How to judge a Brazilian Jiu Jitsu school

Posted on January 31, 2026 by rthtgnbs

The martial arts business inside the United States, and frankly most of the world, is incredibly un-regulated. This means anyone can create a system, call themselves a founder/grand master/grand poobah and open a school to sell their training to the public.

One of the reasons that Brazilian Jiu Jitsu took off inside the United States is that it actually works. But, not always. BJJ is no longer the ultimate dominant martial art in UFC now that everyone cross trains in grappling.

But, not all BJJ is the same. There are different “lineages” and different training philosophies, and belt promotion criteria is often vague and opaque.

So what should you look for in a school?

First, most people choose the school closest to their home. If you do this, and the school doesn’t give you “the ick” as young people say these days, then great.

Second, if you have two schools within equal distance, then people tend to choose the one that offers classes that more align with their schedule.

Third, cost. Most BJJ schools are a significant monthly expense, so instructors who want more students tend to offer lower membership fees.

So how do you know your BJJ school is legit? These considerations are listed in an order, but they aren’t ranked as what is most important for you should be your number one.

First, if the instructors compete and have a public competition record going on years. IBJJF records and FloGrappling records don’t match up, so if they were heavy in one and not the other it isn’t a big red flag, but if you can’t find them at all in IBJJF for FloGrappling, that’s a big red flag.

Second, every technique should be taught how to do it safely first, not as an afterthought. This is an indicator that the instructor is placing on keeping students healthy.

Third, not every class is a fight to the death, but some should feel like it.

Fourth, not every class is an in depth on a technique or series, but some should be.

Fifth, smell. If the facilities smell bad at the beginning of the first class of the day, that means they weren’t cleaned properly the day before. Mats should be cleaned daily.

Sixth, injury rate. Are students routinely injured and training with braces and tape everywhere? You probably don’t need to put yourself in that environment.

Seventh, is the school part of an organization? There are a lot of organizations, Gracie Barra, 10th Planet, Checkmat, Six Blades, Lovato, Atos, etc. Most schools are a part of the organization where the instructor earned their Black Belt (or Brown Belt for some, a brown belt head instructor/gym owner is fine). EDIT: Shortly after hitting publish both Checkmat and Atos had scandals regarding inappropriate behavior from their senior instructors. This doesn’t make any affiliate schools guilty by association, and quite likely the affiliate schools are perfectly fine to train at.

Eighth, is there a good mix of “sport jiu jitsu” and “practical self defense” being taught? I’m not going to say there has to be any particular ratio, only that it isn’t all one and none of the other.

Ninth, is the instructor overweight for their age? If yes this could be a red flag. It could also be a medical issue or something completely unrelated to BJJ, but active BJJ practitioners tend towards being physically fit for their age.

Tenth, is the instructor consistently teaching or reinforcing a competitive ruleset? Priit Mihkelson famously teaches all of his students, even white belts, heel hooks. He also teaches the IBJJF rule set so that students know which techniques are allowed at their competitive belt level. Some “MMA gyms” ignore the IBJJF rule set and incorporate a lot of “catch wrestling” finishes into their grappling, and that is a red flag. Not that catch wrestling isn’t a legit martial art on its own merits (it is), just that calling it BJJ is dangerously misleading if that happens. Pulling a heel hook or “can opener” in competition is a quick way to both hurt your opponent and get disqualified at the same time.

Eleventh, if the instructor ever, even once, ignores safety rules with the explanation “there’s no rules in a real fight.” Huge red flag. You aren’t paying to get into a real fight, you are paying to learn a martial art, which might possibly at some point help you in a real fight. Real fights have eye gouges, single finger breaks, all sorts of bad stuff happening.

Twelfth. Does the school have a published curriculum? If it doesn’t, that’s a red flag. It could be a “white to blue” curriculum or all the way to black belt curriculum and that’s perfectly fine, but having none is a red flag.

Now, which of these are deal breakers? That is for you to decide. I would absolutely prioritize the safety considerations and hygiene considerations first. Learning from a top notch competitor is meaningless if you can’t compete due to staph or ringworm infections on your skin. Of course you can train in a perfectly climate controlled and sterilized environment and not learn anything because the instructor is a fraud, but a few internet searches on the instructor can usually warn a potential student ahead of time.

Leave a comment if I missed a red flag that’s important to you.

Posted in Uncategorized | Tagged , , , , | Leave a comment

Savage Axis Precision II vs Ruger American in KRG Bravo

Posted on June 15, 2025 by rthtgnbs

In the “budget precision rifle” arena there are quite a few good options, but at the low end of that spectrum sits the Ruger American and Savage Axis.

The Savage Axis Precision II is available for under 1000 US dollars after taxes at the time of writing, which isn’t exactly a small chunk of change. But it does come with an MDT Oryx chassis system, and if you had to buy a Savage Axis rifle at MSRP, and an MDT Oryx at MSRP, simply buying the package from Savage would save you a few dollars. If you could get the Oryx at 100 dollars off, and get a Savage Axis for bottom dollar, then it would make really good financial sense to go that route. I wanted a .223 Rem bolt action rifle that takes AI style magazines so that I could load 75 and 80 match bullets long and hot, and I thought the Savage Axis II Precision would perfectly fit that bill until I chambered an 80gr ELD-M load, and saw that the throat really is chambered for .223 Remington, and not a longer .223 Wylde or 5.56 NATO. I’ll run a few hundred rounds through the rifle to see if I need to open up the throat a bit, but we’ll.

The Ruger American Rifle has won praises from gun writers, shooters, and hunters for the performance gained from the price paid. The three lug, full diameter bolt is smooth to operate, the trigger is acceptable, and the twin V block bedding system makes it an easy rifle to make accurate. However the Ruger Precision Rifle (RPR), is over the thousand dollar mark, and I thought I could do better by getting a Ruger American Predator in 6.5 Creedmoor and dropping it into a KRG Bravo chassis. I did end up cheaper than the Ruger Precision Rifle, but not by much, and when I finally pulled the trigger on a KRG Spigot and Harris bipod adapter, the cost was essentially even, but my rifle weighs a few pounds less than the RPR and that does matter to me.

One thing both rifles needed was an upgrade to the bolt handle, which I did at a cost of about 60 dollars per bolt handle. I could have gotten by with the stock factory offering, but I didn’t like the feel and these are my rifles so I put a Scandinavian style “damn big ball” on the Ruger, and a longer tactical “teardrop” on the Savage.

Which rifle is “better” here is a really tough call to make. The Oryx is stiffer in the butt than the KRG Bravo due to the use of aluminum instead of plastic, but I don’t know how much actual difference that makes in performance. The Ruger American in KRG Bravo is lighter than the Savage Axis II in the Oryx, but only by about a pound (half a kilo for those in nations that haven’t put a man on the moon). I do like the feel of the KRG Bravo system a tad better than the MDT Oryx system, although I do like both better than say an “M16A2” style setup.

One interesting note, if you do put a Ruger American in a KRG Bravo chassis, you’ll want to super glue in your bolt stop pin. The normal stock Ruger uses keeps that pin from wandering loose, but the Bravo doesn’t. I used a dab of Loctite brand gel superglue and it seems to be holding.

Posted in ammo, defense, reloading, sniper | Leave a comment

AM4 as home or small business servers?

Posted on May 27, 2025 by rthtgnbs

The AM4 platform is a “dead end” but AMD is still producing Zen3 CPUs for the time being. This means that older Zen2 CPUs are getting pretty darn cheap. A Ryzen 9 3900 will give you 12 cores and 24 threads of compute power, which is quite often overkill for a homelab, and a Ryzen 3950 gets you 16/32, which is more than enough for a small business.

So looking at the benefits of Zen2 CPUs

Pros:
Cheap, because people still on the AM4 platform have one last upgrade to Zen3. The Zen3 had an almost 20% instruction per clock (IPC) uplift, so spending the money on a Ryzen 7 5700x3d or Ryzen 9 5900x makes decent sense for people who don’t need “all the framerate.”

Cons:
Dual channel ram only, which can impact some server workloads (think AI or LLM workload).
The ones with PCIE 4.0 capability have no integrated graphics, and the ones that do have PCIE 4.0 capabilities often only have one slot, and in some motherboards that is dedicated to a GPU.

The interesting:
The bandwidth difference between PCIE 3.0 and 4.0 is largely academic in the homelab space, where even 10 gigabit networking is just fine with a PCIE 3.0 x8 lane. Moving up to 25gb networking would require a full PCIE 3.0 x16 slot, or a PCIE 4.0 x8 slot.

The motherboard options (assuming you want a full ATX with 4x ram slots): Going back to the 300 series motherboards opens up some of the original Zen and Zen+ cpus which can be had for what feels like pennies (I literally purchased a Ryzen 7 1800 Pro for 45 dollars delivered a few months ago). These aren’t going to win any performance battles against newer CPUs from Intel or others from AMD, but for a file server, web server, media server, or game server (host, not remote play game server) they do great, at 65W max power draw which isn’t going to turn your IT closet into a sauna very quickly.

However, I wouldn’t recommend buying a 300 series motherboard (bios upgrades can be wonky) when the B450 and B550 “refurbished” motherboards come on sale for the 50 to 80 dollar range routinely. If you can wait, the B550 is the better choice, but is limited to Zen2 and Zen3 (Ryzen 3000 and 5000 series) only. The X470 and X570 top end gaming motherboards work just fine too, but I don’t see as many deals on them in the various online markets. The B450 motherboards often limit the maximum ram to 64 gigabytes, but the B550 double that to 128 gigabytes. For most homelab uses 64 gigabytes is plenty, but if the cost difference between the two is less than twenty dollars, I think the B550 offerings offer more value.

But, given the cost of purchasing a full size ATX motherboard for 80 dollars, a CPU for 80 to 120 dollars, 64gb of RAM for roughly the same…does AM4 make sense? Honestly no. The AM5 platform has been around long enough that the “premium” for moving to AM5 is largely a wash, and “refurbished/returned” B650 motherboards are showing up in the sub-100 dollar price range fairly frequently, along with low end 7000 series AM5 Ryzen processors.

So, if you already have an AM4 and are looking to upgrade your daily driver to a new PC, then I think repurposing it as a server makes great sense, something that was a Ryzen 5 3600 gaming PC would transition just fine into a home server (with an upgrade path to a higher core count Zen2 or Zen3 processor). Or parting it out to offset the cost of your new PC also makes sense. Getting in to the AM4 platform right now does not make sense, and I’m pretty sure I’ve purchased my last AM4 motherboards and CPUs, at least until a “screaming deal” comes along for pennies. Someone unloading a CPU, Motherboard, RAM combo for 100 to 150 dollars is pretty hard to pass up when you can toss it into a 90 dollar 4U rack mount chassis and power it with a cheap ATX PSU.

Things I really like about the B550 motherboards, the full ATX boards often come with two m.2 slots for NVME drives (one PCIE 4.0, the other 3.0), as well as 6x SATA ports. This would let you set up ProxMox as a hypervisor, run two cheap 1 terabyte m.2 drives for all your virtual machines, and pass through six spinning rust SATA drives to a TrueNas Scale VM for a zfs pool. You could even add in a 10gb NIC and pass that through to the TrueNas Scale VM for high speed network access while running other services through either the onboard gigabit or 2.5 gigabit port, or another 10gb nic.

Now, are these as feature rich as a purpose built Intel Xeon motherboard? No. Out of band management is generally not a feature on consumer oriented motherboards, and dual channel memory is all you are going to get. Unregistered ECC memory is supported with some processor/ram combinations, but that’s expensive and compatibility can be iffy.

In the end, where does the AM4 repurposed gaming PC fit in the server world? Well some place above the low power Intel NUC type mini-PCs and below freshly lifecycled off warranty enterprise gear. The difference between Intel and AMD is that you could in theory have purchased an x370 motherboard in 2017, and yesterday purchased a new in box Zen3 processor to drop into it, and Intel has no socket longevity comparison over the same time period.

Posted in computers, hobbies | Tagged , , , , | Leave a comment

The end of the two state solution.

Posted on February 24, 2025 by rthtgnbs

In 1979 Israel and Egypt made history with the first Muslim country making peace with the state of Israel. In that treaty Egypt gave up all claims to the Gaza strip in favor of the “Palestinian People” and Israel agreed.

In 1987 the Peres-Hussein London Agreement laid forth the way forward for the “Jordanian Option” where the West Bank would be recognized as Jordanian territory by Israel. Israeli Prime Minister Yitzhak Shamir vetoed the agreement. In 1988 Jordan formally renounced all claims to the West Bank in favor of the PLO as the legitimate representatives of the Palestinian people.

Twice Israel has had the opportunity to hand over the “Palestinian Problem” to Muslim countries, first Gaza back to Egypt in 1979, which they rejected, and then the West Bank in 1987, which the Prime Minister rejected.

The current “Palestinian Problem” is one entirely of Israel’s construction, and at multiple times Israel has agreed to the creation of a Palestinian State. In 1979, in 1994, in the Oslo Accords.

And yet….“Everyone knows that I am the one who for decades blocked the establishment of a Palestinian state that would endanger our existence.” – Benjamin Netanyahu

This is why Israel expands settlements. This is why the long process of “settlements” which even Israeli legal scholars recognize are against international law in terms of the UN Charter, are only growing. Israel officially calls the Gaza Strip and West Bank “Disputed Territories” rather than “Occupied Territories” in order to get around this illegality, despite recognizing those territories in multiple peace treaties and accords such as the Abraham Accords brokered under President Trump where countries like Saudi Arabia conditioned their assent on the creation of a Palestinian State.

For the world, the “Palestinian Problem” is that the Palestinian residents of Gaza and the West Bank have been prevented by Israel from self rule through military occupation and colonization.

For Israel, the “Palestinian Problem” is that they just won’t seem to leave the land and head over to Egypt, or Jordan, or Syria, or Mauritania, or anywhere else.

For the Gazan and West Bank non-Israeli people, they have no passports. Jordan already has as many citizens of Palestinian descent as it can reasonably handle for the next few generations (adding more is not a good option for regional stability). Egypt has already felt the influx of 100,000 Palestinians since the Israeli invasion of Gaza, and it is unlikely that Israel will allow those who fled the conflict zone to return.

Simply making the Palestinians citizens of Israel is a non-starter, after all Israel wants the land without the people living on it. The five million or so Palestinians would immediately dominate Israeli politics without some sort of additional apartheid inequality. After all, the only Knesset member to ever be stripped of immunity was an Arab-Israeli (the Israeli Supreme Court overturned that ruling, but the precedence has been set that non-Jews are not truly equal in Israel).

The unconditional support for Israel that President Trump has shown has emboldened the Israeli far right to openly advocate and plan for the annexation of the West Bank. Trump previously endorsed a plan for Israel to annex 30% of the West Bank, but backed off to support the Abraham Accords.

So what is next? Well it is likely that as long as Netanyahu remains in power, and it doesn’t look like elections will be held until October of 2026, it is likely that Israel will formally claim annexation of a substantial portion of the West Bank. Not all of it, not enough to deny the possibility of a Palestinian state at some point in the future, but possibly that 30% which President Trump proposed in his previous administration.

How will the world react? Well Israel has 5th generation fighter aircraft, 4th generation modern tanks, a substantial ground force, and small but professional Navy. None of the Muslim nations around Israel have those things, with Egypt’s aging fleet of F-16 fighters being more maintenance queens than screaming machines (hence the buying spree for European fighters not necessarily tied to Washington’s goodwill), and the massive fleet of T54/T55 Soviet era tanks being mothballed and replaced with export variant M1 Abrams main battle tanks. Jordan has even less, Lebanon and Syria essentially nothing but some Infantry forces at this point. Saudi Arabia has a large, well equipped Air Force, but largely no economic reason for attacking Israel over a breech of Palestinian territory or abandonment of the two state solution.

So that makes WWIII starting over the West Bank pretty unlikely. But it shows again what Vladimir Putin and Benjamin Netanyahu know implicitly, the political party that occupies the land by force gets to dictate terms.

Posted in Uncategorized | Tagged , , , , | 1 Comment

Information Operations, the Cost of Influence, and why the internet sucks these days.

Posted on February 12, 2025 by rthtgnbs

One of the more interesting things to come out of the Department of Government Efficiency (DOGE) is the spending of taxpayer dollars on media outlets like Politico and the NY Times. At least I find it interesting.

There’s a few things you should probably know about the internet if you don’t already know them.

First, there are several commercial and non-commercial software solutions to create and manage bot accounts on social media platforms. There is nothing illegal about this, only that these software suites, some capable of managing hundreds of thousands of unique online accounts, are often used deliberately to create “false consensus” or advance a particular idea or talking point, and because it is really just software, can be managed by anyone with a sufficiently powerful computer (or pay for a cloud service instance).

Second, influence is big business. As loathe as I am to link to it, this article at Common Dreams does a very good job of illustrating the actions of one Israeli influence for hire firm: https://www.commondreams.org/news/israeli-influence-campaign If you don’t like that article, this one from Haaretz is well worth reading: https://www.haaretz.com/israel-news/security-aviation/2022-11-16/ty-article-static-ext/the-israelis-destabilizing-democracy-and-disrupting-elections-worldwide/00000186-461e-d80f-abff-6e9e08b10000

Third, the unholy creation of Large Learning Model (LLM) chatbots has increased the quantity of influence content. It remains to be seen if it increases the quality of influence content, but for sure Artificial Intelligence can spew out large quantities of content.

Fourth, all of this is cheap. Really cheap. So cheap to replace content creators with AI, and use Social Media Account Management Software to update hundreds of thousands of bot accounts, while at the same time “news stories” pop up like mold spores after a light rain….you have the deterioration of the internet. Don’t get me wrong, there’s always been assholes on the internet (real racists, Nazis, pedophiles, etc), but a large part of the Web 1.0 era was content was human created, machine shared. Web 2.0 and the rise of “user created content” and “social media” changed that dynamic, and now “Web 3.0, the unholy rise of AI and malicious actors” means you should trust less and less of what you find online….

After all, how else could the “Flat Earth” movement be actually gaining steam? In an era when people have walked on the moon and currently orbit the planet?

Now, with all of those points out of the way, why on Earth would the United States Agency for International Development (USAID) need to spend millions on various media outlets? Honestly I can only believe it is because they didn’t understand exactly how cheap it is to buy favorable coverage, and were simply going about it the old fashioned way. After all, when you take the King’s penny, you dance to the King’s tune….and USAID and the State Department probably felt it was rather better to be the “King” than a target of actual investigative journalism.

So who would be the target audience for these types of activities? Well if you are funding Politico and the NY Times, you are probably aiming to influence those well employed taxpayers who hit the top of the standardized national test scores for reading and math (literacy and numeracy). And if you are buying online influence for social media, you are probably trying to hit the part of the population that is rapidly falling off the reading and math scores. For better or worse, America is a nation with a LOT of divergence in terms of education, occupation, and lifestyle.

For more information on the diverging academic outcomes: https://www.aei.org/research-products/report/testing-theories-of-why-four-keys-to-interpreting-us-student-achievement-trends/

I cannot prove that the internet is speeding up the decline of standardized test scores in America, but I strongly suspect it. Not just because of disinformation and influence campaigns, but because of the “dopamine hit” problems associated with social media in general. This exacerbates the symptoms of attention deficit disorder and other maladies of the mind. It also explains why so many successful “tech bros” are homeschooling their children.

So what’s the solution? Well I don’t know. The technology upon which the internet is built isn’t designed to differentiate between a human with dumb ideas and machine accounts spreading malicious information at machine scale. To the internet, information is information, just ones and zeros, and any meaning to those ones and zeros is really up to the humans to decide. Educating people on the dangers of massive, coordinated misinformation and influence activities certainly can be part of a solution, but I believe it is often easier to fool someone than to convince them they’ve been fooled in the first place.

Posted in Uncategorized | Tagged , , , , , , , | Leave a comment

2025….a quarter into the century.

Posted on January 2, 2025 by rthtgnbs

From 2000 to the end of 2024 has been 25 years. I got married, had kids, deployed to Iraq and Afghanistan, got gray hair, retired….it’s been a ride for sure.

Things that I’ve learned…

Play with technology. It is always changing/adapting, and if you don’t keep playing with it (building computers, testing software, implementing new features, etc) you’ll end up a helpless old person who has to hand their phone to a grandchild to update the alarm clock. So keep mastering tech.

Relationships matter, not just with your family, but your neighbors and other “tribe” you might have. The older you get the fewer opportunities there are to make meaningful connections with people who aren’t complete wastes of oxygen, so appreciate the good ones while you can.

Education is meant to help you hone your problem solving skills, not make you feel good. If you are paying someone to teach you how to feel a certain way, you are getting ripped off. You always pay for an education with time, money, or both. Whether you get a degree at some point is merely details.

Shooting, gunsmithing, and handloading are three separate hobbies. Being good at one doesn’t necessarily transfer to the other two….

Playing is more important than winning. I still have to remind myself to enjoy the game and not the outcome.

Posted in Uncategorized | Leave a comment

In Defense of the Radeon 7600 XT

Posted on December 25, 2024 by rthtgnbs

AMD used the same silicon in the 7600 XT as they did in the baseline 7600. This has led more than a few commenters to question what AMD was thinking, as elsewhere across the Radeon product line the XT versions generally get a few more Stream Processors as well as a slight speed bump.

The 7600 and 7600 XT are the “base” RDNA 3 gpus this generation. None of the super cut down RX 6400 and 6500 options from the RDNA 2 lineup. That being said, the 7600 and 7600 XT are definitely the “odd ducks” in the RDNA 3 lineup, where several of the previous RDNA 2 generation cards solidly beat them in raw performance. This doesn’t make intuitive sense to people who are expecting solid generation on generation performance. So what sense does it make?

In the RX 580 to RX 5500 to RX 6500 transition from Polaris/GCN4.0 to RDNA1 to RDNA2, consumers all got essentially the same performance, unfortunately also at the same price point (top of the line pricing for bottom of the lineup RX 6500). At the same time, an RX 5700 or 5700 XT will massively outperform any of the GCN 4.0 options, so the “top of the line” RDNA 1 truly was a solid upgrade option over the RX 580. But as much as a 5700 is an upgrade over a 580, a 7600 is the same level of performance increase over the 5700. And the 5700s debuted in the 400 dollar range, while the 7600 debuted at 269. If someone was holding on to the RX 580, the 7600 represents a great upgrade path, as the power draw requirements are essentially within 10 watts of each other (depending on who made the board).

As far as GCN5.0 aka Vega, AMD essentially abandoned the discreet GPU market in favor of RDNA1 and left Vega to be the onboard graphics solution for Ryzen processers up to the Ryzen3 generation on the AM4 platform. Not that Vega was bad, just that RDNA1 was the direction AMD was heading.

Complete aside, I’m thinking about this because I put an old MSI RX 580 4gb GPU in a computer build for retrogaming, and saw that it was pulling about 5w at idle on the Linux desktop. Sure I could do more efficient on the top end using something like a Geforce 1050Ti, but I’m using that one for media transcoding and had the MSI card sitting on a shelf (the RX 580, even with only 4gb of VRAM, is more than enough to emulate a Playstation 3).

So where does the RX 7600 XT fit in the current market? Well, it’s about the cheapest 16gb VRAM video card you can purchase new, the only competitor for that title is the Intel ARC A770 (which has a full 256bit VRAM bus which is double the RX 7600 XT’s). This doesn’t matter too much currently, but the trend is heading towards the reality that 12gb of VRAM is going to sit where 8gb of VRAM sat in 2020, a healthy bare minimum for performance. This is largely because the XBox at 10gb of VRAM and the PS5 with 16gb of VRAM, are driving PC ports that expect similar hardware. But what does more VRAM really mean? Better numbers on the 1% and 0.1% low frame rates as the GPU doesn’t have to fetch information from system RAM or storage.

Price wise, the Nvidia RTX 4060 8gb offering is really the only current gen competitor to the RX 7600 XT. Between the two I think the 7600 XT is more “future proof” simply due to the VRAM advantage. In the used market, RTX 3070s are getting down in price, and represent a great buy for those who want to stick with team green. The ARC B580 with 12gb of VRAM is Intel’s current offering in the price point, and honestly it’s hard not to recommend as even with current drivers is sending out really good performance for the price (and we can expect driver maturation to really help solidify its place in the performance rankings).

To sum up, the RX 7600 XT is going to give playable framerates with very good 1% and 0.1% lows for quite some time. It is as a price point where team green offers an 8gb card and team blue offers a 12gb card. I don’t particularly recommend the baseline RX 7600 8gb model to anyone purchasing new, as I do not believe it offers equivalent value to an ARC B580.

Where you should buy the Nvidia offering is if you need GPU compute. The CUDA ecosystem is mature and AMD is struggling to catch up, and making mistakes like abandoning older architectures prematurely while Nvidia is supporting back to the 1000 series (AMD abandoned all GCN and earlier architectures). As far as Intel’s compute software ecosystem, I have no direct experience. I expect it to eventually be as well supported as Nvidia’s but have no idea if it is useable today.

Posted in computers, hobbies | Tagged , , , , | Leave a comment

Offensive Cyber Effects and Electromagnetic Attack, very little actual overlap.

Posted on November 23, 2024 by rthtgnbs

To all the poor students at the Cyber Captains Career Course or Command and General Staff School struggling to write a concise paper on CEMA, he’s a crash course on why they are two separate things.

United States Code, Executive Orders, and inter-agency agreements are how “Offensive Cyberspace Operations” (OCO) are conducted by USCYBERCOM units. A radio frequency transmitter, attached to a piece of military hardware (or a radio frequency homing munition) is how Electromagnetic Attack (EA) is conducted by United States military units. For the purpose of this article I will only focus on one aspect of EA, that is RF energy emitted to attack a target receiver. There are other parts of EA, such as chaff and flares, lasers and high powered microwaves that just fry circuits, but no one is really making the argument that those are “cyber” effects.

Currently a rather simple test for whether an effect is OCO or EA is this: “Is the effect in my battlespace coming from a USCYBERCOM force structure unit co-located with the National Security Agency? If yes, then cyber, if no, then NOT cyber.”

The confusion exists only because there is no such thing as “Tactical Cyber” but if you squint really, REALLY hard, you can sort of make the case that EA effects, which exist at the tactical edge, are rather like OCO effects. This confusion exists mainly for senior Cyber officers to promise support “at all echelons” while not saying the quiet part out loud, which is the support below the Combatant Command level is going to be Electromagnetic Warfare (EW).

Does the effect require access to a processing unit in a computer? If yes, then Cyber, if no EA.
Does the effect end when the transmitter is turned off? If yes, then EA, if no Cyber.
Does the effect rely on a central processing unit running code? If yes, then Cyber, if no EA.
Does the effect rely on a RF spectrum receiver which is designed to separate “signal from noise”? If yes, then EA, if no Cyber.
Does the effect change information in the physical environment before it reaches a device? If yes, it is EA, if no Cyber.
Does the effect enter the target before an analog to digital converter (ADC) in the target do its work? If yes, then EA, if no then cyber.

“But what about Layer 2 of the OSI model?!?!?!” Someone will always ask. Even a digital signal, such a WiFi or Cellular, when attacked by EA will go back to functioning normally when the transmitter is turned off. “But you are flipping bits, exploiting how the protocol works!” They’ll follow up, but once again, from the receivers perspective that interference could just as likely have been natural. So, still EA, even if a wireless protocol that a device uses is the target of the EA transmission. And, the information is changed before a receiver ever pulls information the carrier wave.

“But, in theory, if USCYBERCOM would align service retained forces with the authorities necessary to conduct OCO then there COULD be tactical cyber!” One might very well argue. And this is hypothetically, highly unlikely, possibly true. But it is very much not going to happen, since the services who deploy forward into areas of hostility do not bring with them their own resources to conduct OCO to USCYBERCOM tradecraft standards. And there will never be a Soldier or Marine with M22 binoculars watching enemy tanks rolling over an intervisibility line calling for cyber fires.

That should clear it up. EA is not Cyber. Anyone who tries to tell you about the “overlap” is mistaken and doesn’t know what they are talking about.

“But what about cyber payloads for EA missions!?!?!” Ok, here IS the actual overlap, where an Electromagnetic Attack system creates a hardware handshake or data link with a target, and then DELIVERS a cyber payload. The EA mission wasn’t cyber, the cyber payload wasn’t EA. Just like when I rode in a USAF C-130 I was still a Soldier and not an Airman, and upon vigorous exit transitioned from the air domain to gracefully hit the dirt like two hundred pounds of air pollution. But it should be noted that EA can deliver all sorts of payloads, overt jamming, covert jamming, information operations payloads, all with different second and third order effects.

“Aha! SO there IS overlap!” A struggling field grade staff officer desperately trying to convince themselves that they have tactical relevance to a land owning joint task force, “We COULD use our organic EA assets to deliver Cyber effects and therefore EA is cyber!” No bucky, no more than a Soldier riding in a C-130 doesn’t become an Airman. It isn’t technical ability that is stopping a land owning JTF commander from having their signal and military intelligence personnel load up Kali or Parrot OS boxes and start hacking on networks inside the area of hostilities, it is policy and authority. The land owning commander has all the EA authority, and no OCO Cyber authority. It is what it is.

Tactical OCO is a myth, just as much as a Strategic Bomber can do “close air support.” (The USAF has tried to sell that one more than a few times to kill off the A-10 with the mantra that “CAS is a mission, not a platform!” Which is why they don’t really train F-35 pilots to do CAS….)

Posted in Uncategorized | Tagged , , , , , , , | Leave a comment

The Final Upgrade for AM4

Posted on November 11, 2024 by rthtgnbs

Years ago when I came back from a deployment to the middle east, I built a very modest gaming computer for my then ten year old son. A Ryzen 3 3200G, 16gb of RAM, and a SATA SSD in a cheap case, with a cheap power supply. This was just fine for playing Minecraft and other games he liked at the time.

Then he discovered Steam. And rather suddenly a four core, four thread, integrated graphics computer wasn’t cutting it. So I upgraded him to a Ryzen 5 3600, and a Radeon RX 560 (this was the height of the GPU shortage during COVID) and then eventually to a Radeon RX 580 after Ethereum was done mining. The Ryzen 5 3600 and RX 580 is still a viable option for entry level gaming at 1080P and medium settings for a lot of games, but then my son discovered “Helldivers 2.” The system would run, but the 1% lows were a little distracting and so upgrading was again on the menu.

One birthday he got a 1TB m.2 NVME drive, and we moved his Windows partition over there using Clonezilla. We also took advantage of the free upgrade to Windows 11. And another Christmas he got a 2 TB SATA SSD for game storage.

The move from an RX 580 to an RX 5500 was a “sidegrade” but the RDNA1 architecture still has driver update support and optimization from AMD, and I got the card for 80 bucks on Facebook Marketplace. This really helped out the Helldivers 2 performance, and my son was quite happy, playing at mostly medium, with some low settings, at 1080P.

Then I found a Ryzen 7 5700X3D on sale for a very reasonable price, and made the purchase. The 5700X3D is only a 3 GHz base clock with 4.1 GHz boost, but the massive level 3 cache really helps in some games. What we found was that by going to the Ryzen 7 5700X3D (after putting the 3600 back in to do a BIOS firmware update to recognize the new chip), was that the 1% and 0.1% lows really improved. Now the game plays a lot smoother, still averaging over 60fps on medium settings.

There is one final upgrade for my son, possibly a Radeon 6750XT or other RDNA2 based graphics card, or Nvidia RTX 3000 series gpu now that they are getting affordable on the used market, and maybe even an Intel ARC. But that’s it, any GPU beyond that RDNA2 or Nvidia 3000 series will be “CPU bottlenecked” even with the 5700X3D. That’s just the nature of desktop gaming.

But with that one last upgrade, I believe my son will easily get another two to four years of fun gaming out of his PC. That’s not bad at all for a CPU socket standard that first came out eight years ago in 2016.

And this points to a winning market strategy for AMD. From 2016 until now, AMD has only used two desktop socket standards, AM4 and AM5. The initial launch of AM5 was questioned by quite a few industry talking heads because AM4 was still being supported, had massive adoption, and was cheap. Those were all things that AM5 wasn’t, but AMD pushed forward, and the performance gains with the new 9000 series CPUs is enough that even people with the absolute best AM4 setups are considering the cost of upgrading.

So do I recommend buying or building an AM4 system in 2024? Not really, unless you are building a low budget homelab server where you need more IO than you’d get from a surplus corporate drone box. If I were building my son a new gaming PC today, it would be on the AM5 platform.

In comparison from 2016 until now, Intel has gone through four desktop CPU sockets. LGA 1151, 1200, 1700, and 1853. And while it is true that you can’t get some of the early 300 series chipset AM4 boards to work with 5000 series processors, Intel had that same problem with LGA 1151, as the 200 series chipsets wouldn’t support the next generation so you needed a 300 series chipset motherboard. This really, really separated out the hardcore Intel fans from the more pragmatic computer builders.

Posted in Uncategorized | Tagged , , , , | Leave a comment

Virtualization: Intel Xeons vs. AMD AM4 Ryzens

Posted on October 27, 2024 by rthtgnbs

In the homelab world it is a very common practice to buy last generation or even older enterprise servers and repurpose them into your homelab. There are a lot of advantages to this approach, the first being that enterprise grade hardware generally sees longer support timeframes (for example my current laptop surplussed “corporate drone laptop” is a 7th Generation i5 cpu, and HP updated a new BIOS December last year for CPU security). This longer support window isn’t always there though, some companies wipe their servers of any reference to abandoned products and “buyer beware.” Generally you are fine with Dell or HP.

On the AMD side of things, the consumer oriented Ryzen CPUs were never marketed as server chips, that would be the Epyc line, and even the Threadripper line was marketed at the “High End Desktop” power user. However, cpu architecture is what it is, and just like the Intel Xeons share commonality with their iCore workstation CPUs, so do Ryzen and Epyc. This means that for 65 Watts of power (which can often be configured down to 45 Watts through bios settings), you can get 6 core/12 thread or 8 core/16 thread Ryzen 5s and 7s on a platform fairly inexpensively.

For a simple homelab server, the 6 core/12 threads of a Ryzen 5 represents “72 vCPUs” if you believe in overprovisioning (vCPU = cores times threads). That’s enough CPU horsepower to run a pihole, an opnsense firewall/router, a NAS (using hardware passthrough for the actual storage), a media server (Plex, Emby, or Jellyfin). And if you choose something like a Ryzen 5 4650GE, you get to use error correcting RAM, at 35 watts power draw. The problem with the Ryzen Pro lineup is that they were OEM sales only, so you have to pick them up used. But even a Ryzen 5 Pro 2400 GE has 4 cores and 8 threads, at 35 Watts, even if it is on the original Zen architecture that is plenty of horsepower for running ProxMox or your particular hypervisor of choice (as 4 cores/8 threads = 32 vCPUs).

The biggest downside to going with a consumer AM4 system is that motherboards aren’t built with server workloads in mind. The most budget friendly only have two RAM slots, and often limited PCIE expansion options. A full size ATX AM4 motherboard is still a fairly pricey purchase compared to the smaller boards, largely because the market is willing to pay more for four RAM slots and four PCIE slots. So any potential budget savings going with AMD hit the iron law of diminishing returns, everyone is looking for value, and the market determines the price. One particular downside I ran into was IOMMU groupings, a B450 motherboard by Gigabyte would only assign the main x16 PCIE slot its own IOMMU group number. The other two x4 and x8 slots would get lumped in with other chipset devices, and I had to implement the ACS patch to artificially separate out the chipset devices from the PCIE slot devices. Enterprise grade server motherboards are designed from the beginning to support features like IOMMU, and so there is less “software hackery” to force the computer do do what you want.

On the Intel side of the equation, if you are buying used server gear, the downsides become noise and power draw. I honestly think if you are looking for a quiet homelab experience you look for used “High End Desktop” workstations that come with a full sized motherboard (even if it isn’t ATX compatible) and a lot of RAM slots. The biggest risk here is motherboard failure, as old Intel CPUs are dirt cheap, because they last longer than the motherboards.

I’ve done the “whole range” of homelab setups, from dual socket Xeon systems to permanently attached Celerons with a passive cooler to a couple different types of single board computers. What I’ve found is that the AM4 platform is a “sweet spot” in terms of power draw and performance (something that can be done with the equivalent generation Core chips from Intel if you want to go that route). I keep my “silent server” running a passive heatsink on a quad core Celeron and a pico PSU in my bedroom, the only moving parts are the two 12 TB hard disks in a mirror array for my wife. My homelab server is in my closet, running a Ryzen 5 4650GE, an ancient Coolermaster Hyper 212 heatsink keeps the temperatures down no problem, and 48 GB of RAM lets me run a PiHole DNS server, a TrueNAS SCALE VM with 4x 12TB disks in a 3 active, 1 spare RaidZ configuration (21.2 TB of available storage), and an Ubuntu 22.04 server that handles Jellyfin. Between all three VMs I’ve dedicated nine vCPUs (only one vCPU to the PiHole, as it doesn’t need more than that, 4 vCPUs to the other two VMs). This leaves me with plenty of headroom for things like a Minecraft server or virtualized firewall/router (I have exactly one PCIE slot still open on that server, a multi port NIC with passthrough to opnsense might be just the trick in the future).

Now, am I a little jealous of the guys running a full rack with 220v power distribution in their garage? Yeah, a little. When you have more you can do more, for sure. But I also still have an upgrade pathway in the AM4 platform, as eventually a Ryzen 7 5700G or Ryzen 9 will come down in price low enough that it could make sense to go to an 8core/16 thread or 12/24 CPU. I would have to go to 65 or 105 watts of power at that point, but with more you can do more. If I were much more financially successful I probably wouldn’t be building servers with “low power draw and very quiet” as my primary concerns. And to be honest, the guys running Intel systems seem to be winning the “most efficient homelab” setups on average, at least last time I checked.

Maybe in the future I’ll need to do some fluid dynamics calculations or “big data analytics” that require things like GPU clusters and “data lakes.” But not today, but it sure would be cool to build it if I could.


Posted in computers, hobbies, science | Tagged , , , , , , | Leave a comment