ross can use it! New FM coming this month, maybe in HD! LINK TO FORUM COMMENTS

something Freeman-related is coming very soon

Yeah yeah, everybody seems to think this takes 20 minutes to make, but the reality is a little different. RGD is just another experiment that I can produce more quickly than Freeman's Mind. What I really want to do is make a full length machinima movie, but the time required for that is massive and people tend to get pissed the longer I work on a project. All I'll say is you probably don't believe me, but I think I have the capacity to make something better than Freeman's Mind, and better than all my other videos. You might be surprised how much food and rent money can get in the way of other things. While his tone could use improvement, I'll leave this open for now. If there's no sign of progress, you can relock it.

While they could be related, the video I linked to proves that REAL views are being cut also. The reason being is the guy set up a link where more people clicked on the link to his website, coming from Youtube, than Youtube reported as viewing the actual video. If you're confused by this, that's because what I'm describing is impossible in reality. Since that's what YT is reporting though, the only conclusion to draw is that YT isn't reporting all the views. Stuff like the like-to-view ratio and statistically improbable events are circumstantial evidence, the website results however, are direct evidence. I can't say WHY Youtube is doing this, just that they are.

Oh okay, I misunderstood. I'd even say you won't see THAT much of a performance upgrade for upgrading the CPU / RAM in gaming, unless it's a RTS, uses lots of physics, or is specifically multi-core. 4GB of system RAM is fine as long as you're not running anything else big at the same time as your games. BTG is right though, the 1GB of VRAM is probably your biggest bottleneck towards gaming performance more than anything. That's part of the card however, so you can't exactly upgrade that without getting a new card and it sounds like you have an okay midrange one as it is. The main thing that will impact is how high the texture resolution you use is, and your non-shader-AA performance. Also if you do like AA on the cheap, you might want to look into the "FXAA injector" rather than the MLAA that AMD includes. I don't know if this is still the case, but FXAA has traditionally had better performance than MLAA. Also BTG, you mentioned SMAA, I have mixed feelings about that. While it DOES give a sharper pictures on edges it handles than other shader-based AA methods I've seen, I think it creates more of a contrast effect also, because it draws my eye to the parts that it misses, making them stand out more for me. FXAA tends to be a tad blurrier and more inclusive than SMAA, which for me personally makes the problem areas a little less noticeable, but it's a personal preference thing.

Well out of all the old games to pick, it was Crysis on very high settings, which is about the most punishing benchmark you can possibly get from the time (Crysis is still more stressful on a system than many modern games). It was also done on Vista 64 bit, so it could definitely address more than 2GB. As for what you're describing, that's true, but again, that comes back to the GPU. You WILL see a performance difference if you're loading so many textures that it spills over past the VRAM, but at that load, the performance is likely to drop more dramatically elsewhere without a lot of GPU muscle anyway and would show much more performance for the dollar from having a card with a larger buffer rather than worrying about the RAM. I don't know man, I just feel like a lot of the arguments you're making for gaming seem more like fine-tuning performance on an already performing system rather than primary considerations. By that I mean what you're saying is correct, but if you're on a budget, it's sort of the last step to be worrying about, since you have rapidly diminishing returns on performance for the dollar compared to the GPU. Stuff like memory controllers, RAM speeds, CPU power (with a few exceptions mentioned), you worry about AFTER you have the beefiest GPU you can handle. Otherwise, you might have a more rounded-out system going your way, but you'll have lower FPS in games.

Unless I'm missing something important here, no, not really. I found benchmarks on it actually: http://www.tweaktown.com/articles/1782/amd_phenom_ii_ddr2_vs_ddr3_performance/index.html There is a little bit of difference, but nothing close to the 15fps you're claiming, it's basically the 2fps difference sheridan is talking about. Granted, that's on an older CPU, but it's the RAM that's being tested. RAM really isn't the bottleneck for the majority of programs, especially games. Yeah, a GPU is pretty much the across-the-board performance gain for gaming, outweighing the benefits of a faster CPU and RAM in all but extreme cases. In any event, the money you put into a GPU is generally going to get you much more performance than the money upgrading a CPU (unless it's VERY dated or underpowered). There are exceptions, RTSs in particular, games that have a bunch of physics done on the CPU, a few multi-threaded games can show benefits from more CPU muscle, but by and large the bottleneck is the GPU. This could change in the future however, as the latest gen consoles are focusing on lots of threads, so that may trickle down to the PC side as well. I've read Battlefield 4 is one of the very few games that will really use all the cores you throw at it and actually does put a lot of the load back on the CPU, something that hasn't been the case for a long time.

That example I think is more grey. I see Obamacare as both helpful AND harmful to people, depending on what bracket a person is in. What I see as clear-cut evil is our system that rewards health insurance companies for denying coverage to people that need it at the same time they take in massive profits. That's literally profiting on the misery and death of people who need their help. Collectively, they spent about 500 million on lobbying in 2012. With Obamacare, Obama clearly compromised heavily with these forces, so that makes him "not good" at least. As for him being evil, his actions seem to deviate from what he says a lot. I can't really say if he's evil or not, his intentions seem totally unclear to me. I mostly see him as a conduit for larger forces at work. I've thought many times that we should just rename the USA to "Crazyland" and it would make us far more consistent. You can point to Christianity as a source of morality, but in the USA, things get surreal. My favorite example of this was at the republican presidential candidate debates where Ron Paul was for suggesting we abide by "the Golden Rule." So in other words, you have a demographic of people who are largely Christian booing somebody for quoting a central theme from Christianity.

I doubt that's the case. Ad block just means the view doesn't count towards any ad revenue, if it actually meant views didn't show up, this would have been a massive issue long before now. They never told me, I never got a reply email, but they fixed it, 4-5 days after release.

No, I'm ahead of you on this unfortunately. Blip supports re-uploading media to the same URL created, so I did that, no luck, same problem. As for the settings, these are the exact same settings I've used for the last 50+ episodes uploaded to blip, so that's a dead end there too (and I'm already using 441.khz for the audio). I think something changed on blip's end. I'll look into it some more later.

Unless you're talking on a super-meta level, I have to disagree with this. I mean in the big picture, we're a speck of life the universe probably doesn't care about, so yeah, anything that happens, isn't good or evil in any sense that matters. On the planet Earth however, there's plenty of things that can be defined as good or evil in almost any context. For example, if I go around kidnapping people and chopping their hands off because I want to decorate my fence with them and I don't care about their pain, well, there's not really any defense of an action like that. I tend to get defensive when people say there is NO good or evil because that moral ambiguity is the sort of thing sociopaths purposefully exploit. For their purposes, it only benefits them for people to believe there is no evil as that provides more of a tolerance for actions that are clearly evil. There are grey areas to be sure, but to deny it entirely I think only works in a cosmic sense. A rock doesn't have good or evil, but people do. I like this example. I think besides how intentional something is, responsibility is also an issue. Taken at face value, assuming the politician couldn't have predicted these consequences, the politician is not immediately evil, even if the consequences of his actions ARE. The "test of evil" in this scenario is what the politician does AFTER learning the outcome of his actions. Politicians are in a position of power, which carries responsibility behind it. If the politician later realizes what he has done is harming people and tries to rectify it, then he's not evil, he's just flawed / ignorant. If he realizes what he's done, feels bad about it, does nothing to remedy it, but will never pursue action like that again without more careful research, then I'd say he's still not evil, but not good either. If he ignores the negative consequences and feels no remorse about his actions and / or CONTINUES to do further actions like that, then yeah, I'd say evil. Unintentionally committing an act that causes evil doesn't make the person evil, but not caring about the consequences of those actions does.

Absolutely, I sent it to my family also. It has almost a million views, it needs more like 100 million.

Maybe, although I really hate the idea of holding the episodes hostage for money. Believe it or not, exploiting my fanbase for income is NOT one of my goals (unless they're loaded). They have a de facto monopoly on video streaming, I thought I read before it's about 70% of all free streaming video. Makes it hard for alternatives. Old Google is very different from modern Google. I recommend reading #1 of this article to get a better picture of what's happening. Yeah it makes no damn sense. Like Windows 8, there are quantifiable flaws with their new design. "Inbox" used to take ONE click and ONE page load. Now it takes FOUR of each. To turn off annotations, that used to take 1 click, now it takes 2. Increasing the resolution used to take 2 clicks, now it takes 3. What this says to me is one of the largest companies in the world can freely ignore fundamentals of good UI design. It must be maddening to people who have degrees in that sort of thing. Thanks, glad to see you got Ettore Majorana right. BTW no one can disprove he wasn't killed by ninjas at sea.