|
priznat posted:Fun fact: my co we would usually do green or black solder masks and we thought hey we want to have another colour in the mix so we can tell what project it is just from sight, let's make this one black. Hooo boy. The fab house had all sorts of problems with a black solder mask, it was totally bizarre. If they're not familiar with it it's a big hassle. We went back to just green/black after that, nothing fancy. That's not the first time I've heard about unusual soldermask colors being a problem for some PCB fab houses. I'm guessing there must be some differences in material properties between the different colors which need process tweaks to deal with, and if your fab house does 99% green, the ride might be a little bumpy.
|
#
?
Nov 10, 2022 21:25
|
|
|
|
| # ? May 20, 2024 07:52 |
|
|
Not sure how PCB manufacturing gets done but maybe there are some optical inspection/alignment steps and those get screwed up by different colors? Seen it happen with wafers plenty... if the wafer comes out a little too shiny, some tools will have issues finding fiducial markers to align to.
|
|
#
?
Nov 10, 2022 22:16
|
|
|
I’ve had impedance geometries change slightly based on soldermask color. The AOI I’d also believe could be an issue… have anecdotally heard stories of white / black soldermask causing issues in the reflow oven from a pure thermal emissivity standpoint.
|
|
#
?
Nov 11, 2022 16:15
|
|
|
hobbesmaster posted:I’m curious to see how Elkhart Lake does, especially with the FuSa stuff. I guess some people got through with “real time is no longer possible on x86” what's up with real time on x86?
|
|
#
?
Nov 11, 2022 16:29
|
|
|
Arivia posted:what's up with real time on x86? I used to develop Xenomai-based real time systems on x86 / x86-64 / ARM (ignore ARM for now) and any kind of platform that has poo poo like SMM and layers and layers of potential glue between you and hardware will / should have a healthy skepticism of actual achievable determinism.
|
|
#
?
Nov 11, 2022 17:05
|
|
|
movax posted:I used to develop Xenomai-based real time systems on x86 / x86-64 / ARM (ignore ARM for now) and any kind of platform that has poo poo like SMM and layers and layers of potential glue between you and hardware will / should have a healthy skepticism of actual achievable determinism. To elaborate a little, traditionally ring 0 was it. If you had a (hard) RTOS kernel running at ring 0 it could manage everything and give you deterministic scheduling. Virtualization extensions add a ring -1 for use by a hypervisor. That’s fine if you’re designing the system. You can use that or not. You still have full control. System management mode adds a ring -2 and here is where you have issues. When a system management interrupt fires the bios or whatever takes over and from the RTOS’s perspective a bunch of time is just lost. That means you no longer have a deterministic real time system. You can determine that SMM was entered at least. Still, that time was lost and your kernel doesn’t even have the opportunity to enter a design safe state until whatever SMM task completes. However SMM has been around since the 386. You could often disable (almost) everything that might cause an interrupt. This was considered fine for many uses. Intel Management Engine adds a ring -3. This has all the issues of bios entering SMM except you can’t turn poo poo off.
|
|
#
?
Nov 11, 2022 17:31
|
|
|
Rexxed posted:Gonna keep these SCSI cards going forever in PCI slots. HP ScanJet 4c will never die. Adaptec 2940 right next to my Sound Blaster X-fi, OG PCI forever. I think the SCSI card is close to thirty years old at this point.
|
|
#
?
Nov 13, 2022 19:40
|
|
|
JnnyThndrs posted:Adaptec 2940 right next to my Sound Blaster X-fi, OG PCI forever. I think the SCSI card is close to thirty years old at this point. I remember when I put one of those cards in a Novell server for work and it was a big deal because it was very expensive, we had to do it after hours so nobody would be accessing files in the late 90s. I think I got mine for $20 or something on ebay a while back just to keep the scanner going when my ISA card was a little too old.
|
|
#
?
Nov 13, 2022 23:55
|
|
|
Put together my last DDR4 build   i5s are back baby, they're good again
|
|
#
?
Nov 22, 2022 20:47
|
|
|
I wasn't really sure where to put this, since this is more of a generic CPU question: does either brand (or model/generation within the brand) perform particularly well when it comes to Paradox titles? A lot of the benchmarks revolve around shooters and other 3D-heavy games, and I'm wondering if there's some idiosyncrasy that would make, say, running Victoria 3 or EU4 better that wouldn't be reflected in such benchmarks. Is it still the 5800x3D? Still the i5-13600K?
|
|
#
?
Nov 28, 2022 12:43
|
|
|
would expect the 5800x3d to come out on top there, those sort of titles typically love the giant cache
|
|
#
?
Nov 28, 2022 12:50
|
|
|
repiv posted:Put together my last DDR4 build MemoryExpress has 13700K+DDR4 board+32GB DDR4 bundles for less than the cost of a 13700K+DDR5 board and I'm wondering if it's really worth the extra $200 to go for the DDR5-5600 or whatever option. A lot of what I've been playing recently is noticeably bottlenecked on my 8700K, as my 3080 is usually hovering around 70% utilization and I really want MOAR FRAMES so it's time for a new CPU.
|
|
#
?
Nov 28, 2022 12:50
|
|
|
gradenko_2000 posted:I wasn't really sure where to put this, since this is more of a generic CPU question: does either brand (or model/generation within the brand) perform particularly well when it comes to Paradox titles? A lot of the benchmarks revolve around shooters and other 3D-heavy games, and I'm wondering if there's some idiosyncrasy that would make, say, running Victoria 3 or EU4 better that wouldn't be reflected in such benchmarks. Is it still the 5800x3D? Still the i5-13600K?
|
|
#
?
Nov 28, 2022 12:54
|
|
|
That would be an interesting benchmark, yes. It's likely much more branchy and pointer-chasing code than the usual benchmarked games. Someone on reddit did post a benchmark of sorts comparing 7th with 13th gen:  That seems to point to the game scaling pretty well with number of cores. It might also be scaling because of memory BW improvements. On Stellaris benchmarks, the 5800x3D demolishes Alder lake, but that's not a guarantee it will do so for V3.
|
|
#
?
Nov 28, 2022 13:29
|
|
|
gradenko_2000 posted:I wasn't really sure where to put this, since this is more of a generic CPU question: does either brand (or model/generation within the brand) perform particularly well when it comes to Paradox titles? A lot of the benchmarks revolve around shooters and other 3D-heavy games, and I'm wondering if there's some idiosyncrasy that would make, say, running Victoria 3 or EU4 better that wouldn't be reflected in such benchmarks. Is it still the 5800x3D? Still the i5-13600K? aim for cache size and mesh cache I do wonder what Stellaris or KSP would look like on an EPYC with Infinity Architecture 3.0 Potato Salad fucked around with this message at 14:24 on Nov 28, 2022 |
|
#
?
Nov 28, 2022 14:18
|
|
|
Potato Salad posted:aim for cache size and mesh cache by this, you mean that something like a Ryzen 3600 is not necessarily going to be better than, say, a 10th gen i5, just because the Ryzen has on-paper a much larger cache, because the cache design is different?
|
|
#
?
Nov 28, 2022 14:21
|
|
|
Anything with Infinity Architecture 3.0 has, amongst other improvements, L3-IMC transport improvements. Pair that with high frequency ram and you'll see memory movement take up a lot fewer cycles in dtrace it's unreal how much pchem simulation, as a workload example that's bound to single core performance and memory architecture like Stellaris, benefits from this in the real world Potato Salad fucked around with this message at 14:32 on Nov 28, 2022 |
|
#
?
Nov 28, 2022 14:29
|
|
|
Kazinsal posted:MemoryExpress has 13700K+DDR4 board+32GB DDR4 bundles for less than the cost of a 13700K+DDR5 board and I'm wondering if it's really worth the extra $200 to go for the DDR5-5600 or whatever option. A lot of what I've been playing recently is noticeably bottlenecked on my 8700K, as my 3080 is usually hovering around 70% utilization and I really want MOAR FRAMES so it's time for a new CPU. I wasn't very convinced by the gains from DDR5 given how much it adds to the cost for now, my thinking is that by the time an 13600/13700+DDR4 starts showing its age DDR5 will be much faster than it is today and then will be the time to buy in
|
|
#
?
Nov 28, 2022 15:21
|
|
|
I just splurged today on a 13900kf because it seemed like a steal at $570. I was even able to ship it to a no sales tax address . I'm upgrading from a 2500k, more than double singlethreaded performance and 28 more threads lol. I figure I went more than 10 years on my last build, I might as well make sure I can do another 10 for this newer build. Can I actually air cool this thing, or do I basically need to go watercooling. I'm a power user and will be running multiple VMs at once and doing dev work. Except now I'll actually be able to do it well. https://www.amazon.com/gp/product/B00L7UZMAK/ref=ewc_pr_img_1?smid=A1Z5H6ZGWCMTNX&psc=1
|
|
#
?
Nov 29, 2022 00:31
|
|
|
Methanar posted:I figure I went more than 10 years on my last build, I might as well make sure I can do another 10 for this newer build. If you plan to keep it like this for 10 years, I would definitely get one of the largest air coolers on market. Many water coolers are not going to last a decade of 24/7 usage.
|
|
#
?
Nov 29, 2022 00:39
|
|
|
Air is best for longevity but I'd say watercooling for the usage. I have power limited a 13900K for a gaming rig and the NH-D15 is close to the limit of what it can do while remaining silent. If you are going to be using all those threads and want the performance, an AIO will soak up heat for longer before spinning up the fans. You'll probably have to replace it in several years but that beats having a noisy case in the meantime.
|
|
#
?
Nov 29, 2022 00:45
|
|
|
You definitely need a big boy air cooler, Noctua or Bequiet for example as a minimum. If you arn't overclocking you can use those kinds of coolers. I have a 140mm Noctua w/2 140mm fans and my 13700K is locked to 5.3ghz all P core and 4.3ghz ecore. I hit about 80-90C at most doing burn in tests. Good enough for my uses.
|
|
#
?
Nov 29, 2022 00:46
|
|
|
I'm more likely to undervolt than anything tbh. This thing is so much stronger than what I have been using its ridiculous My ambient room temperature is cool too, so I'll just go for air. Thanks 
|
|
#
?
Nov 29, 2022 00:55
|
|
|
Just FYI, userbenchmark is about as reliable a source of anything as The National Enquirer. Their benchmarks and weights are chosen with the sole purpose of making Intel look good and AMD look bad. It's actually quite funny.
|
|
#
?
Nov 29, 2022 05:44
|
|
|
Noctua makes static pressure case fans for industrial applications that aren't very expensive at all. I replaced all my cafe and cpu fans with them and even at low rpm they blow anything else I've tried out of the water
|
|
#
?
Nov 29, 2022 05:59
|
|
|
Potato Salad posted:Noctua makes static pressure case fans for industrial applications that aren't very expensive at all. I replaced all my cafe and cpu fans with them and even at low rpm they blow anything else I've tried out of the water If you're buying static pressure fans to run them at a low RPM, I'm not sure you're getting what you're paying for.
|
|
#
?
Nov 29, 2022 08:12
|
|
|
gradenko_2000 posted:I wasn't really sure where to put this, since this is more of a generic CPU question: does either brand (or model/generation within the brand) perform particularly well when it comes to Paradox titles? A lot of the benchmarks revolve around shooters and other 3D-heavy games, and I'm wondering if there's some idiosyncrasy that would make, say, running Victoria 3 or EU4 better that wouldn't be reflected in such benchmarks. Is it still the 5800x3D? Still the i5-13600K? I played a lot of launch Victoria 3 on a 5950x, and when the forums lit up with people complaining about the game simulation crawling to a halt in 1880 I had no idea what the hell they were talking about. The older games don't scale well with core count, but the newest ones certainly do.
|
|
#
?
Nov 29, 2022 23:01
|
|
|
gradenko_2000 posted:I wasn't really sure where to put this, since this is more of a generic CPU question: does either brand (or model/generation within the brand) perform particularly well when it comes to Paradox titles? A lot of the benchmarks revolve around shooters and other 3D-heavy games, and I'm wondering if there's some idiosyncrasy that would make, say, running Victoria 3 or EU4 better that wouldn't be reflected in such benchmarks. Is it still the 5800x3D? Still the i5-13600K? Blorange posted:I played a lot of launch Victoria 3 on a 5950x, and when the forums lit up with people complaining about the game simulation crawling to a halt in 1880 I had no idea what the hell they were talking about. The older games don't scale well with core count, but the newest ones certainly do. Yeah even my 3900x didn't see any of the slowdown people kept complaining about and glancing at the task manager shows that V3 will in fact utilize every single core you can offer it. That said, I did start to see very mild slowdown once I started using that AI mod, but the pop assimilation mod minimized that. Stellaris is the one that has the worst slowdown for me if I mod it even slightly heavily or use the Star Trek New Horizons mod on any of the galaxy maps.
|
|
#
?
Nov 29, 2022 23:44
|
|
|
I'd really like to see reviewers add V3 to their benchmarks. It would be interesting to see whether an 8-core X3D chip can outperform amd's or intel's biggest chips.
|
|
#
?
Nov 30, 2022 00:57
|
|
|
Now I'm wondering how it will do on a 400MB cache Xeon
|
|
#
?
Nov 30, 2022 09:48
|
|
|
BlankSystemDaemon posted:Sure, but they are not noticably more quiet when one compared at the equivalent static pressure, and even then they run at a higher RPM. Low at idle use, sure, but you best believe they run at high RPM when needed
|
|
#
?
Nov 30, 2022 20:28
|
|
|
Decided to put this here instead of making a separate thread in GBS or something. So, you know how you can buy some cars where all of the features are already installed, but they won't work unless you pay to activate them? (Either with a one time fee, or through a subscription service.) Intel is doing that now with CPUs!  https://www.tomshardware.com/news/intel-officially-introduces-pay-as-you-go-chip-licensing
|
|
#
?
Dec 1, 2022 07:35
|
|
|
Potato Salad posted:Low at idle use, sure, but you best believe they run at high RPM when needed Normally they're used to force sufficient air through heatsinks, between DIMMs, and more heatsinks found in servers where the fans are usually found close to the front and no other component has an active fan - so there it doesn't make sense to run them at a lower RPM.
|
|
#
?
Dec 1, 2022 09:00
|
|
|
Bula Vinaka posted:Decided to put this here instead of making a separate thread in GBS or something. God the comments section. 
|
|
#
?
Dec 1, 2022 09:07
|
|
|
Bula Vinaka posted:Decided to put this here instead of making a separate thread in GBS or something. As the article points out, Intel already did this over a decade ago with consumer CPUs https://en.wikipedia.org/wiki/Intel_Upgrade_Service
|
|
#
?
Dec 1, 2022 11:04
|
|
|
Eh. I see it as not having to pay for the dozens of features I do not use nor care about. It gets fused off over a network instead of the factory.
|
|
#
?
Dec 1, 2022 11:21
|
|
|
the defense of this is going to come down to "because Intel (and others) do not really manufacture different SKUs as their own discrete physical chip, then it ultimately wastes less 'sand' for the 'whole' chip to be sold to you, and then you pay for unlocking the parts of the chip that you can afford or want or need" and they're doing it to the enterprise segment [first], presumably because: - there is enough accountability in asset management in that space that you won't get as many problems with what happens to chips after they've been 'unlocked', but it changes hands to someone else - there is enough specificity in how enterprise customers use these chips, that it would actually matter to them to save however many dollars to turn off a feature on a CPU, multiplied over however many CPUs - and conversely, that such firms might actually take advantage of unlocking a feature that they didn't want before, but want now, which then yields some 'green' returns in the form of avoiding having to throw out so many otherwise perfectly functional chips in the upgrade the utopian answer would be that if they're selling you all that silicon in the first place, they might as well sell it to you completely unlocked, but of course el problema en es capitalismo
|
|
#
?
Dec 1, 2022 11:32
|
|
|
Beef posted:Eh. I see it as not having to pay for the dozens of features I do not use nor care about. It gets fused off over a network instead of the factory. You've already paid for them, you're paying extra on top of that to get to use them. Not sure why that's a difficult concept for some people.
|
|
#
?
Dec 1, 2022 11:48
|
|
|
ijyt posted:You've already paid for them, you're paying extra on top of that to get to use them. What do you mean with that? The whole idea is that Intel offers cheaper SKUs that have a bunch of special-purpose accelerators disabled. I don't need nor want to pay for DLB, because I'm not running DPDK on it. But at least with this on-demand thing I have the option of paying to enable it when I do. If you mean that you somehow pay for that hardware feature anyway: chips are not cars. It is astronomically more expensive to make a chip die that removes the circuits (IP blocks) for the relevant feature. That's why fuses exist and that's how SKUs are created at assembly.
|
|
#
?
Dec 1, 2022 12:49
|
|
|
|
| # ? May 20, 2024 07:52 |
|
|
Product segmentation generally benefits average consumers because it helps milk corporate (or just plain wealthy) customers more. The main ways I see this type of licensing scheme being a problem in the enterprise space are if it bleeds over into consumer space, and if it gimps hobbyists or small businesses who either are playing with older hardware or can't afford the license fees. If it comes to the consumer space we need to fight like hell, at least with two players it probably won't happen soon but at some point they'll push for it.
|
|
#
?
Dec 1, 2022 13:14
|
|