Be Excellent To Each Other

False advertising. As you say, it is not the product itself but the way it is demonstrated. It's being demonstrated with a completely biased "best case scenario". You see the performance figures and think "wow, this is the fastest gaming CPU on the market !" and then you go off and buy one. However, the board you have can only run it at the guaranteed base clock, so you are actually getting 50% of what you saw on Youtube.

By this stage you may well be thinking "So it's down to the guys on Youtube?" but most of the time they do completely as instructed by the company providing the review sample.

If reviewers sat down and explained this before reviewing the product (IE completely open and honestly) then most people would buy a Ryzen. Mostly because the guaranteed base clock is actually higher than the Intel, and it is much cheaper to boot. IIRC the guaranteed base on both is 3ghz, 200mhz higher than the Intel. But all we see are best case scenarios....

It used to really piss me off too back when the Core 2 Duo etc were around, because you would read a review showing one being overclocked to bejesus yet you then buy that board and CPU you saw in the review only to find the reviewer was using the fastest RAM available that cost three times what yours did and because of that you are heavily restricted on the FSB meaning your CPU is running 900mhz slower than the review you saw.

Pawns. These people on Youtube are just pawns. They do as instructed by their master, and the money fame and popularity ensues.

Whereas Adored? he's "doing a Stu". IE - telling the truth and not giving a shit whether he ruffles some serious feathers.

Why can't he just say that then? Preferably in a 500-word article.

But then you wouldn't get to hear his lovely Scottish accent.

Seriously though, whilst I take the point that some people would rather have this sort of material written down, I personally find I can consume a lot more information via videos, simply because I can have them playing and listen to them whilst I'm doing other things (mostly when I'm in the kitchen, as we all eat different things at different times, so I tend to spend a of time in the kitchen), and that's information I can consume that I wouldn't otherwise.

I think that half of the problem is that Jim covers very complex issues. These issues can not be simply explained in a minute or so, because then people would just say "He made that up and pulled the figures out of his arse". Usually when people complain about him they have not watched the video itself. I see that a lot. Then you ask some one to make a nice cup of tea, sit down, watch, and actually digest the information he is giving to you. If you are capable of understanding it? it's very valuable information.

Any one who speaks out about anything these days is labelled a hater and a conspiracy theorist. It's actually amazing how people just don't want to hear the truth now, and will refuse to accept it no matter what.

Let's take the past few days for example. Intel CPUs are called out for being insecure. First information is that basically the CPU has a fault that has been there for ten years, which allows a hacker to walk straight into your kernel, allowing the hacker to gain all of your private info. Now do note, this is Intel. Not AMD.

Intel make a press statement and immediately start slinging shit. Right at AMD, who they claim are just as bad. But the truth is they are not. Because as any one who has spent a few hours reading and digesting the information will know, only one of these vulnerabilities even applies to AMD. IE - Spectre. Meltdown can not touch AMD because the fault with the Intel CPU is not there. However, Intel post their press statement and then within hours you see stuff like this.

"Gotta love the US sue everything policy ugh. AMD and the other one are equal 'at fault'. Not a fan of these sue tactics personally."

And immediately they have made some of that shit they are throwing around stick. Now this post (the line above) was said before Spectre became apparent. At that moment it was only and specifically an Intel problem but instead of being told that the shit was already being hurled around. Then if you read the actual facts it turns out that Spectre can be an issue for AMD, but only in specific circumstances and AMD can patch it out without any performance loss.

Intel on the other hand? I wouldn't want to be a datacenter owner, that's for sure.

So most of the time the short answer is not good enough. And that is why Jim talks on as much as he does, so that you can fully understand it *and* see where the info and facts are coming from. And that isn't easy. However let's be honest most people these days just want the short answer.

You literally are.

No, that's literally the opposite of the issue. Video doesn't allow you to read and digest at your own pace, doesn't allow you to link to other sources for background reading or additional information. Complexity is an argument against video as a medium.

If only there were a way to pause a video, or rewind it to listen to a particular bit again.

If only people like Jim would put a load of links in their video descriptions to direct people to further reading and reference materials.

IF ONLY.

Who's pausing a video to copy down an on-screen, unclickable link?!
And yes, you can pause and replay. Can you slow down or speed up without it becoming nonsense?

I'm not saying it's useless, I'm just countering the ridiculous idea that people don't like these sort of videos because they're 'too complex'. FFS.

Oh I agree with that completely. I would still rather see complex technical things written down in print. I might even suggest to Jim that he does that.

I first saw the issue with Youtube on South Park. Kids these days all live on it, and don't even bother to read any more. So most of the time you are wasting finger energy... And the shit they watch? crikey. 8 million views for some spoiled little twat going out cruising in his Lambo.

Oh yes, computers !

So for about seven months now I have been happily running my upgrade. 14 core 28 thread Xeon that I have bus clocked to gain an extra 50mhz on all cores, as well as a force clock using the Genie to make certain cores run at their single core speed. Basically in a nut shell this forces four cores to run at 3150mhz, two after those ot run at 3050mhz and then all of the remaining cores to run at 2850mhz. Great ! and leaves a Ryzen 1800x in the dust when threading which is exactly what I was after. It cost less, too.

Any way, back in the summer I also bought a Titan XP (which has not lost a single penny in value since then which is awesome) and I soon got very tired of the air cooler. It was loud, and the back plate got obnoxiously hot. So I full out water cooled it on a slightly larger than single 120mm rad.



It took a couple of weeks of planning, measuring and mocking up in order to get it to fit. I had to take the slant degree of the case into account before doing anything, and then trying to make everything run vertically so it wouldn't leak all over the place and the res not be pissed at 60'. Yeah, wasn't easy but I managed to cram it all in.

Sadly I didn't realise that every review of the Titan XP under water were using a double rad. I simply didn't even think about it. Well when I finished it I realised why they had done that. Basically on a single rad I was reaching temps of 88c under water. No throttling, but yeah, pump is only rated ot 60c water temps and even the hoses became rather soft and a little floppy. That was during summer though, and it was a very hot summer down 'ere. So I decided to lower the power limit, lower the overclock to 200mhz and ended up getting 1900-2000 depending on the temps. This is not possible on air, your max boost is around 1800mhz depending on temps so I was happy I guess.. However, I wasn't happy that it was still hitting mid 60s.

So I did what every sane person would do, bought a lovely big old pile of shit to improve it.



240mm rad as opposed to 120, top end rad too (Bitspower) to match all of the fittings and res, and new fans that have lights in them. First time I have used LED fans in many a moon. I also ordered a USB computer (Poweradjust there) with a module that allows you to run 9 fans or pumps and control it all within Windows. I then bought a heatsink for the computer within a computer.



Because I hear they run quite hot, and then many fan extensions and a power extension.



Then I pulled it all apart and drained it.



Coolant was OK, but you can see here why you don't use a single rad on something so hot. It doesn't allow for much coolant at all really.



At least I saved it, it is rather expensive at £12 a bottle. Only a little bit of evaporation too over the past seven months.



Not bad considering how hot the GPU was running. I will be tackling it all next week, but, if I did my maths correctly it should all just literally slot in and I shouldn't have to use any new hose. The rad is the same thickness, same width and the same spacing between the G 1/4 ports. All I need to do is flush it all with distilled vinegar then I can build.

Can you do a video and put it on Youtube?

I'd prefer a 500 word article.

500 word video?

South Park.

That's the best of both worlds; a video showing someone typing their opinions out in a Notepad window.

I think I've seen a few videos *exactly* like that

Yeah me too haha. I think they obviously don't have confidence in their voice. They still go too fast

To be fair, whenever I've seen it done it's been quick and dirty stuff rather than lengthy lectures so it's probably someone that doesn't have a mic but can grab screen recording software easily. I've never seen anyone use that technique to present a lot of text, as that would be properly stupid.

It's annoying when you are following something verbatim and they open something by mistake lmao. Straight back to Google....

Oh, for sure. And if they don't put any required commands into the video description box so you can c&p then it quickly gets really stupid. On balance though I've been very grateful for some of these kind of things in the past and realise that YouTube gives people an easy way to 'publish' something that doesn't require too much effort on their part.

Some big hits to home PC performance with Windows and BIOS updates applied to mitigate against Meltdown and Spectre.

How does your SSD performing 41% slower sound?

These are all Intel CPU benchmarks, the chap will be doing AMD benchmarks shortly.

It may be that AMD are hit a lot less or not at all, since their CPUs do this branch speculation stuff in a fundamentally different way (as I understand it), and are only vulnerable to 1 of the 3 exploits anyway.

My PC is toast in a really interesting way: Windows has marked various chunks of RAM, splattered widely across the address space, as bad (ie put them in badmemorylist in the BCD) but memtest86 reports no errors on a 90 minute (and counting) run. Current working hypothesis is an L1/L2 cache fault resulting in intermittent memory read/write errors, which Windows runs for long enough to detect from time to time but memtest86 hasn’t (yet.) This exhibits as an error when attempting to install Windows updates at boot time with the message “cannot allocate ramdisk”, presumably because it’s trying to find a large chunk of contiguous memory. Either way, I’m fairly sure it’s deader than disco.

Sounds like your upgrade just moved up the schedule!

Still going for Intel?

http://www.businessinsider.com/linus-to ... tel-2018-1

https://lkml.org/lkml/2018/1/3/797

Linus Torvalds isn't very impressed with them.

To be fair, the random crashes I was seeing was what prompted me to start pricing out systems this week. Now I have a better handle on what’s causing them though, and it appears to be a fatal wound!

Wow those SSD drops suck

I'm really hoping that my M.2 doesn't take a nose dive. It shouldn't because it's not NVME but yeah, a little worrying.

I did have this all explained to me by a guy who writes for CPC (Gareth Halfacree). He said that basically imagine a bike that you ride that crashes every time you turn left. So basically you have to learn to ride that bike only turning right. Eventually you will get to your destination, but it will take longer. That's the linux side (he writes Hobbytech).

And that may explain why the SSD is taking such a battering in performance. Ugh, trust me to go bloody Intel and be pleased about it.

I guess the fallout of this will be being noticed for quite a while to come yet.

I remember him from some of the uk.comp newsgroups 15+ years back. A very nice chap indeed.

Yeah Gareth rules. TBH it's probably one of the only reasons I still buy CPC. The Hobbytech section at the back is fantastic

Subject if not thread relevant -

Nice to know that android is fine with the latest security patches.

However my Moto g5 is still on January 2017 because it turns out Motorola as a Lenovo imprint is disgustingly lax. I don't WANT to throw our phones in the bin but I was already getting twitchy before this...

Sure, the Oreo rollout is coming any day now, but will it actually be a remotely recent patch level, and even if it is, it's likely to be the final update.

(Ah, it's an a53 so not allegedly vulnerable anyway)

Re: Meltdown/Spectre

A good nontechnical writeup of how they were simultaneously discovered by four independent teams: https://www.wired.com/story/meltdown-sp ... discovery/

A fantastic writeup from 2005 when the Xbox 360’s PowerPC CPU suffered from a very similar issue: https://randomascii.wordpress.com/2018/ ... -xbox-360/ (don’t miss the final sentence about the job interview either... Plus ça change.) The bit where they make it stop crashing by changing code that was never run is just ::Italian chef kissing two fingers gesture::

A description of Retpoline, a Google-developed binary patching technique which protects against the bugs and claims to have almost no performance penalty (I have no insider information here, I’m quoting the doc): https://security.googleblog.com/2018/01 ... cpu_4.html

The whole thing has had me thinking a lot about this seminal essay by Joel Spolsky: https://www.joelonsoftware.com/2002/11/ ... tractions/

Interesting. I note they also say they've rolled out the KPTI OS patches in addition to that mitigation.

I'm not reading all that. Can you make it into a 1 hour YouTube video please?

I don't think PCs should be triangular.

Just about finished patching a 200 CPU farm running on SUSE

This is normaly used by scientists 24\7 so would have been a good test to see if there were performace drops

However they all like to use the GPU farm these days, so had to ask them to run some benchmarks and come back to me

There is a plan to increase\replace some CPU this year so interested to see what performace impact there is as it will affect what we need to buy.

Maybe I'm just an old-fashioned arse, but the idea of the CPU running code that it hasn't (yet) been explicitly told to run is horrifying to me.

Well, to be fair, branch prediction and speculative execution is decades old and a significant speed up compared to not having it so... It’s worth noting that Itanium apparently doesn’t suffer from this because (I’m told) it enforces thread-level memory protection on cache lines. But the market didn’t want Itanium, it wanted more teetering hacks piled on top of x86.

Don’t forget your Intel CPU is also running an entire OS in microcode, too, with remote access and everything. In fact there’s three whole layers of stuff under your kernel you can’t see: https://en.wikipedia.org/wiki/Intel_Management_Engine The days of CPUs dumbly chewing through in-order ASM instructions ended 20 years ago, we just didn’t notice at the time.

Or as my old internet buddy put it on the day Meltdown was announced

https://twitter.com/mjg59/status/948820694376173569

Good morning to everybody in Europe waking up to discover that computers (a) don't work and (b) have never worked

— Matthew Garrett (@mjg59) January 4, 2018

It'd be fine if Intel didn't design their CPU architecture as Linus Torvalds gracefully phrased it, like 'shit'.

Why does a peanuts character have an opinion on transistors?

Good grief

Let us note that Mr Torvalds has spent the last 20 years writing kernel code to take advantage of the performance gains provided by that exact CPU architecture and hasn't had cause to complain until now.

Maybe he was running with the assumption that Intel's CPU would appropriately protect restricted kernel memory areas?

There's nothing wrong with the concept of speculative execution (it's a very good idea, in fact), it just needs to be implemented properly.

Also Linus is an abusive bellend and his opinion that something is shit is basically meaningless. See e.g. https://mjg59.dreamwidth.org/38136.html

Then his assumptions are foolish. You don't work that close to the hardware level while making assumptions about how the hardware works. I'm not defending Intel here, I'm saying Torvalds views are (as they usually tend to be) opportunistic finger-pointing.

Five year old PSU never opened..



Opened.



Not as bad as I thought. Cleaned it all through.



Put back together.



Caps were all fine. Then worked until 2 A.M removing everything, cleaning and then putting back.



Just a Quadro in there for now so I can listen to my music



Not seen it this clean since the day I got it over two years ago.

Meltdown / Spectre like flaws on the Xbox 360 chips -found and resolved a long time ago (interesting read)

https://randomascii.wordpress.com/2018/ ... -xbox-360/

That's almost as good as when Gaywood posted about it!

Apologies - i missed that

It did also take me to this which i also found interesting (hardware bug on the original PS1)

https://www.gamasutra.com/blogs/DaveBag ... g_Ever.php

Anyone's work patching the clients yet?

In addition to the Windows patch levels Grim posted on this thread, I was told that we may need to do Bios update as well

Have a call with the client lead tomorrow to see if this is true.

One silver lining to this is that all the fence sitters in IT that were blocking the deployment of Anniversary (because it took so long to deploy) can now tell users its Intel fault for the disruption so we have a green light for that at least.

Speculative Execution would be a great name for a metal band (and a logo with nearly unreadable lettering)