>be rx6600>release 2021 amidst covid and height of cryptoscam>msrp 300>life is not good, but not bad either>FF 2026>ol rx6600 getting pretty outdated>still runs 1080p like a champ, but not any higher>$60 on eGay>DRMcuckbox comes out>$1400>gets curbstomped by rx6600>next week>rx6600 costs $200 againcant make this shit up
It's always hilarious to see how reactionary and volatile the diy gayming market is
Gamers have always been fucking retarded, as evidenced by the fact that they continue to buy Nvidia despite being responsible for the AI apocalypse.
>>109278241>you're retarded for buying superior productslet me guess: you run linux and think gimp is not only as good but even superior to commercial alternatives
>>109278268Larper
>>109278281stop using words you dont understand you retarded groupie
>>109278288Do you use Solidworks or Adobe slop cloud?
>>109278297it's like the average nvidia shill. AMD? but CUDA nvenc DLSS OptiX RTX remix. I don't use any of that but still you won't have access to them so you better buy buy buy.
>>109278241>>109278307AMD GPUs are only useful for gaming while costing almost as much as Nvidia. You might only care about gaming right now but if you want to do anything else in the future then you will hate yourself for being stupid enough to buy AMD. Saving maybe 10-20% compared to buying Nvidia is just not worth it when you will be using this product for the next 5+ years.AMD offers absolutely nothing except for badly copying Nvidia and always being at least 2 generations behind but they still have the audacity to charge big money for it. At least Intel knows that they are far behind and price their GPUs accordingly.
>>109278374Wouldn't your employer provide you with the tools you need for your job? Why would you bring your own GPU to work?
>>109278374you're exactly as described in>>109278307word for word.
>>109278374That's fair, and I half agree with your point, but I use Linux, so I'm beholden to AMD, wherein NVidia would constantly give me heart attacks during every update.
>>109278268you can do everything with gimp if you cant you are retardedyou can also do anything with emacs
>>109278418It doesn't have to be for work. Maybe you want to learn 3D rendering and animation in Blender as a hobby next year or you want to play with local AI models. Or there is this new game you really want to play but you have to use DLSS to run it which is just better than FSR. The reality is that AMD just offers nothing except for a small discount compared to Nvidia. Having no features is not a selling point.>>109278421That is just how consumers think since they make rational choices when they spend a huge amount of money and are locked into something for the foreseeable future. The consumer doesn't care about brand loyalty, they only care about how much they get for their money. AMD is heading head first for a 5% marketshare (they are almost there) until they reduce the price of their products by at least half.They copy every bad design from Nvidia but make it even worse. They cheaped out on the memory bus on their new lower end cards which they copied from Nvidia but they went even further and also cheaped out on the memory and gave it GDDR6 instead of GDDR7. They supposedly did this because they love gamers so much and they want to offer them a """cheap""" gaming card. The end result was a card that is 20% cheaper and also 20% slower while having none of the features that Nvidia offers. Why would anyone buy this?
I have one to sell, but everyones already moving on here
>>109278557Hobbyist ML and 3D rendering are fair use cases, but how many Nvidia users actually do this and how many just parrot Jensen Huang's marketing points?FSR 4.1 can nowadays be injected to most single player games that support DLSS2+, XeSS or FSR2+ using Optiscaler and looks sometimes worse and sometimes better than DLSS 4.0, which was already touted as looking better than native. I guess if you really insist on having path tracing you'd go for Nvidia as AMD Ray Regeneration is still Coming Soon.
>>109278374>AMD GPUs are only useful for gaming while costing almost as much as Nvidia.So fucking true now.After RDNA, they stopped offering entry level and cheaper gpus and pegged their cost to performance relative to nvidia.In my fictional country using a fictional currency called credits, RX 7600 has an asking price of C$1000. RTX 5050 is C$1100, RTX 5060 is C$1200 and RTX 5060Ti, C$1300. All 8GB models.Do you wanna know how much a 9060XT 8GB is? C$1400. With nothing but FSR and Lisa Su's bath water.AMD GPUs might as well be Nvidia GPUs with less features.
Lol novidia jeets are desperate to make a sale
>>109278493>wherein NVidia would constantly give me heart attacks during every updateSame. I would also have to pray that my pc would boot after a driver update. Even some small updates to the linux kernel headers could leave me with broken drivers. It would just boot into a TTY because it was so broken, other times there were weird phasing issues that left parts of the driver not updated or it missed the 32bit libraries. Absolute nightmare.
>>109278634I'm actually upset that AMD is shitting the bed againThey had a good run with 5850, 7770, R9 280X, RX 580 and (arguably) 5700 XT.Now there are no fucking good AMD cards, they're all nvidia but worse. It's no wonder that leathery slimeball fucker slipped: "we captured 100% of the gaming gpu market".
>>109278623It is hard to believe that Ryzen CPUs and Radeon GPUs are made by the same company. One offers competetive products at every price range that is sometimes even unbeaten while the other offers products that make no sense in any way price wise. There has to be some secret deal between Nvidia and AMD that stipulates that AMD will always be around so that Nvidia isn't counted as a monopoly while they produce GPUs that no one will buy. The only place where their GPUs are relevant is consoles and Nvidia probably gave them this market so that their GPU devision at least breaks even. There is no other explanation to this level of incompetence.
>>109278503i can also do my groceries just using my feet, but why the fuck would I do that?
>>109278634>undisputed market leader is so desperate that they need to shill on some random forumAt least make the cope believable. Notice how it is always Microsoft, Apple and Nvidia (aka the biggest companies in the world that are printing money at this point) that are supposedly "shilling" their products because they are that desperate for money. If anything AMD would need a shilling campaign to make people buy their subpar products.
lol. lmao, even
>>109278634>>109278813
>>109278718Well yeah the CEOs are cousins
>>109278751Notice how Nvidia advocates never reveal if they personally actually use any of the features they refer to besides DLSS
>>109278374i've used my amd card for llm's, gaussian splats, text2image, text2video, ray tracing, etc. lots of tools these days aren't nvidia-only
>>109277859You know what makes me hate AMD the most? The fact that they made RX 7400, which is a bit more powerful, uses RDNA3, uses only up to 43W power and would've been a smash hit for budget GPU market. But AMD thought "nah, let's keep this thing OEM only". Fuck them.
I just bought an rx6600 xt to replace my rx 570. Did I fuck up?
>>109279065Yep.Return it and get a RX 7600 or RTX 5050.
>>109279082I bought it from a guy in a car park
>>109279139Hunt him down
>>109278374False, I use my 9070 XT for LLMs
>>109279283Gemma 4 here. 9060 XT 16G. It's shit thought: 35 t/s.
I use a AMD Radeon RX 7600 XT and I have no issue with it because I've tapered off gaming for the past decade because the whole space is just so fucking lame. I don't care to generate images or video and 12GB VRAM is decent enough for basic chatbot functionality like slamming together cooking recipe and basic python or bash scripts. I went AMD this gen after being on nvidia forever because Nvidia didnt work with sway, hyprland or whatever else under the Wayland ecosystem and I didnt expect that to change because lol nvidia and linux.
>>109278374typical nvidia shill trying to get us AMDChads to buy their NVIDIA AI Cards that generate frames instead of processing frames. My 7800XT has dealt with GPU utilized workloads just fine. Blender, LLMs, etc.
>>109279065Depends how much you paid, but it's a decent GPU. I used to have one and it ran mostly everything I played up to 1440p and 4K/60 in some cases.
>>109277859I almost got one but ended up going with an xfx rx6400 because I realized I needed a single slot card to fit with the other two GPUs I was using at the time, I've since upgraded it to an rx9060 xt 16gb but the rx6400 did it's job of running a display on linux while my tesla p4 was running vidya in a windows vm.
>>109279283I tried but just gave it away to my brother and bought a used 3090. AMD is just shit, damn thing just crashes your whole pc if you get an oom.
>>109279391>>109280504Vulkan llama 60+ tokens per second, it's fine
>>109278241I want to play games tho
>>109280548That's much better than mine. I hope llamacpp nets me a few more tokens, but at 35 things are ogre.
>>109280593Actually it's about 110 t/s35 seems way too low for yours, are you using rocm or vulkan?
>>109278374What are you talking about fag? I literally made a small Ai lab at home running amd. Fuck Nvidia even though amd are also kike faggots but a bit less.
>>109278557>3D rendering and animation in Blender as a hobby next year or you want to play with local AI modelsI do this on my AMD card>DLSS... FSRI'm not a retard or a cuck so I don't use those. I run my games at native resolution like god intended. If I have to lower some settings or buy a faster GPU then so be it, but I've yet to find a game made in the past decade that's both worth playing and demands more than my GPU can offer.>AMD just offers nothing except for a small discountShit just werks on linux. All nvidia offers me is a housefire.
>>109278751Why do you think Coca Cola advertises during the Super Bowl? Do you think consumers don't know what Coca fucking Cola is? No, they do it to flex and say "WE are the product you're going to buy. We have so much fucking money that we can waste it on expensive ads, so we MUST be the best." Same deal with Nvidia. Notice all the posts sneering at the mere thought of owning an AMD GPU? Some of those are genuine shills, but many are consumers who have been conditioned to think #neverAMD by those shills.
>ol rx6600 getting pretty outdatedstill using my 1080ti (never updating)
>>109280706>are you using rocm or vulkan?I'm using through LMStudio, so I'm going to assume Vulkan, even though I think they are just a wrapper over llamacpp. Pic related. I checked settings and didn't find anything related to underlying API.
>>109280862Based beyond belief, but eventually get at least a 5060 Ti, anon. Say during this or the next Black Friday. It's going to be a hella of an upgrade, at least 50% better, and if you want to go cheaper and is just gaming, then AMD is even better value with much bigger VRAM (16GB on a 9060).
Bought a 6600XT recently for $150. Anything better had a worse price to performance ratio. The next cheapest cards I could find were 6700XT and 7600XT for $300 but those are only like 20% faster lmao.
>>109280849i've been conditioned to be #neverAmd by my previous AMD purchases. >but works on my machinedon't care, didn't work on mine. ngreedia works so that's what im using. I buy lower end anyway at deal prices
>>109280948If you are playing on 1080p that card still has some juice, but I would worry about the lack of upscaling support. A 7600XT would be nicer.
>>109280753Congrats for being part of a tiny minority that use AMD GPUs for local AI but that doesn't change the fact that Nvidia is just better supported and faster for AI workloads while AMD costs just as much as Nvidia. Even their CPUs are more popular because they actually offer something of value like up to 128GB of memory which you can't easily get on GPUs.
>>109280984Same but with Nvidia.>but works on my machinedon't care, didn't work on mine. AMD works so that's what im using. I buy lower end anyway at deal prices
>>109281019The most demanding games I play are unoptimized indieslop like Peak so I'll probably be good until 2030. Also been playing MGSV a lot lately and damn this game never so much as stutters even on max settings. Why can't all games be this optimized?
>>109281050and that's the only sane position. No corpo is your fren, buy what's at good price and works for you.
>>109281130But anon I've bought from intel, amd, and nvidia and have gotten duds from all of them.
>>109281076>Why can't all games be this optimized?Trust me. It's not that MGSV is optimized, but it's capped to hell and back to look highly detailed when everything is close to you. The game is just empty spaces with aggressive culling applied every-fucking-where. Same objects, same NPCs, pop-ins and LOD jumping to hell and back. I was playing a few months ago, still have it installed. It's pretty, but the more you play it, the more the shortcomings become obvious.
>>109280862>>109280925>my ancient piece of hardware is le based!!not trying to harass you or anything but this is just pure cope, sure it might be recognized as one of the greatest pieces of hardware ever but its ancient at this point, its like 10 years old right now. you are just copingbonus point if you are linux imbeciles
>>109280906You can see what you are running by going to Settings (Gear icon in the bottom left)->Developer and turning on Developer Mode. Then you will see a Runtime Menu in the settings.And if you are running a MoE model with GPU+CPU inference then you need to make sure to load the correct layers on the GPU because it makes a big difference in the performance.>I think they are just a wrapper over llamacppIt is. The big advantage over LM Studio is that llama.cpp automatically loads the model according to your available memory so you don't need to manually fine tune your model loading parameters so you always get the best performance right out of the box and this is why some people think that llama.cpp is way faster. But it is also updated faster and you can use way more launch parameters which sometimes can make a difference depending on the model. llama.cpp also has a webui and I think it works pretty well now but it isn't as plug and play as LM Studio.
>>109281216No need to shit on one's socioeconomic status. /g/ and /v/ are full of poorfags larping to their heart's content. In this case, it's just good business these days if you can avoid to give money to these fucking ghoulish companies filled to the brim with anticonsumer practices. And I say this as someone who updated their CPU and GPU recently.
>>109281253i will shit on them though>my ancient piece of hardware is le based!!!!! its just poorfag cope, im not trying to harass poor people though don't get me wrong but this is just typical poorfag cope from this website
>>109281278How is a reliable piece of hardware for a reasonable price not based? It sounds like you're unhappy with your life so you justify your existence by being a consoooomer.
>>109281222Trips witnessed. Thank you for the mini write up. I found it, yes, and... it's Vulkan. Pic related. I'm downloading ROCm now. Thought I had it installed already, but I guess this is it. Thanks again.
>>109281294its fine to use old hardware if you like it but im just saying is people running around claiming how its giga based don't actually like it, they are poorfags and are just coping. you know im right
>>109281325Why would they not like it if it works? Sure it's not as fast as a more expensive card, but by that logic I can say whatever GPU you have is shit and cringe because I doubt you have the fastest GPU in the world. Maybe you think $2000 is reasonable to spend on +10FPS in Cunterstrike but for me $200 to play every game I enjoy at my monitor's refresh rate is ultra comfy.
>>109281360there is no need to continue this discussion, you are trying to make this about me and are trying to dox me?if you are happy with whatever you have that is fine but running around coping >this thing from 10 years ago is le best!!!!is usually poorfag cope
>>109281397Ok you are objectively schizophrenic and illiterate. Nobody said a 1080ti was the best card in existence but it is widely regarded as an excellent GPU. And now your pedo diaper loli ass is claiming I'm doxing you by not personally following the same purchasing preferences as you. You're write, there really is no need to continue this discussion. You're cooked. Go offline and cool off for a while.
>>109281424am i not wrong though?>my ancient piece of hardware is le best!!!is poorfag cope, is it not? i found plenty of people start harassment campaigns on this website when someone confronts them that their 1080 or whatever is now ancient garbage
>>109281424fyi this guy is a troll and it looks like computer hardware is his current flavor of the week interest
>>109281301Glad I could help. Another thing that might help is enabling the new LM Studio Engine Protocol. It fixed one issue I had with partial offloading at least which in turn boosted my t/s. It is under Developer->Runtime settings->Use LM Studio Engine Protocol. They also say it makes engine updates faster and more frequent so that is also a plus. It might already be enabled by default on newer installs though.
>>109281615It wasn't activated for me. Weird. Must be the version of LMStudio I'm using it. I don't think I updated it since I installed a few weeks ago.Unfortunately the ROCM installation didn't quite work. It couldn't locate my GPU, I googled it and it's something to do with the lack of privileges, which I'm not that eager to give to these guys. I will save that to my own tinkering with llamacpp. Having said that, a nice surprise: I managed to get 68t/s using gemma-4-e4b. Small 5.89GB model. Probably can't do much, but still, nice. The bigger 26B A4B QAT which fits on my 16 VRAM GPU sadly isn't getting anything further than 40t/s on its best moments. I downloaded Qwen3.5 9B, and it's hitting 40t/s. I'll experiment a bit more, but so far nothing that a little bit of patience won't help. Rocm probably would do wonders, of course, but that's for another day. I just want to leave one model on my PC in case I need help troubleshooting something if I find myself without internet (sometimes the link fails when it's raining).
>>109281710I only used the Cuda backend in the past so I don't know how big the difference between ROCm and Vulkan is but Vulkan is compatible with every GPU so I just tried running Gemma 4 12B with my 5060Ti 16Gb with Vulkan and Cuda and got ~52t/s both times. Vulkan was only a bit slower.I also have Gemma 4 26B A4B QAT and with all layers on the GPU I get over 115t/s so your 40t/s seems really slow. But it only barely fit inside the Vram so maybe it is actually offloading some layers to the CPU on your system. Might be worth checking how the model gets loaded and if the GPU layers are maxed.Another thing that might be useful is looking into MTP and speculative decoding. Gemma 4 and Qwen support it and it can significantly speed up token generation especially for coding. I haven't figured out how to use it in LM Studio yet though and have only used it with llama.cpp. With Gemma 4 12B I went from 52t/s to 125t/s but it depends on what the output is and it performs best with coding. The extra MTP layer does take up a little bit of extra Vram though but with the 12B model it stil fit inside my GPU completely.
>>109282271>I also have Gemma 4 26B A4B QAT and with all layers on the GPU I get over 115t/s so your 40t/s seems really slow.Man, 115t/s would be fucking great. You are using Nvidia, though, so there's that, too.>Another thing that might be useful is looking into MTP and speculative decoding. This is EXACTLY what I'm doing right now. Testing that on the 12B before moving to the 26B A4B as my main. Unfortunately, again, it's a pain and I can't find a suitable assistant model to work (my guess). I just get an error >Engine protocol predict stream returned an error: {"code":500,"message":"decode() failed: failed to process speculative batch","type":"server_error"}Or if I try to use another gemma model (which is obviously not ideal since the draft model is suppose to be really small, right?), like the E4B, the output gets slower, down to 20t/s. I heard that speculative thingy can give me good +20t/s. That would be indeed great. Alas, I'll trying a bit later. Suffice to say that at least that decoding I have to get it right, so I'm positive it's going to work. Thanks.
>>109277859Real inflation over the last 5 years is much higher than being reported. The total fiat money supply is much higher than it was back then, which is how you should actually calculate inflation. The costs to produce hardware has risen beyond that because of AI demands. It's a miracle you can still get a PC that is reasonably affordable, especially second hand. The supply will dry up and more people will realize its value. It's going to go up at least another 50% within the next 2 years, if not more. I don't even think worthwhile consumer products will really be released in the coming years, and if they are they will be extremely expensive. Which will again drive up the price of current products and the second hand market since it will show how much more valuable it is than people think.
>>109282332MTP is newer and should be better than normal speculative decoding because the MTP layer is really small unlike normal speculative decoding where you have to use a second model that is also pretty big. It also gives way more of a speed up. For Qwen this MTP layer is already integrated into the model for example here: https://huggingface.co/unsloth/Qwen3.5-9B-MTP-GGUFYou just have to toggle the speculative decoding to "MTP" instead of "Draft model" inside LM Studio and it should just work but the speed increase might vary.Gemma is a bit weirder because Google added this MTP layer later on and released it as a seperate thing so you have to load it separately. Here it is for Gemma 12B https://huggingface.co/unsloth/gemma-4-12B-it-qat-GGUF/tree/main It's the 254MB file.I don't know if it is possible to load this MTP layer in LM Studio but in llama.cpp it works.
>>109278374Lol I would have just flipped the word Nvidia for AMD here. Nvidia is the one locking you into their garden, not AMD. If you actual have the longest outlook, AMD wins. If I was a poor third worlder I would go AMD easily. Yes you get less at the point of purchase, and their basic to approach bleeds into other stuff like memory controllers and audio etc. At the same time, people go with the open standard/source approach for futureproofing when products are fucking sheit (see blender, freecad, and a million other examples). If we quantify AMD as being 20 - 0 % worse, then it's really not so bad.Also the fucking smarmy apathy at something vaguely ideological (perhaps seen as naive), is fucking inane. You don't walk into a car repair shop and laugh at give people a hard time for buying cars based on repairability or gets less for their money, as anywhere else. It's bizarre and embarrassing. Nvidia has always been one or two notches more slimy, and it's ok to push for a more healthy and ethical industry.And betting on future software improvements is more than just a meme, especially now that things are slowing down. AMD is hiring more, and as things gets more settled, patents expirations catch up with the global corpo, a GPU becomes a more general use item. Except for those who cuck themselves with locked down hardware of course.
>>109282578Tried that, hopeful that it would work, but it did not. llama.cpp is going to save me. I'm sure.
>>109277859>8gb of rami guess that might be good for a voice model
>>109278503You cannot work on CMYK documents in GIMP for print projects.
>>109277859>DRMcuckboxA PS5?
>>109281461The one time when it's not a meme is when old hardware has better build quality than new, like certain input devices
>>109277859DILATE THIS
>>109279060we really never moving from these gt710 and 1030... it has to be manipulated market...
>>109278374>>109280984skill issuei've had only ati/amd, cpus as well only amdATI Radeon 9550AMD Radeon HD 5670AMD Radeon RX 570AMD Radeon RX 6900 XTAMD Radeon RX 9060 TI 16GB - downgraded from the last one because it was overkill for meI never had much issues on Red side. Sure in ATI days drivers sucked and computer crashed, but it was common for anyone's computer to crash and often it wasn't caused by gpu but something else.things amd can do:- better support for older games- just works linux support- better support for console emulators- blender works, you are just a snowflake- llm works and has more VRAM, and on cheaper tierswhat amd lacks:- physx- cudaI've always picked amd because it was always cheaper for what i've got at that price. I've went full retard once with 6900 XT but never again - it was BIG, power hungry, noisy and hot150-200w is the ceiling for me. I'm not going further anymore. It's retarded.
>>109285446forgot about the fact that amd still uses 8pin power instead of 12/16pin housefire which makes it superior. The last Nvidia gpu with non-housefire 8pin was 4070 12 gb (non super, non ti)
>>109281397you are retarded nigger. How is it coping if it works for the anon. You are just a blind retard that chases imaginary numbers. If someone plays at 1080p at 60 fps they will hardly need a better gpu. And the higher fps that your monitors are for feces smelling retards that have no idea what they are doing. 1080p is most common resolution still. The only need for higher resolution is artificial bullshit like godawful unreal 5 slop looking like thrash on lower resolutions than 1440p. Games in 2016 looked and played (physics) orders of magnitude better than unreal 5 slop. The modern cards are only needed to offset the garbage that unreal 5 produced - to hide the shit like that weird shimmering and blur/afterimages. So it's actually the opposite. Modern card is cope, because it hides the shit you don't want to see. You don't want to see how shit modern unreal 5 games are in their native resolution. You are watching fake upscaled and sometimes made up frames. As for the other games than that unreal 5 trash then 1080 ti is probably still capable to max out the settings at native resolution.
>>109283396you dont do this, you are a pedantic nitpicking little bitch that shills adobe for free. modern printers can convert rgb to cmyk if necessary. cmyk has a smaller color gamut so you would work in rgb first regardless. if you REALLY needed to directly edit in cmyk, krita can do that. but you dont create anything, youre just a little bitch on the internet trying to score points for a corporation that would kill you for a dollar.
>>1092855291080p is third world tier. Pc monitors are supposed to be higher resolution than televisions.
>>109285446>blender works, you are just a snowflakeNo one said it doesn't work with AMD. It is just way slower than the Nvidia equivalent because it doesn't support OptiX. Nvidia is somewhere between 40% to 80% faster because Blender is optimized to use the Nvidia hardware directly.>llm works and has more VRAM, and on cheaper tiersAgain, no one said it doesn't work. The support for AMD is just way worse than Nvidia and every inference engine is optimized for Cuda which means better performance. You are lucky if it is supported at all.And the 9060XT has only a slighty lower MSRP than the 5060Ti and both have 16GB of Vram. But the 9060XT has both a crippled memory bus and last gen memory which gives it a really slow 322.3 GB/s memory bandwidth compared to the 5060Ti with 448.0 GB/s and even that isn't that good nowadays. LLMs are bottlenecked by the memory bandwidth so AMD is at least 30% slower just because of this and it is only because they wanted to cheap out even harder than Nvidia while charging almost the same price.The reason no one buys AMD is because their cards are way too expensive while offering nothing.Nvidia is 20% more expensive but you get slightly faster rasterization, way faster raytracing, way faster 3D rendering, way better AI support and hardware. And on top of that you get every new technology while with AMD you have to wait a few years before they can badly copy Nvidia. The difference in the performance in Blender for example is like jumping multiple tiers of graphics cards but you only have to pay a little bit more so it's a no-brainer.tl;dr AMD cards cost just as much as Nvidia while being awful in everything outside of gaming where they are only a little bit worse. It's not rocket science.
>>109280222£150
Man people argue a lot about somthing thst only costs $400-1000. Right now you only have two choices with good value, do you get a 9070xt or a 5070ti? Those are the only cards worth buying right now anyway. If you want to just play games you get the 9070xt and if you want to also coomgen you get the 5070ti, it's that simple.
The unpaid Nvidia shill ITT talks like the schizo from UserBenchmark lmao
>>109286535Seems to be the going rate desu. I don't know how long you plan to keep it but you could probably resell in a few years at a minor loss.
>>109277859I'm still running this one and I don't think I will be able to afford to upgrade in at least 5-6 years.
>>109286507my rx7800xt has 624.1GB/s memory and cost me nearly half that of a new 5060ti
