>UPGRADE & BUILD ADVICEPost build list or current specs including MONITOR: https://pcpartpicker.com/Provide specific use casesState BUDGET and COUNTRY or you will NOT be helpedGuide: https://igwiki.lyci.de/wiki/Build_a_PC>CASEmATX: AP201, Lian Li A3, O11 Air Mini, XT M3, CH260ATX: XT PRO (ULTRA), AIR 903 Base/MAX, Lancool 207, Flux Pro, Y40, Meshify 3, 4000D FRAMEDual Chamber: Y60/70, O11 Vision (Compact), Antec C8>CPUBudget (inc. gaming): 7500F, 7600/X, 9600/XGaming: 9800X3D, 7800X3D, 9700XWorkstation: 9950X, 9900XPremium: 9950X3DIntel: If deeply discounted (265K for value & reliability, 14400F budget, 14600K midrange)>COOLERAIO: Thermalright Frozen Edge, Cooler Master Atmos, Liquid Freezer 3/Pro (unhandy mounting solution)Dual tower: A620 PRO SE, Peerless Assassin 120 SE V3, NH-D15 G2ITX/>42mm RAM: Mugen 6 Dual Fan, AXP120-X67TIM: MX-6, Duronaut, PTM7950 (phase change)>MOTHERBOARDAvoid ASRock AM5 boards, CPU degradation issues still reported.>RAMDDR5: 2x16GB or 2x24GB, 6000CL30 (AMD), 6400CL32 (Intel)Workstation: 2x 32GB (budget), 2x48GB or 2x64GB (high-end)>SSD (Keep firmware updated)Budget: SN7100, NM790High-end: SN850X, 990 Pro (Windows)Premium: SN8100https://borecraft.com/>GPUBudget: Arc B580 (for new builds), used cardMidrange: 9060 XT 16GB, 5060 TI 16GBHigh-end 1440p & Midrange 4K: 9070 XT, 5070 Ti Workstation: 5090, used 4090 or 3090 Ti>PSUBuying guide:https://hwbusters.com/best_picks/best-atxv3-pcie5-ready-psus-picks-hardware-busters/>MONITOR1080p: KTC H24T09P, ASUS XG259CS1440p: KTC H27T22, Gigabyte M27Q, ASUS XG27ACS, KTC M27T6 (miniLED)4K: Gigabyte M27UA/M27UP, KTC M27P6 (miniLED), ASUS PG32UCDM (QD-OLED), XG32UCWMG (WOLED)>OSActivate Windows @ >>>/g/fwt>CASE FANSMeta: Case with good stock PWM fansMidrange: Arctic P12 Pro, P14 Pro (5-pack, loud @ higher RPM)High-end: Fractal Momentum, Noctua G2 (140mm or 120mm)Prev: >>106494670
>News & RumorsRTX 50 Super: might launch in Q4 '25, 24GB VRAM for 5070 Ti & 5080 Super, 18GB for 5070 Super.>OmissionsRTX 5070: buy if it's reasonably priced & if you plan to work around potential VRAM bottlenecks.RX 9070 non-XT: worth considering if it's significantly cheaper than the 9070 XT.Intel 13 & 14th gen i7/9 -K CPUs: BIOS update required due to reliability issues. Recommended to set a CPU VR voltage limit (1.4 or 1.45V).>Product NoticePNY RTX 50 cards have QC issues (rattling fans, reports of 5070 & 5090 with blown capacitors)Gigabyte RTX 50 & RX 9000 video cards: thermal gel has QC issues, Gigabyte claims only early production batches were affected.https://www.igorslab.de/en/update-on-thermal-putty-on-gigabyte-graphics-cards-after-further-storage-oil-separation-to-an-alarming-extent-and-silence-as-a-strategy/>PSU Reviewshttps://www.tomshardware.com/pc-components/power-supplieshttps://www.lttlabs.com/https://www.kitguru.net/reviews/?category_name=power-supplieshttps://www.techpowerup.com/review/?category=Power+SuppliesFor tier-list enjoyers:https://docs.google.com/spreadsheets/d/1akCHL7Vhzk_EhrpIGkz8zTEvYfLDcaSpZRB6Xt6JWkc/>Storage Reviewshttps://www.phoronix.com/reviews/Storage (Linux performance)https://www.tweaktown.com/cat/storage/index.htmlhttps://www.techpowerup.com/review/?category=SSDhttps://www.tomshardware.com/pc-components/storage/ssds/reviewshttps://www.servethehome.com/category/storage/>Monitor Hunterhttps://docs.google.com/document/d/1illeNLsUfZ4KuJ9cIWKwTDUEXUVpplhUYHAiom-FaDo/
Asking again, is a delta of 30 something to be worried about if the hotspot caps at 85 Celsius?
reposting in the proper thread because i was fucking retarded getting baited like that.guess since i posted it recently enough i point at >>106507795 (You) (Cross-thread) from the last thread, though I forgot to take into account case fans.
>>106508151Run a gaming benchmark or other 3D benchmark that isn't FurMark.If the temperature delta is around 20°C in other benchmarks, it's fine.
>>106508187Do you know of any free ones? I ran the 3dmark Steel Nomad demo benchmark but that just shows GPU temp and not hotspot. GPU topped out at 57 Celsius.
>>106508168That CPU cooler is retarded, you're paying an insane premium that has fans that were designed over 10 years ago.https://www.techpowerup.com/review/id-cooling-frozn-a620-pro-se-cpu-air-cooler/9.htmlIs that CPU/memory/motherboard from a Microcenter combo? Fine if so.I wouldn't bother pairing a 7800X3D with a 2080 Ti, get a 7600X as a temp CPU and buy the Zen 6 X3D CPU in 1~2 years when you decide to get a better video card (keep the 7600X box and resell it).This is the kind of PC I would build for that 2080 Ti, you must update the ASRock board BIOS.https://pcpartpicker.com/list/nZ7VpKIf you want to avoid ASRock boards, the B650M Gigabyte Gaming PLUS WIFI you picked out is fine.
>>106508262Download Unigine Superposition and either run 1080p extreme or 8K Optimized mode.https://benchmark.unigine.com/superpositionDownload and leave HWInfo64 running in the background while you're running Superposition to see the hotspot temp.https://www.hwinfo.com/3DMark's in-house performance monitoring app can conflict with HWInfo if they're both running. You can try it if you want.
>>106508286Thank you will do.
>>106508273>Is that CPU/memory/motherboard from a Microcenter combo? Fine if so.all handpicking off of ratings and initial hearsay to temper afterwards with proper opinions afterwards, Microcenter is kind of a pain in the ass to navigate. 7800X3D was initially considered for an RX 9070 Sapphire Pulse combo but that might've been too steep a consideration on both fronts and sticking to the RTX 2080 Ti may be in my better front for the time being.
I just tried to swap out my old RAM kit https://www.newegg.com/a-tech-256gb/p/1X5-006W-003E6with this https://www.newegg.com/a-tech-512gb/p/1X5-006W-000J5and now it stalls at "0xAD - DXE_READY_TO_BOOT" and does not ever go further. The motherboard is an asrock romed8-2t. The old RAM worked just fine. Anyone have any ideas? The board supposedly supports LRDIMMs in addition to RDIMMs, so that shouldn't be the problem.
>>106508286So this was the result running the 1080p extreme benchmark. Slightly smaller gap of 26ish degrees Celsius. Anything to worry about here if I am not having any stability or performance issues? I am pretty new to this whole thing. I read on reddit that more than 20 degree delta is bad but no one ever really explained why so I am not sure if they are repeating what they heard or if the delta itself actually matters for whatever reason as long as the max hot spot temp isn't insane ( correct me if I am wrong but 85.9 on a hotspot during a benchmark isn't awful? )
>>106508449It seems fine, I would learn how to undervolt it to keep it running a bit cooler.https://youtu.be/naBqsSfGX3A
>>106508499This is running with a -50mV undervolt and nothing else. I should have specified. I tried -70 to -80 and it crashed during actual gameplay but it seems very stable at -50mV. Thank you for your help friend, you have put my mind at ease.
>>106508499Is there a video like this for nvidia gpu's?
>>106508273they seem to have issues with cooler manufacturing maybe lack of motivation after parked car
>>106508514NTA but you can simply experiment with undervolting. At worst you will experience instability but even if it crashes you won't do any permanent damage to your hardware.
>>10650844920C+ delta is bad on older GPUs but not this one. It seems AMD has improved the sensor on Rdna4 which makes it better at detecting hotspots than previous generations
>>106508499dumbfuck 33 or 45 not 10k
>>106508566Ah I see, so just outdated info for the hardware I am running. Thank you for letting me know!
>>106508514Most of the GeForce undervolting guides on YouTube make the card run power inefficiently @ lower voltages.Download this latest beta version of MSI Afterburner.https://forums.guru3d.com/threads/msi-afterburner-4-6-6-beta-6-with-unofficial-rdna4-support.456792/For the simplest OC+UV, raise the [Core (MHz)] offset slider in the main screen of Afterburner between +120~150Mhz.You can start at around +180~200MHz offset for RTX 50 cards (they have more overclocking headroom).Then lower the [Power Limit (%)] slider to 70~80%.You're not voltage-capping the GPU with this method, but the lower power limit will restrict the card from requesting high voltages.If you want to actually voltage cap the GPU, follow this guide.https://www.youtube.com/watch?v=LPpW9yXHvOUThe video raises the GPU core offset like I suggested earlier, and it modifies to curve to cap the maximum voltage draw.Read this guide for details & how to apply & save changes (including applying preset on Windows bootup).Don't set the OC+UV profile to apply on Windows startup until you've run plenty of stability test (gpu demanding benchmarks & games, including RT tests).https://github.com/LunarPSD/NvidiaOverclocking/blob/main/Nvidia%20Overclocking.mdYou can also OC the VRAM [Mem (MHz)] slider to around +800 MHz for RTX 30 & 40 series cards, and easily past +2000MHz for RTX 50 cards.
>>106508420Update your bios on the old ram sticks and try booting with the new sticks
>>106508132Doesn't that mean Intel b580s should only have slightly less compute to the 5060 ti? even more so if Nvidia doesn't omit dark silicon>19.6M compared to 22.9M An increase of 3.3M transistors shouldn't be as decisive as it seems to be desuTechpowerup for example lists the 5060 Ti as 143% more performant
>>106508597My current BIOS is v3.8 and newest is v4.1. It isn't really that out of date. The motherboard inherently supports LRDIMM, right? A BIOS update wouldn't be able to just add support for that right?https://www.asrockrack.com/general/productdetail.asp?Model=ROMED8-2T#Download
>>106508607You're assuming Intel's GPU design team is a good as NVIDIA's, it's simply not true.Intel is trying to make dGPUs out of Xe core iGPU IPs, and they also have inferior drivers.https://chipsandcheese.com/p/intels-battlemage-architecture
>>106508595Thanks for the info, all this computer stuff makes me feel like a big retard. Glad I decided to wait a few years and try to learn some things before attempting to build on my own.
Should I get the Asus PG27AQDM for vidya? I can't really justify spending more than that for some damn vidya.
>>106508707>2025>OLED
>>106508718
>>106508699The offsets I specified are just rough guidelines; you can push higher if your GPU can run it stably (silicon lottery).If you end up setting an unstable OC to apply on Windows bootup, enter safe mode and uninstall Afterburner.
>>106508733That's a photo taken by a customer who reviewed the monitor you are asking about on Newegg
AI is going to kill us all after they have stolen all our money. https://www.youtube.com/watch?v=hJ2tqs_vksc
Recs for a good cheap box fan I can put in front of my computer to cool it off better? Also would that even work to cool down my temps? I don't use ac.
>>106508846I tried it once and it made almost no difference for me.
>Me when I see the jpr numbers and AMD stockguy at my local computer store started laughing hysterically at me when I asked for a 9070xt. I asked him what was so funny he just turned his screen towards me and I saw it... I was so upset I ran out of the place crying and tripped on one of the parking things and then 3 black teenagers outside started laughing at me too...Team red bros I'm so tired of us always being losers, what can we do to change this situation?
>>1065088799070 XT is too expensive for casual gamers to meaningfully increase Radeon's market share.AMD needed to make a very strong 9060/XT priced at around $300 if they really cared about capturing the mainstream market.AMD & Intel are investing in dGPUs to make better APUs (besides the AI market).
>>106508879>Team redHow is it that this place has more corporate cheerleaders than Reddit?
>>106508961He's a shitposter for one, but the cringy corporate tribal shit was part of Radeon's official marketing.
>>106508665As far as i understand they got the compute right but can't use it wellAre the actual important bits are the>20 Xe cores(iGPUs next to each other)Vs>18 TPCs(dGPUs next to each other)I'm going on a whim here but does this mean each TPC is about 1.58 times more powerfull than an Xe core as per Techpowerup so i should evaualte accordingly or are there bandwidth/ GDDR shenanigans involved? As a bonus are "Tensor cores" just an entirely different grouping of transistors doing matrix calculations that AMD fears? Since that would explain why the RX 9060 XT, despite having significantly more transistors cannot do AI as well as even a B580.Also thanks, post more
>>106508956No they need a proper flagship. Anyone can make a cheap ass card.
>>106508607architecture matters a lot tooit doesnt matter if intel has 30B transistors if most of them are sitting idle because of having low occupancy in each wave causing most CUs to operate below capacity, or constantly waiting for data because of high cache latency and/or low cache hitrate
>>106508961>more corporate cheerleaders than Reddit?What are tyou talking about? I was trying to buy a 9070xt to replace my obsolete 7900xtx and now you're making fun of me too??
>>106509011Also a good example is to look at vega64. I believe it has more theoretical compute than the gtx1080ti but it had problems with wave occupancy in games which made it perform barely at a level of a much slower gtx1080
>>106508999Game devs care about install base, not maximum performance.The halo effect is overrated if those products are out of reach for the vast majority of PC gamers.The $200~400 video card market matters.
>>106509023Vega was god tier, 1080ti sissies could never ever play Indiana Jones and The Great circle + Doom The Dark Ages.
>>106508718Do you think the upcoming "true gloss" 540hz monitors are going to fix this? That fucking shill on youtube literally never mentions this issue in his reviews.
>>106509033i regretted getting the 1080ti over the 2080 for a $50 savings. the 1080ti was only ever good for dx11 games and never did well in vulkan or dx12
>>106509041Doubt the panel finish will change it.Tandem OLEDs have diagonal dithering issues.https://www.avforums.com/threads/lg-g5-oled-primary-rgb-tandem-2025-avf-review.2525553/page-69
>>106509025>Hey, we can get to 40% market share with this strategy.>Only gets 6%lol lmao
>>106508956But AMD fears the tensor core>>106509033I wonder what it would be like if we could mod silicon like we can solder VRAM
>>106508879>Provoking a brand war for the 50th timeYeah but the radeon posters the real problem. Kys
>>106509057I think they actually lost shares somehow.
>>106509057nvidia have been caught and sued for not disclosing crypto sales from gaming gpu sales beforeyou're an idiot if you think that 94% market share is purely because of gamers
>>106509067And you think AMD omitted Radeon revenue from cryptocurrency miners?AMD cards were also selling well during the mining boom.
>>106508879>>106509017>>106509057>>106509058>>106509064>Trolling is not permitted. Do not instigate or participate in flamewars
>>106509073nobody is buying amd radeon for ai today so you only get inflated numbers on one "team"
>>106509011Can there be a hypothetical modular GPU card? Like it has absurd bandwidth and is on PCIe 5.0 x16 and you just slot in extension cards containing compute units or VRAM on the extension card. There would be one card per PCIe gen and people would swap GPU chips or VRAM themselves>Excellent upgradability and repairability desu
>>106509077Shut up nerd.
so done with these fucking driver issues on my 5070 ti. i was in the middle of typing up a shitpost dunking on the AMDrones ITT when my screen went black and I had to turn my whole pc off. im rmaing and switching to 9070 xt fuck it idc anymore.
>>106509033The only game that mattered at the time was pubg and vega was terrible for it compared to the nvidia cards. It probably gets smoked in fortnite too.
>>106509077>Trolling is not permitted. Do not instigate or participate in flamewars
>>106509057>>Only gets 6%Because they are prioritizing margin over market share in reality.The 9060 XT 16GB sells for $380 without AMD's rebates atm, it's literally NVIDIA -$50.Hopefully they're just lying about increasing dGPU marketshare because they're so shit at this.
>>106509082like a socketable gpu board with built-in IO like a mini-motherboard? its theoretically possible but there's no incentive to build something like that
I'm at a loss. 5090 Astral for damn near four grand and the drivers just won't fucking work. Every game I open works for a few minutes then just a blackscreen. Help me pls.
>>106509081>ai todayThere's no inflation going on. AMD desperately wants to attract more AI customers.They're just shit at it.
>>106509113are you trolling or serious? It only crashes when you play games? What psu do you have and is it fully plugged into the card? what drivers are you using, what OS?
>>106509123its a ritual post
>>106509123He's just a very insecure Radeon card owner who always retaliates against the shitposter's bait.https://desuarchive.org/g/search/image/Req1Zuzilo2iQQJbhVBOMw/https://desuarchive.org/g/search/image/bNIzchFETF7pSm5X2znqXw/https://desuarchive.org/g/search/image/VFzlqtR4Gh8n9j9edsfvrA/https://desuarchive.org/g/search/image/c8HB6p-rD-Q_QZ3-A8E08w/
>>106509133I thought as much lol.
>>106509123>It only crashes when you play games?It also crashes after a few minutes from furmark or any other stress test>What psu do you have and is it fully plugged into the card? 1000 watts seasonic platinum and yes it is fully plugged in>what drivers are you using, what OS?I've tried latest multiple times on fresh windows installs and an older driver once. Nothing works.
>>106509118maybe they should consider making rocm work on an out of box install on windows and linux for their existing radeon lineup before throwing buzz words around
>>106509141get a 1200W instead
>>106509098Basically an GPU extension card with it's own extension cards, you pick how many Compute units and for what are on the card>Want AI?Fill it with tensor core card>Want Ray tracingRay tracing core card>Want VRAMVRAM card>Part brokeRepair easier because the cards can be swapped>GPU card brokeTake out the core cards and VRAM if unrepairable and use on another card, if a slot is broken, possibly carry on without it
>>106509135>The guy who is also making the same shitpost every thread breaking the no brand wars rule is at no fault for inciting other trolls to do the same>It is only the one shitposting in favor of the red tribe which i hate that is badKys fanboy they both should be permabanned
>>106509144>own extension cardsthat wouldn't be feasible because those units are tightly integrated and will lose significantly more performance over the high latency pcie bus than it would gain from larger compute units
>>106509118Add more tensor cores and ROCm is shit for Consumer grade
>>106509148Where did I say the NVIDIA shitposter's fine? The mods/jannies treat this place as a containment thread for retards, including the likes of (you).
>>106509141>>It only crashes when you play games?>It also crashes after a few minutes from furmark or any other stress testcap your fps at 120 fps globally in the nvidia app and trying playing a less demanding game like counter strike. If it doesn't crash it is definitely your psu being underpowered or maybe something wrong with your power cable.
>Bought 5070 ti cuz Im a compfag cs2 nigger and fell for the reflex meme>Get constant stutters and crashes no matter which driver version and even with reinstalled windows>Days of troubleshooting and no fix>Switch to 9070 XT>Locked 240hz and the frametime graph is literally a straight lineWell that was a week of hell but at least it's over now lmao
So insecure over a single shitpost.
>>106509159calling out one side makes you a supporter of the other side in fanboys' minds
How much vram do you think nvidia will have on the 6090?
>>10650921336gb from 384bit bus with 3gb models maybe
>>106509210I called out the NVIDIA shitposter in >>106508969It's a failure of moderation.I also wouldn't bring up CS2, the game fixes keep improving Intel CPU & NVIDIA GPU performance.
>>106509202>Anyone ever makes a post in this thread having an issue with a geforce card>Nvidiot trolls on their mind game bullshit repost it later so they can then post the screencap of the og and then dismiss the story as a shitpost and not a real negative experience someone had with an nvidia productI've seen this happen over a dozen times now and Im getting really sick of it. I remember the story back in the 2000s about nvidia hiring a marketing farm to shill in forums and Im starting to think they are still doing it to this day.
>>106509081>>106508956>>106509118>nobody is buying amd radeon for ai today so you only get inflated numbers on one "team"Okay but real talk when the gAIy bubble finally bursts will I be able to get a 9070xt for half of the original MSRP? Honestly I'd settle for that.
>>106509153>Here's another thing to look upI love youAny ideas on how to reduce that?From what I understand you mean the SLI question would be problem in this case as well. Best i could come up with is either to have them all connected to the same U shaped memory card and multiplexer 2
>>106509228Not my problem you're shit at creative writing, and you can't help yourself from standing out like the abortion you are.https://desuarchive.org/g/search/text/Bought%205070%20ti%20cuz%20Im%20a%20compfag%20cs2%20nigger%20and%20fell%20for%20the%20reflex%20meme/
>>106509237>Everyone who disagrees with me is the radeon trollYou are becoming as schizo as nico troon calling everyone dlss schizo
>>106509224I hope the 6080 could have specs like that and they keep the 6090 as a 512 bus card. i've been trying out some upscaling/interpolation and i've seen full 16GB vram + 32GB system being used with topaz so 48GB would be great for the 6090.
>>106509247There's nothing to agree or disagree with what I've said. You're shit at this. That's why you stand out.
>be me>have a 3060 Ti>coworker says "hey why don't you try my leftover 6950 XT?">fuck it, why not>boot it up, run time spy just to check>crashes on the first fucking run>adjust drivers, clocks, etc>try Doom benchmarks to see if that saves it>creates artifacts from the last menu I was on (akin to OLED burn-in) when I close out of the game>immediately yoink it and drop in my 3060 Ti (no issues)Fuck this, I'm just going to wait for the 5070S.
>>106509187It sounds like you had a faulty card or were doing some kind of weird software setup. In any case, locking at 240hz on cs2 is a bad idea.I'm rooting for you to make it out of silver.
>>106509234The Steam Hardware Survey captures gamers.Don't give us the laptop/legacy card inflation excuses either.The new RTX 50 cards' hardware % are also affected by those devices.There are nearly 8x more 5070 cards than the RX 9070 in the Steam Survey; the 9070 XT install base is too small to appear.
>>106509235latency is a very hard problem to solvecpu makers have resorted to tricks like integrating the memory controller onto the same die as the cpu where it used to be a separate chipset called the north bridge, a lot of logic in cpus like sophisticated prefetchers and multi-level caches were invented to reduce the latency impact of dram which hasn't improved since the 1990s or 2000s the separate expansion cards that you're thinking of used to be the case when 3D accelerators were just introduced and worked alongside 2D accelerators across the pci bus and it wasn't a large deal back then since there were lower hanging fruit for performance improvements like year-over-year 2x clockspeed increasesalso the customization of compute units that you're talking about is kind of being done right now by manufacturers who are resorting to tiles or chiplets with specific uses and put together to create products like the various arrowlake socs and strix halo. but these are done with highly sensitive and precise processes (think micrometer scale vias/interconnects with zero contaminants required) which isn't field replaceable by the user
>>106509254maybe they'll go full chiplets next gen and continue with a 512bit bus who knowsgb202 is already technically two gb203s stitched together logically but still produced as a monolithic chip so it seems they're testing out multi chip modules already
>>106509187>even with reinstalled windowsMaybe your Nvidia card would work fine if you'd stop using such a shit OS. Funny how the cards have turned an getting Nvidia to work in Linux is actually easier if not just works right out of the box just fine even with the fact that Nvidia hates Troonix users.
How come progams run 30% slower when running 16 at a time on my 16 core 9950x3D?
>>106509398Thank the cross-ccd latency Oscar.Keep your BIOS updated.https://www.tomshardware.com/pc-components/cpus/amd-microcode-improves-cross-ccd-latency-on-ryzen-9000-cpus-ryzen-9-9900x-and-ryzen-9-9950x-cross-ccd-latency-cut-in-half-to-match-previous-gen-models
>>106509265Did you run DDU (twice if not in safe mode) before you installed different drivers?6950 XT is quite strong on paper, slightly behind the 5070 in raster.But it draws a lot more power, probably more than a 5080 in average gaming loads.
Is a 7800x3d worth the $100 AUD increase over a 9700x?
>>106509443Yup. Did it twice and I undervolted it to prevent power overdraw. I'm running a 750W unit though, I wasn't expecting it to be THAT hungry.
>>106509457If the games you play scale very well with the 3D V-cacheI’d stick with the 9700X for 4K gaming.
>>106509457Yeah 9700x is a massive disappointment
Is it just in my particular shithole or are cases getting expensive? All i see is a bunch of chink labels with the actual good ones shooting up in price.
>>106509414ty anon! I will try and update my bios