Phasmophobia – a low tech Ghostbusters game

Halloween is coming and I recently discovered this game thanks to @LomoFear, which is ideal for this winter season if you like scary games/movies and play with friends. I’ve been completely caught up with this game and I can’t stop playing it.

In Phasmophobia, you’re part of a crew (1-4 people) that got a contract to investigate a ghost sighting in a house. Your objective is to collect evidence and identify the ghost. You can’t expel the ghost but the ghost can kill you.

Here, we begin on our truck to get the gear before heading out

For me this is like the Ghostbuster movie but in a low tech fashion, maybe more “realistic” in the sense that it’s more accurate to the actual lore. If you like mystery and ghosts in real life you’ll see that everything is more or less depicted in the game as it is in real life. I don’t (or no longer) believe in ghost myself, but I like all the stuff around them, it feels scary and interesting.

Phasmophobia will make you paranoid. It’s a game where almost never anything happens, at which point you’ll doubt yourself about if you actually saw or heard something or not. But, when something actually happens, oh boy… you’re about to die. Run!

You count with basic tech to locate and identify the ghost, which it is all unreliable. Some of the stuff is an EMF Reader (Electro-magnetic Field reader), UV Light (fingerprints), a radio (electronic voice phenomena/EVP), and a Ghost book. With these devices (and some extras you can buy) you go inside and keep testing to see if you can detect anything.

The game can be played alone, and up to four players together. There is a voice button to talk, and a radio button. The voice only reaches a few meters, while the radio is everyone together. But the radio will stop working when the ghost starts hunting.

Ghost hunts are freaking scary. When the ghost gets angry and fed up with your presence, it will go all-out hunting for you, and it will try to kill someone. When it happens, the lights start flickering, sometimes even if they’re off, and the flashlight as well. Run! But you cannot outrun the ghost, your best bet is to find a hiding spot, shut up, and turn your flashlight off.

Some of the optional objectives are taking a photo of the ghost, or stop a ghost hunt with a crucifix. These can be risky and sometimes it’s best just to leave them out if you’re not confident. Safety first!

We’re dead and we can see our old body lying on the ground

If the ghost haunts you, you’ll watch a 5-sec terror scene, and then you’ll move into the world of the dead. You cannot communicate with your teammates anymore, your body will be on the ground and everything has a blue tint. You can hear and see them, and you’ll be able to watch the game more effectively. I’m not sure, but I think the ghost is easier to see when you’re dead already. If you’re alone, or all teammates die, the game ends. If you die, you lose your equipment. In Amateur difficulty you’ll get back 50% of its value from the insurance, while in Professional is all lost. If you didn’t invest any personal equipment that you bought, nothing will be lost.

I don’t know why it’s so appealing to pursue something invisible, but I’m totally caught with this game. Also, I cannot play for long periods (>3h), I get tired. Probably this constant sense of danger and paranoia gets me tired. I can sense my heart ramping up at times.

My tips and recommendations

I’ve played a bit already (~20h) and I’m noticing that people tend to do the same mistakes again and again. If you’re kind of new into the game, maybe these tips help.

1. Don’t rush

The game is not timed. Yes, there’s a 5 minute timer for amateur & intermediate difficulty to set things up, but that’s it. If the timer reaches zero, staying in the truck for 1 hour will not make things any worse.

The timer it’s only to make sure the ghost will not start hunting straight away, you will consume it in the first trip anyway, so make the most of it. Even in professional when there’s no timer, the game doesn’t actually start until you open the front door. So, plan ahead calmly, get the stuff you need near the front door, and think of a plan.

2. Always run

If you enter the house, I expect you have a plan on what to do, which path are you going to follow and what evidence are you trying to collect. If you have that, then make it as efficient as possible. Don’t roam more than enough, and hold the shift key all the time to make the trip as fast as possible.

I know, holding the shift key doesn’t provide you much more speed, but if it’s a 20% faster, you’ll end in 20% less time, and that counts. Also you don’t get tired, so there’s no point on not running.

Unless you’re hiding. Run to reach your safe spot, close doors, and then stay quiet, calm, and don’t run. Crouch as well. You’re playing hide&seek with a ghost, you don’t want to make noises or light changes.

3. Buy & carry a thermometer

The first pass is the most important, you want to locate the ghost room or at least narrow down the area. If the place isn’t too big (i.e. a house), a thermometer is the best tool to do this.

Most ghosts don’t produce freezing temps, but most of them cool down the room where they are (but not always). The thermometer is an infrared one that displays the temp where its pointing at. Just pointing inside a room for 0.5 sec is enough to get a reading, so that makes the thermometer a tool that can very reliably locate the ghost in no time.

This item is not on the default ones, so it’s the first one I recommend to buy. Having 2 thermometers and a friend will speed up this process quite a lot. I don’t recommend having more than 2 people with the thermometer unless you are in a big map (university or asylum) and plan carefully each one to sweep a different area.

Some ghost don’t lower temperature, while other ghost could do it after 5-10 minutes of activity. So, not finding anything is normal, and it might be worth a second sweep later. 80% of the time the first sweep of the thermometer is enough to pinpoint the room.

The second most useful item to sweep rooms is the EMF Reader. The problem with this one is that it won’t detect anything unless the ghost touches something or has strong activity. So sweeping with this is unreliable. This tool both works in directional (scan what you point to) and in omni (scan nearby). If you put it on the ground and the ghost is active very close to it (0.5 meters more or less) it will lit up, and when handling it if you point to objects it will scan them, so you can understand which object was touched.

4. Be quiet and follow the sounds

The ghosts usually like to make their presence known, they will tell you where they are. If a sweep is not a good option or does not provide any result, the best way is to just hear.

If you anticipate that this is going to be a problem, bring some sound sensors, and place them on strategic locations across the map on the first sweep. But they’re expensive and you’ll want from two to four of them. These cover a great area and hear trough walls. Just be sure to face them properly as they’re directional.

If you don’t touch anything or open most doors (or if you can remember the state of the doors and light switches) you can also locate the activity because something has moved. On some houses you can see it clearly because there’s stuff on the ground that belongs on a table (cutlery, dishes,…). On other maps you’ll rely on doors and lights.

But in the end, the most important thing here is that the ghost is constantly making some noises. It activates sinks, moves objects, rings the phone, activates alarms, opens and closes doors… all these things make noises. Follow them without fear, if you can’t locate the sound source, try to go to the same point in the map you’re hearing from but in a different floor. The ghost tends to roam near its room, so even if the room where you found activity isn’t the ghost room, it’s nearby.

The parabolic sound sensor doesn’t work for me. My hearing seems to be better than this sound sensor. But the in-wall sound sensors (the regular ones) are nice because you can wait on the truck to locate activity. You can stay as much as you want there, it’s safe and there are no penalties for doing that.

5. You don’t need to locate the actual ghost room

Yes, the ghost has a room, and this is kind of it’s base of operations. It will roam from there and go back to the room. It’s useful to locate it because EMF Readers, Spirit box, fingerprints, etc will most probably appear in that room. But, remember this isn’t a must. These things activate not on the ghost room, they activate upon a ghost.

So, if you know that a ghost is usually roaming certain place, you may get results just in a corridor. It takes more time because it needs both the ghost passing by and willing to do something (talking, writing, touching some stuff, etc).

For example, if you saw the ghost touching a door or a light switch, you can test them for fingerprints. They don’t always appear, but if the ghost just did it, it’s likely to be visible if the ghost leaves fingerprints.

Also, on camera, if you see the ghost moving stuff but you can’t see orbs, you can almost discard the ghost having these. The orbs also follow the ghost. Cooler temps also follow the ghost, so the thermometer can be used to know if the ghost is in the same corridor or room as you are.

6. There are secondary tips to classify a ghost

Different ghost show different behavior. For example, if you put down the ghost guest book and you notice it kicks it, assume it doesn’t like it. Most probably that ghost is not one that writes. Poltergeists have a tendency to do this.

Jinns for example like to mess up with electrical items, and have a tendency to prefer lights up. If you hear that it starts radios, heaters, and strangely it doesn’t stop the lights,it might be a Jinn. Remember that the breaker might trip if too many things are on at the same time, and this is not a ghost doing anything. If you think it’s a Jinn, turning off the breaker or making it trip by adding to many lights will weaken the ghost.

Demons tend to be really active, aggressive. While the Mare is really scary and will shut down all the lights and make the scene as dark as possible. Revenants are hard to pinpoint because they don’t get active unless you get close, but when you get close they get really angry, these are quite dangerous.

Also, lack of evidence can also be evidence. If you’re sure that you got the location right and it doesn’t do something, well, give it time, but you can start thinking on discarding that evidence altogether. The EMF Level 5 is the hardest to discard because you need really strong activity in front of the reader; I use to place the reader in the view of a camera to monitor it from the truck. The problem is this does not work well with Revenants, as they will become inactive when you leave.

7. Stay on the truck, someone

Rookie mistake, except for the first sweep, it doesn’t make sense to have all four people in the field, you’re wasting precious team sanity. Set up the basic stuff and leave one teammate in the truck. Someone has to watch cameras, activity, sound sensors and sanity.

After a trip or two, rotate the teammate, this will even the sanity. You want sane people inside, insane people outside.

Setup the area to do the stuff from outside, remotely. The activity will go up eventually, even without anyone inside (except Revenants). Give the ghost time to write, to leave fingerprints and to roam around.

8. Don’t call the ghost name

A lot of people like to shout the ghost name again and again. This angers them. There’s no need of doing that until you have it classified. Angering a ghost, provoking him, is useful just to do optional objectives: taking photos, surviving hunts, etc.

On the Spirit Box, avoid the ghost name and just ask for something easy like: How old are you? What is your gender?

Some phrases may anger it, and you don’t need an angry ghost, at least until is properly identified.

Running Phasmophobia on Linux

If you know me, you know that I don’t use Windows or Mac, just Linux. So if I’m recommending this game is because it does work properly.

Steam is able to install it properly with proton out of the box, no tweaks needed. The game will not use voice recognition, but the spirit box will talk alone. So it’s perfectly playable, single and multiplayer.

The only issue I found is that there are FPS dips, it lags sometimes without explanation. I tried a few things that helped quite a bit:

I updated to the beta-unstable channel. The devs are working on these issues and the beta should be more performant.

I forced the use of a newer proton. In my case I tried the 5.2-GE which is a fork I manually installed for other games. I don’t think we need the fork, just the latest 5.x proton should suffice. This seemed to help.

Finally I reduced the quality of the shadows (resolution), I was seeing input and redraw lag, which was solved after lowering the shadows.

Out of the box was playable, but after the tweaks it’s performing quite nicely.

100% Recommended

I love the game, give it a try. I got it for about 12 euros, which I think it’s quite cheap. There’s also a cheat-sheet here that I found useful: https://un0btanium.github.io/phasmophobia-cheatsheet/

Let me know what you think!

AMD Ryzen 5 5600X, a gaming beast?

They just announced yesterday the new line of Ryzen processors on Zen 3, where we got the first three processors to the new family:

  • Ryzen 5 5600X: 6c/12t, 3.7/4.6GHz, 3/32MB, 65W TDP (cooler included)
    Price: $299
  • Ryzen 7 5800X: 8c/16t, 3.8/4.7GHz, 4/32MB, 105W TDP
    Price: $449
  • Ryzen 9 5900X: 12c/24t, 3.7/4.8Ghz, 6/64MB, 105W TDP
    Price: $549
  • Ryzen 9 5950X: 16c/32t, 3.4/4.9GHz, 8/64MB, 105W TDP
    Price: $799

On the benchmarks AMD released, we saw between a 20-30% uplift in single core, yet I did not saw any benchmarks about multi-core performance. Yes, I know the event was all around gaming performance, but it’s still strange/funny the multi-core wasn’t stated. If they added a 20% of uplift also in multi-core, wouldn’t they brag about it?

I think they focused this time around on single-core performance, because it was the only part where Intel had some ground left, and not anymore. Well played, indeed.

Something went unnoticed here, while the boost clocks are higher, the base clocks are around 100MHz slower than the 3000 series. This, in conjunction with having same TDP and same node (7nm), it will probably mean that the multi-thread performance is going to be roughly the same (maybe around 5-10% better?).

Understanding how they’re built

In order to see which one to chose from we need first to go back on how they get different models. No, they don’t build three different chips, in reality, the chips contains small chiplets (CCX) where each one can hold up to 8 cores. These chiplets are all the same across all 5000 processor series. So, how they get 4 processors out of a single chiplet?

The 5800X and 5950X are simple to understand, these are simply one CCX or two CCX, period. The other models, they get CCX that are less performant (or slightly defective, depending on how you look at it), disable some parts, and ship them. For example, by lowering the clock they can make a processor run even if it’s not as perfect as it could be. In this way, AMD can sell almost everything that they produce, giving lower costs.

Here’s my guess on how I think they’re building them:

  • The 5950X, I believe it would get highest quality CCX, two of them.
  • Then probably the 5800X will be a single CCX, of high quality possible. So this processor gets the full 8 cores and the full 32MB cache.
  • Next comes the 5900X, which will have 1 core disabled on each CCX.
  • And finally, the 5600X which comes with 2 cores disabled.

What does this mean? For gaming, unless you have a huge budget and don’t care, I’d say the 5600X is the most attractive option, as it includes the cooler and it’s quite cheap ($299), it is going to be the most bang for the buck. Why? As all CCX are the same, a slight decrease in frequency is not going to make more than a 5% difference in performance. And games only abuse the first core. So, more cores does not yield higher FPS.

Remember, those are going to perform 10-20% better in gaming than the best Intel out there. If you can get around 5% better than any Intel for $299, I would say it’s a bargain. It’s only 6 cores, yes, but c’mon! 6 cores is a huge amount of cores. Even for professional workloads, 6c/12t is quite good.

For those like me that plan to use the computer for other stuff, like compiling programs, I see the 5800X as a great high-end option. It’s most probably to have one of the highest quality CCX, and an all-round performance in any type of scenario. This is going to be my target for my next computer.

Then the only thing that I’m now intrigued about is how these chips will overclock, specially the 5950X. This chip holds way too much power for the package. Its TDP is still 105W TDP, and it can be clearly seen that the base clock had to be reduced, this is probably to put it into spec for the given TDP. I think this chip will really like expensive cooling setups, for example a custom water loop with a 360 radiator. If the motherboard has capability to increase the TDP (by fooling the processor about the current draw), and given the boost frequency is so high, it’s possible that the chip will auto-overclock itself.

My guess on future releases

It’s possible that the 5950X overclocked could get really near the ThreadRipper 3960X. And probably the reason of AMD to not to release higher core counts is to avoid competing with the ThreadRipper line of processors. This means that ThreadRipper needs an overhaul before the next generation of processors, adding the same technology here to get that extra 20% performance, same core count but maybe discarding the lowest one.

But then the problem becomes Epyc, the server line, if they release a new ThreadRipper that could be 40-50% more performant, the server market would get hit. Remember that AMD is selling a lot of the Epyc Rome processors. So this means that whatever they do, they need to release the new Epyc codenamed Milan.

It also needs first to get an idea on how the new line goes, how good are the yields and so on, prepare the new APU (CPU with graphics) line, and get some low-end processors out. At some point it has to get with some refreshes to fight Intel back when they release their new platform.

So my guess here is more or less what I expect:

  1. Release Epyc Milan around May 2021.
  2. Present new processors with graphics + low end CPU from the lower yields, around April 2021. The APUs could be really good this time around.
  3. Refresh of processors + new ThreadRipper around November 2021.

Those are 100% guesses.

What about Intel?

They lost the crown on gaming, productivity, laptops and servers. On laptop they still hold a bit of ground but not by much.

On laptops, AMD is probably trying the 5nm process to get a 6xxx series processor for low consumption, if they release something like that, Intel will be done.

I think they should stop worrying about consumers/gamers and let AMD have the crown for a bit and focus on servers heavily. It is unlikely that they can beat AMD in 1-2 generations anyway. In the other hand, servers should give a lot of revenue, as the chips are really expensive. AMD has the Epyc Rome for a while and Intel has no response for that. With the amount of datacenters that exist, as they begin to refresh their CPUs, they will be all AMD at this rate.

Even if I’m a fanboy of AMD I really want Intel to be there, because if not, the prices will increase a lot. We need competition.

Time to upgrade our computers

I am still running my old i7-920 with a GTX 1060 and in January 6th I talked about why I was not upgrading yet. This is going to change quite soon! The wait is almost over.

What changed here? Well, first things first, my habits on the computer are changing. I have been trying to encode video and to program in Rust, both of these really benefit of multi-core processing and my old i7 does not cope well with them. I also upgraded my monitor to 1440p, and the GTX 1060 is suffering quite a bit when gaming.

On the other end, the market is about to release new stuff that is really compelling for an upgrade. On graphics cards, both NVIDIA and AMD are pushing hard for the crown, which is driving prices down and performance up, by a noticeable margin. On CPU, AMD is about to release a new line of processors that seem would have both a big IPC and frequency jump.

Last time I compared how bad it seemed that after 10 years we got only 2x faster on CPU. But in one year and a half, AMD is releasing CPUs that should be ~30% faster than the 3000 series. So, to put in perspective, the previous years we had an average uplift of 7.5% performance per year, while this new generation should be around 19% better per year (I accounted here for 1.5 years, in a compounding effect). This is 2.5x better than before, and adding this to the fact that I own a really old system, makes me really excited for this.

The new GPUs are more or less on the same territory. The RTX 3070 should be as fast as the RTX 2080 Ti, costing only around 550€ while the old one was around 1500€. Same performance, a third of the price. AMD is also going to present the new GPUs, and leakers point to them being on the same territory as NVIDIA ones, with less power consumption.

Everything should be out and in stock for Black Friday, as it seems to be what both companies are targeting. But, given the uplift and the amount of hype, I don’t expect the stocks to hold. In fact, the 3080 and 3090 stocks have lasted seconds. I just hope that AMD comes with a huge amount of production to cope with it. I don’t expect NVIDIA to do much better on the 3070 launch, even after delaying it.

CPU for gaming

I don’t consider myself a gamer, I’m more onto productivity and I can afford to lose some FPS if I’m going to get better performance on other stuff that I do. So I will probably target a Ryzen 7 5700X along with a beefy cooler. But if you use your computer mostly for gaming, here’s something interesting: Most CPU’s should perform more or less the same for almost any games.

Most games don’t use more than 2 threads intensely, so what matters here is the single thread performance. So for example, compare a Ryzen 5 3300X with a Ryzen 9 3950X:

https://cpu.userbenchmark.com/Compare/AMD-Ryzen-9-3950X-vs-AMD-Ryzen-3-3300X/4057vs4076

The 3950X, being five times more expensive, it doesn’t provide any performance benefit in one or two core tasks. On 8 or 16, it completely destroys the 3300X.

So if I were recommending any CPU 3 months ago, I would say, 3300X FTW! But now we have a dichotomy. The new generation will have around 25-30% performance uplift, also in 1-2 core loads. This means better gaming. But as usually they don’t release lower-end processors on the first month of a new architecture, now you have to choose: Should you get the 3300X or a 5700X?

I really hope AMD also releases a 5600X or similar as well, as this will make the price gap smaller. If that’s the case, a lower-end 5000 is probably the best choice for gaming. If not, the 3300X is probably the best gaming CPU for the price.

Graphics card

On this front there’s a lot of unknown yet as there’s not much information on what AMD is baking, and hopefully we’ll get a lot more data in the next 2-3 weeks. But basically, the RTX 3070 or similar GPU from AMD (probably to be named RX 6800 XT) are the most compelling ones in a 400-600€ price range.

Here, be aware that 1080p is going to be dead quite soon. I do expect 1440p to be the common resolution in less than 4 years. It has a noticeable increase in detail, and productivity wise is also really wonderful. Monitors have seen recently a huge change where IPS panels can get onto refresh rates and latency similar to TN panels, while holding color accuracy. And big 4K monitors are becoming more and more affordable. If you’re still using 1080p and going for a new graphics card, be sure to size it sensibly for 1440p, as it’s quite probable that you’ll find yourself upgrading the monitor before upgrading the computer again.

This is why I would not target RTX 3050 or similar when they come, unless for very constrained budget options. The 3070 is going to be the cheapest card that holds the bang-for-the-buck here.

I’m talking a lot about NVIDIA cards and not much about AMD, this is just because we still don’t know anything about them, and current offerings do not compete. But I believe AMD will come out with outstanding products this time around and if possible I would hold buying. AMD should have the same performance this time around, with less consumption which turns into less heat to be removed from the case.

I truly think it’s going to be good because NVIDIA moved off quite scared, pushed an aggressive pricing and released early even with no stock to offer. AMD is so quiet that sounds even overconfident on their product. As if it’s going to erase competition once released. We know from leaks that AMD “Big Navi” is going to be on par with the RTX 3080, and if they’re not going to rely on hype this time around, this probably means that it’s going to compete in price aggressively. Most people (including me) it’s quite loyal to NVIDIA, and launching so late it makes it hard to sell to most of us. So if they want to sell this late, they need to cut prices. We’ll see in a few weeks.

Hard drives / SSD

Bad news for hard drive manufacturers. SSD are really taking off, and there’s no point anymore on getting hard drives except for archival purposes. While my 1TB SSD SATA costed me around 300€, now 2TB NVMe is around 240€:

https://ssd.userbenchmark.com/Compare/Samsung-850-Evo-1TB-vs-Sabrent-Rocket-NVMe-PCIe-M2-2TB/3576vsm847986

This should be enough for most people for a few years, and it’s worthwhile waiting for getting extra capacity, as the price seem to be falling pretty quickly.

We’re starting to see now a new trend, SSD oriented towards capacity, not speed. This is going to become more evident in 2021-2022. This is a huge gap currently on datacenters and it’s most probably going to be available to the consumer in the next years. For example, I could find a cheap 4TB SSD:

https://ssd.userbenchmark.com/Compare/Corsair-Force-MP600-NVMe-PCIe-M2-2TB-vs-Samsung-870-QVO-4TB/m841469vsm1243154

This is around 91€ per TB, which is really, really outstanding. Speedy SSD are going to be NVMe, and capacity oriented ones are going to be either 2.5″ or 3.5″.

It turns out that slow SSD still leave mechanical HDD into the dust on performance, leaving HDD only for those that do RAID 5 or 6 with 4 or more disks, and for archiving only purposes.

https://hdd.userbenchmark.com/Compare/WD-Gold-12TB-2017-vs-Seagate-Barracuda-3TB-2016/m391336vs3898

Notice that we have 3x the SSD capacity at a price slightly cheaper. This is only 3.4x cheaper per TB, and I say “only” because it used to be way, way more.

Also, given the slower speed and that huge capacity, the amount of time required to backup that data is going to be days. Just a sequential read of the full disc is around 17 hours straight. If we’re going to copy files, not just byte-copying the platter, it’s going to be much slower.

I also don’t trust HDD manufacturers anymore. They’re being pushed really hard from SSD to be competitive, so they’re using everything they can to make their products compelling. This means rushed technology which could be faulty.

If you really need the capacity and cannot afford SSD, refrain from buying cheap HDDs; at the very least use a bit of the budget to get into reputable brands and models. If they have been independently reviewed, the better.