Search the Community

Does disabling Simaltanious Multitreading and or Hyperthreading hurt preformance in other games more than just marginally? So I just found out that I have a game that dislikes multithreading. It runs well on my Laptop with a i7-9750h but not on my PC with a i7-13700kf. Both have Hyperthreading enabled, so I am guessing that the 13th gen has SMT where to old 9th gen does not. This brings me back to the question above. Also if I disable SMT how badly will that impact other programs out side of gaming? I apollogies if this is not the right place to post this topic. I was not certain since it kind of covers a multitude of areas.

I have an AMD Ryzen 3 2200G and a Gigabyte B450i PRO. SMT isn't enabled, and when I go to enable it, there just is "Auto" and "Disabled." Nothing I try works, and I'm tired of having CPU usage pinned at 100% and my game stuttering when playing even a moderately demanding one.

I saw a build on PCPP a while ago where a guy bought a 3950X for his workstation and then disabled SMT. He claimed it had given him benefits. I don't remember what the benefits where but I assume it would be power consumption, frequency and probably single core performance. I had questioned his decision but didn't put a comment on his build. Clearly his workstation's task didn't benefit from SMT, but then I was wondering what are the specific things you lose and gain when you disable SMT? What is the difference in performance of a 16C/32T chip turned into 16C/16T vs a 8C/16T chip. From what I can tell, at least on Intel, the other thread uses silicon/resources within the core not being used by the other thread. For example, the FPU, if used by one thread can't be used by the accompanying thread. But that is on Intel, does each thread have a FPU and ALU in Zen 2, or do they compete just like they do in hyperthreading? (Then I remember AMD FX where there was 1 FPU for every 2 cores iirc) Like do you lose access to some cache. I really wish I could read+understand the block diagrams for Zen 2 on wikichip, I even looked up the block diagrams for Coffee Lake refresh since some of the products already have half of the threads cut off, and it would prove the differences w/o testing. (9900K has 16MB Cache, and the 9700K, which doesn't have hyperthreading, had 12MB cache. Actually it says MiB, which is mebibyte, but doesn't everyone just say megabyte in this context anyway?) Obviously Zen 2 is well known for their multi-threaded performance, so why would he choose to throw that away beyond the fact that his workload wasn't benefitting from SMT? (This all assumes that we are locking the frequency at say, 4.1 GHz)

Hi All Following some issues with my current setup, I've got a 3950x on the way as well as the majority of other pieces and I'm getting things prepared in advance. One of the things I'm looking at doing is settings core affinity for specific programs I'm using frequently or require more compute to run optimally. Here are my questions: What is the most reliable way to find your best performing/consistent cores/threads? Do you run a single threaded benchmarks on each core/thread and see which scores highest? Whilst researching I found people that claimed a performance/stability loss if setting core affinity to threads that weren't matched to the cores they are paired with. They said something about it being better to match them as this way it uses the same allocated cache. How do you know which thread belongs to which core (using a quad core for example)? Is it something like CPU 0+1, CPU 2+3, CPU 4+5, CPU 6+7 Or is it something like CPU 0+4, CPU 1+5, CPU 2+6, CPU 3+7 Is it possible to stop Windows from using specific cores/threads for programs you want to ensure run at their highest level? Eg, I set a game or work application (I don't game when working so I can be safe knowing they will not be operating at the same time and overload these cores) to run on 6 specific cores/threads, I don't want Windows to ever load these 6 cores/threads with tasks in the background etc. Ideally I want to use the best performing cores for this, so if they are say 0-5, can I reserve these? Alternatively if this is not possible, then it sounds like I could set them to use the last 6 cores/threads on the CPU as Windows wouldn't load these unless there was enough work to require it, or is Windows indiscriminate and will try to load cores evenly? Here are some example issues of what I am doing to try and keep things clean/stable: As above, test to find which cores/threads are best and allocate to programs Overclock and tighten memory timings Undervolt and overclock CPU+GPU Run tests on performance matching infinity fabric with timings etc Raid 0 the 2 or 3 drives Actually another question if someone knows? I've bought the Aorus Xtreme x570 and in my haste realised the m.2s are split between CPU and chipset. Is it possible to raid 0 all 3 or can I only raid the 2 on the chipset? I don't believe I had an issue in other builds but they were using gen 3 and I'm not sure if they were split between CPU and chipset. This is mostly just wanting to test things out and play around/learn, I know that it's not really necessary for what I'm doing to raid0 gen 4 drives. Raid backup the install onto an SSD. Keep installs of programs tidy so that I can more easily test how programs/background tasks are impacting workload. One of the main reasons I started researching core affinity. I love my Razer Orbweaver, I have a Corsair Scimitar I quite like and a decent Corsair keyboard. I plan to use RGB Fusion to handle lighting (I'm not a fan of RGB but white/very pale blue I like along with the occasional change to switch things up) and may have missed out some others. This will leave me with at least 3 different pieces of software for lighting/peripherals and having watched a good amount of videos on this along with GN's extreme look at this, you can see that they're a problem with hampering performance in terms of lowering averages but can cause major stutters. I hate iCUE with a passion and Synapse can be a pain. I looked at AHK for reallocating keys on both but I couldn't find anything concrete saying that they work well due to lack of onboard memory. It would also be a bit of a pain to have less control on lighting software. Ideally with these kinds of software I want to allocate them to a specific core/thread that will affect more intensive programs the least. I will be testing before/after installing them as well as combinations to see how performance is affected (hopefully after spacing out the programs with affinity) and if it is still too affected then I may get rid of my keyboard and mouse as I'd prefer to go back to a Naga anyway.

Does CompTIA not know about SMT or am I missing something?

Budget (including currency): $3000 Country: US Games, programs or workloads that it will be used for: Want Base as Linux - version TBD - this will host two passthrough Windows VM's for Windows 10 gaming. All other VM's off this box will be either Linux or Windows Server 2019 Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): Use existing drives/case/etc. Parts RTX2070 (2) for Windows 10 virtualization https://www.newegg.com/asrock-rack-romed8-2t/p/N82E16813140044 Epyc Server Board - see elow - think option 2 is bette EDIT - Decided on the same Gigabyte MZ32-AR0 that Linus used - has IOMMU in BIOS whereas ASRock does not. Has MORE memory slots. Has most of the features I need (not as good as ASRock) https://www.newegg.com/amd-epyc-7402p-socket-sp3/p/N82E16819113592?Description=epyc rome cpu 24 core&cm_re=epyc_rome_cpu_24_core-_-19-113-592-_-Product 24 Core Epyc CPU So this motherboard is more expensive than the Gigabyte motherboard Linux used. This one has all 7 pice x 16 gen 4 slots, 2 oculus, 1 thunderbolt, etc. Adding Adaptec 71605 - not sure if I will do passthrough to Windows Server 2019 as all drives attached are on current Windows Server 2019 implementation Using 45 drive case to replace 36 bay that had motherboard, aIso will need to buy a new case for the ASRock motherboard Questions: 1) I cannot really afford all 8 sockets of RAM to be populated - can I start with say 2 or 4 memory sockets populated? 2) Did Linus ever get around the SMT issue? If not, they could be a deal breaker for me. I think the Threadripper TR3 has a workaround for VFIO and SMT, but there are no 7 pice 4 7 slot motherboards. I would be fine with 7 slots provided maybe 3 were pcie 4 and the rest were pcie3 so I can put in expansion card 3) Thoughts for a case? 4) Should I watercool? Will that lower the overall temperature in the room? Or which is the ideal air cooler? 5) Any good tutorials on VFIO? 6) For the motherboard provided above, how would i find out the IOMMU groupings and if this will do what I want 7) Any ideas for a case? I do NOT need any drives in their except the ones populated on the motherboard as I will hook up to external 45 bay All help is is sincerely appreciated. While I have been running Linux, I want to cut down on physical Windows machines (ideally want to get off Windows eventually completely). New to VFIO and GPU passthrough. Timothy Atwood

Wassup people, I got my Ryzen 5 1600, have overclocked it to 3,8GHz and wanted to try out games without SMT, to see if there'd be any difference. After concluding that there was almost no difference, i tried to re-enable SMT in the UEFI (Asrock AB350Pro4 on the latest bios, 2.2 i believe), and even when toggled ON in there, it isn't in Windows. Does anyone else have this problem? Does anyone know the solution to this problem? Cheers! also, pic related, it's ENABLED in the UEFI but only shows 6 cores (logical/physical) in windows.

Just curious, is there any actual reason why a 4C/8T chip, for example, should be more expensive than a 4C/4T chip of the same clock speed, architecture and all that? Do they cost more to produce, or even produce any more heat? Or is one thread per core, like locked multipliers, just an arbitrary limitation?

I was watching the "Use your Gaming PC's Extra Power as a NAS Ultimate Guide" video and Linus brought up the "hyperthreading" term when explaining what CPU you would need for a desktop/NAS that supports virtualization. Now I was searching about the interwebs and was seeing that Intel uses this technique whereas AMD uses SMT technology. Can anyone give me a clear description between the two if there are any differences or if they are similar?

I have a little problem about my single core performance. So i decided to disable my SMT from auto to disabled. What i want to achieved is stable framerate, less stutter. Now half of it was solved. From 12 to 6 core. Games can use my cpu with more usage. But the usage is higher than before. I will do this until next upgrade to 3600 (already asked more than twice) about this solution. Is it okay if SMT off? (mostly used for gaming)

What are the differences between Intel's hyperthreading and AMD's simulatneous multi threading?

Could this come from a motherboard or gpu? The little white thing, it seems to be plastic.

I'm wondering what the current thinking on these settings are for gaming. I know games aren't threaded enough to benefit from smt on high end processors, but I'm wondering if modern games and windows 10 work well enough to make sure to send game threads to separate physical cores and thus not lose performance from poor scheduling with smt on. And similarly, what about intel speedstep? Does windows 10 recognize games are running and force the cpu to the highest p state? Is there any concern of cpu throttling and reduced performance in game from poor power management?

I tried to find an approriate thread but very few places cover this. SMT and Hyperthreading can seemingly reduce single core performance. Many CPU dependent games prefer HT or SMT disabled to improve on FPS numbers while GPU intensive games seem to not care either way. I know there are going to be some games that make good use of SMT and HT however is it worth shutting off for most instances if you're not doing hard multitasking / multi-threaded work loads and you just want more bang for your buck on every day games? I think this would make for a cool video idea since not too many places touch on this as a simple solution to improving the hardware you have safely, if it does make the difference that most people say it does. Also, would be good to throw in Precision Timer as well since that's said to hinder performance. Would like to see if it's worth changing for most users for the small improvements they may cause. Very curious to see how it effects multiple different chips in benchmarks / gaming.

SMT on or off, what's better gaming-wise for Ryzen?

i was looking through my bios and came across this setting, i cant set it to enable. is that becouse you have to set a other setting of? motherboard: msi b350 pc mate cpu: ryzen 1600

(Well this is kind of a review of how SSD's are made.) ("final look at some of Kingston's test & assembly setup. mostly because the sheer amount of hardware is gleefully entertaining") Steve Burke the writer of this article recently visited Kingston Technologies' headquarters in Fountain Valley, CA, where he was able to get a tour on-site production facilities and talk about RAM & SSD assembly. This article is all about the production of RAM and SSD's at Kingston. Here Kingston Senior Technology Manager Mark Tekunoff gives us a walkthrough of the SMT lines and packaging machinery. How long does it take to make a stick of RAM? You can read and see more about this article here: gamersnexus.net -by Steve "Lelldorianx" Burke.

In gaming scenarios CPUs with smt have lower utilisation than those without smt (for example i5 and r5), shouldn't smt increase CPU utilisation because it allows the software to use a core while other thread is doing nothing (and CPU without smt should be under utilised)?

Hey all, I have a ryzen 5 1600. And from what I know - on some forum posts and videos e.g. from 3kilksphillip turning off hyper threading on skylake intel chips increased performance in a small value for CS:GO (5%) Does anyone have data to show the differences in gaming with simultaneous multithreading turned off. As SMT is the processor splitting the core to utilize the resources but gaming data is large? If somebody does not have data I am willing to make an excel spreadsheet and do benchmarks. I, am a person who doesn't want to overclock as I know it reduces the lifespan of the chip - I am willing to buy myself a new processor in the last/2nd to last year of AM4 socket chips - but there is no harm in giving the chip to a household pc to my family.. So I thought i could do something less harmful/not harmful to the processor. source -