ivanrcks95

This was just a fun curiosity experiment, that ended up with great results. For the longest time outlets like Gamers Nexus made comments about certain cards blocking a lot of airflow by having the sides of the cooler blocked off. My particular card has the same "issue", there is a small gap between the cooler and the LED bar that goes across, but it does definitely block a majority of it. I decided to see if it 1 - could be removed and 2 - if it will make a difference. Removing it was very simple, 6 small screws and a plug later it all came off, and the difference it made was not insignificant. I wish I recorded all data I had before, but I did this mostly for shits and giggles not thinking it would actually make a difference, so i dont have actual screenshots of the before results, but without changing anything else, fan profiles all being the same, I saw a 4c drop in temps. With both attempts I ran cyberpunk with RT on for 30min, and the edge and hotspot temp both dropped by 4-6c, going down from 71c and 90c to 67c and 84c respectively, those are the max temps. Average temps were more closer to 4c all round. Surprised how much of a difference its made and how easy it is to do it. I also noticed that XFX stopped putting this LED bar on the 7000 series, so maybe they noticed these results too.

Its definitely strange, mostly unexpected to be honest. Plays one of the most demanding games of the last decade - cool temps. Plays a game thats nearly 19 years old - reaching temp limits. Ill do more digging online to see if its perhaps a 6900XTXH specific situation, since it was their biggest and highest clocked die with most limits raised.

Yeah that makes sense, thats what I thought initially as well, its just confusing that the less demanding games are the ones with the issue not the more demanding games lol As I said, Cyberpunk with ray tracing has a cooler hotspot temp than a game from 2005 with no ray tracing.

Just an observation I made thats very odd to me. Ive noticed that depending on the game, the hotspot and edge temperature difference gets bigger? On average in games my GPU edge temp sits between 62-67c, depending on the demand of the game. Games like Cyberpunk maxed out (including ray traced reflections), NFS Unbound maxed out, will all sit at 67c edge temp, and the hotspot would hover around 85-88c. With less demanding games, and really old games with lots of mods its does the opposite? Games like Automation Car Company Tycoon, or a modded version of NFS Most Wanted 2005, the edge temp will be lower, 60-62c, indicating its less demanding but the hotspot would be higher at 90-95c? Is this just poor optimization? Seems odd that a less demanding game would have a lower edge temp but a higher hotspot temp? The temperatures are still safe for a 6900XT, so I am not worried about that but I am confused as to why the delta increases when the game is less demanding.

Thanks! I forgot to actually include what the 5800X3D swap cost, but the CPU was $300 new, and I sold the 5600X for right around $100, so another $200 on top of the $475, so the final total spent is $675usd for the full upgrade path. Obviously its not fair to say I spent $675 to build a 5800X3D, 6900XT specced PC, as I still had to buy the original parts I sold off, but when you wait 3+ years between upgrades, youve pretty much paid off those original parts anyway, and upgrading is always cheaper than starting from new.

Thanks! And yeah, its came with the card. As far as I know, only the 6900XT Limited Black had it, no other XFX 6000 card.

Cleaning day, so thought id grab the most up to date pic. Yes custom GPU cables had to go (they were causing power delivery issues)

Yeah ive heard about the Arctic hum, maybe I got lucky, but my sets of P12/P14s and F14s dont do it, or I cant hear the frequency they do it at lol Also see the comment just above about the pump, after a quick research session, seems like its just a X63 thing, no reason to suspect its faulty if its been 3 years, I think theyre just noisy pumps in general.

I decided to look into X63s online, and seems like its the crap pump part. Pretty much every video or post I come across is people complaining about them being noisy above 60%, same exact noise mine makes. So seems like its just a noisy pump in general. Its lasted 3 years, so no reason to switch it, and if its completely silent at 50% with the temperatures under control, seems like theres no reason to leave it running faster anyway, or set up pump curves for it to ramp up/ramp down.

NZXT Kraken X63. Ive had it for 3 years now, and the noise it makes above 50% has been the same since day 1, so I know for sure its not a recent issue. I run Arctic P/F fans all the way through, the case is a Lancool II Mesh Performance, so the fans, even under load, dont need to go above 50% fan speed, which is around 1040rpm based on my fan software, and at those speeds theyre dead silent. The pump makes like a humming noise above 50%, and yeah it is louder than the fans.

Anything above 50% i start to hear it, as soon as 55% its audible. My fans are dead silent even under some load, so it makes any pump noise more obvious. Thats why ive kept it locked at 50%. Is there any benefit of having it go down to 40% and ramp up to 50% instead of just having it fixed at 50%?

More of a curiosity question than anything else as ive heard mixed responses about this topic. Is it better to keep an AIO pump running at the same rate all the time, or have it ramp up with the fans? Ive personally ran my AIO pump at 50% locked all the time, with only my fans ramping up as temp goes up. This way the temps are still perfectly fine (76c max after 10min R23 on a 5800X3D, 65c or less in games), and I dont hear the pump at all, its dead silent. Ramping up the pump does help drop these temps by another 5-6c, not an insignificant amount, but then the pump gets very loud, louder than the fans. Is it worth it though? Does the pump experience any extra wear when constantly ramping up and down compared to being locked at 50% speed? Would be curious to get some extra insight.

Today was the day, finally made the plunge to commit to a 5800X3D. it was always in the plans for this PCs "ultimate" version, just didnt expect to be switching to it this year, but an epic black friday deal was too hard to refuse. I do a lot of PS3 emulation on top of my normal gaming, so this change has been extremely rewarding in both aspects, pretty much a 25-30% uplift in both scenarios over my 5600X, do not regret it one bit. GT6 emulation has pretty much gained 10-15fps on average, which keeps it in the high 30s now at its lowest point, whilst staying locked at 60 most of the time. 5600X couldnt stop the game from dropping into the 20s, and on average it would stick in the 40s, never locked at 60. It responded very well to a PBO curve optimized undervolt, managed to do -25 all core (i never tried -30, as -25 got me what I wanted, so who knows might even be stable at -30, will try at some point), which has stabilized clocks at 4450mhz, dropped the temps by 7-8c, and improved scores and performance all round. Cinebench R23 went from 14,406 on stock settings to 15,150 with the -25 TimeSpy CPU score improved from 11,433 to 11,662, not as big of a jump, but the jump did still happen while having much lower temps. First untouched results the chip would hit 84c, now it maxes at 76/77c, doesnt even hit 70c when gaming. Overall extremely happy with my purchase, and im hoping this rig will serve me well for at least 2-3 years, before needing to swap to AM5 or whatever other platform will be available by them.

Well, its been over a week now since switching back to just using the PSU cables without the custom extensions and the problem has not happened again, so I think its safe to say that it was most likely the issue. If anything changes ill report back!

That adds up, doesn't help when the GPU draws 350w as well. Really hope its just the cables and not the PSU, its a RM850x they're usually highly recommended so fingers crossed

Really weird issue has been happening to my GPU over the past few days. The screen would randomly go blank, and the PC would be unresponsive, forcefully shutting down the PC and powering it on again would bring me back to windows, but the GPU driver would be corrupted/reset, and the only way I can install a new driver again is by using DDU, and then installing the driver, without DDU the installation would fail. Its happened 3 times, twice while I was just browsing, and once during a game. I never paid attention to the PC lights the 2 times it happened while browsing, but I did during the gaming one, and I noticed one of the power connectors on the GPU had a glowing red light on it (bad power/no power). I am running custom extension cables to the GPU, yes extensions not direct to PSU cables, could they be the culprit? Ive removed them for now and just running the stock PSU cables directly now to see if it goes away. Odd thing is, it only happened once under load, and ive been running these extension cables for almost 2 months with no issues, could the PSU be failing or some other power related issue?

Its the same graph top and bottom, just if you notice, ive placed the data point on the top graph of where the cards stays at 99% of the time, and the bottom graph ive placed the data point where the spike happens. So you can compare the stats side by side of 99% usage and the random spike.

Top graph is showing what the card normally stays at, and shows how flat the temp curve is. Bottom graph is highlighting exactly when the spike happens, and all the other variables staying the same when said spike occurs.

Not really paranoia rather more about curiosity. It happens in every single game I play, so I know its not something a particular game is doing. Just seems completely random for a spike to happen when every other thing stays unchanged.

Havent really seen something like this before so thought id ask here. My 6900 XT has weird, less than a second temperature spikes occasionally, whilst nothing else changes. Ive attached a graph below to show what I mean. Ive highlighted all the main points. The card stays pegged at 67c, you can see how flat the graph is, but every so often it peaks to 70c/71c out of nowhere, with no other indication anywhere else that anything else has changed. Hotspot temp is the same, clocks are the same, power draw is actually a bit less when it peaks to 70c. Anyone have any ideas as to why this is? Havent seen this before on my old cards so unsure.

After turning the display down to 60hz, and then back up to 144hz in windows display settings, it seems to be smooth again with no jitters or choppy animations. Seems like the settings didnt save after a restart or something? Will monitor it over the next few days to see if it happens again.

Ill try this for the zoom app thanks, but other apps are affected as well, so I suspect it is something to do with windows 10 itself

I recently upgraded to a new monitor. My set up before was 2 24inch 1080p 144hz displays, didnt have any issues with them. The new monitor is just a single setup, 1440p 144hz display. Everything was all good yesterday when I got the monitor, but today morning I am getting weird choppy animations with certain apps, not with all. For example, my zoom client app is very jittery when I move around the window, as it VLC. But regular windows explorer is smooth. The windows snipping tool animation is also ultra jittery. Ive tried turning on an off variable refresh rate, as it is a g-sync monitor, as well as changing the refresh rate, none of it has helped. Curious to know if anyone else had any ideas.

I had acquired this card just over a week ago. Despite being on AMD CPUs since 2018, with my first Ryzen, the Ryzen 5 2600, and sticking to Ryzen CPUs only since then, my entire GPU collection up to this point has always been nvidia. From my first card in 2007, the GeForce 8500 GT, up to my last card, a Gigabyte RTX 2070. However, since 2021 ive been more and more tempted by the red team. The 5000 series was a good first attempt to be back in contention on the mainstream GPU market, but I feel like they made the biggest dent with the 6000 series cards, and 7000 series is looking to continue that trend. Therefore when I was looking to finally upgrade my 2070, which I had used since 2020, AMD cards were on the top of my list. The 40-series and 7000 series cards were a bit out of my budget, even in the used market. I couldnt quite get myself to swing $700 for a card, those were the minimum prices of 7000 series and 40 series cards that had at least a 50% uplift in performance over my 2070. A 50% uplift in performance was always the limit I set myself to consider upgrading, otherwise I personally dont see it being worth it. This left me with the 30 series and the 6000 series cards. With my 50% uplift target, I was left with RTX 3080s and above, and RX 6800XTs and above. To my surprise, 3080s here are still at least $600 for a good used one, whereas 3080tis and 3090s were all $700 and above. I honestly didnt expect them to hold value so well, and that basically decided my fate. On top of the somewhat shady business practices nvidia had done over the past 3 years, it looks like the market made the decision for me. After a few more weeks of digging, my card popped up, a XFX Merc 6900 XT Limited Black, for $450usd used. The seller had 1000s of positive reviews, and from all the photos and videos the card looked in very good shape, so I decided to commit. As I was waiting for the card to ship, I did more research into it. Turns out, the Limited Black card is actually a pretty significant upgrade even over the stock 6900 XT. Firstly, its an XTXH chipped 6900, for those who arent aware, 6900s came in 2 variants, the XTX and the XTXH, with the XTXH chips being better "binned" and "unlocked" for overclocking. They had higher clock limits on the core and memory, and were usually seen in the water cooled cards, very few air cooled 6900s had them. Secondly, its 2495mhz advertised boost clock also puts it pretty high on the list of highest boosting 6900 XTs overall, air or watercooled. Safe to say I was very exited to play around with it, as overclocking is something ive always enjoyed. With my previous 2070, and a AIO strapped onto it, I even managed to break into the top 3 of 3dmark scores with it for my hardware. However, I did have my doubts, especially given the driver history and overall instability of AMD cards that was reported online. The card does live up to the expectations I had for it. On completely stock settings the card gets a 22,274 graphics score in TimeSpy, this is a 54% boost in performance over the 2070s MAXIMUM overclocked graphics score of 10,421. If we compare the stock 2070 graphics score, then its a 62% boost. The story continues in FireStrike, with the 6900 XT setting a graphics score of 62,929, completely stock, which is a 59% uplift over, again, the MAXIMUM overclocked graphics score the 2070 ever got of 26,108. Stock for stock, this difference goes up to 62% again. Great, the card meets my 50% uplift requirement, and it manages to do so whilst running the fans at a maximum of 35%, which is barely audible, and the core doesnt see anything above 67c, with the hotspot staying around 80-84c, something the 2070 struggled with, as the stock gigabyte cooler wasnt very good. *Take this with a bit of a grain of salt, as the seller did say he repasted the card with MX4, and this could be why it runs so cool. In any case, this means this card is going to be completely insane when overclocking right? Not quite. This is where I came upon my first hurdle with this card. I have to give credit to XFX, they have absolutely squeezed every single bit of performance out of this card on its stock settings, with the card hitting and holding 2600mhz constantly, even when gaming, well above the advertised clocks. But it does mean that there is hardly any room to go above the stock settings. With a maxed power limit, overclocked memory and core, the absolute best score I managed to get whilst being stable was a 22,782 in TimeSpy and a 64,911 in FireStrike. Only a 3% and 4% uplift respectfully, so much for that "binned" and "unlocked" XTXH chip. Turns out, after spending some time on this forum and other AMD subreddits and forums, the card is power limited. Despite drawing over 360w with the 15% power slider, it is not enough for it to do anything more, and most people have to resort to MPT to get more out of it, something I wasnt willing to do. Fine, so the overclocking card is so heavily overclocked from the factory that anything I do on top doesnt result in any more significant performance gained. This does mean I should have some room for undervolting, and maybe I can get the card to run even quieter and even cooler at the same settings right? Once again, the cards clocks are so high from the get go, that anything below 1200mv is not stable, and would require a drop to the factory clocks. Something I was not willing to do. So im stuck with an overclocking card that I cant overclock without messing with BIOS limits, and a overclocking card I cant undervolt without sacrificing performance, great! Not quite what I was expecting, but its not the end of the world. The card performs so well out of the box that frankly the lack of any overclocking performance gain isnt a deal breaker. As even on stock settings the card is setting numbers that are 6% higher than the average 6900 XT and 4% higher than a 6950 XT. Its actually performing closer to a 4070ti, and outperforming the daddy of 30 series, the 3090ti, if the 3dMarks own results are to be trusted. So I cant really be sad about its performance. How about those infamous AMD drivers then? And the constant game crashes and shitty optimization? Well, I havent experienced any of that at all, everything has been rock solid, every game I play, and all my work related tasks have been flawless, fingers crossed it stays this way. In a weird turn, the card being as good as it is, has actually revealed bottlenecks in my system that I did not have with the 2070, take that as you will but it has forced me to look into the 5800X3D a lot sooner than I wanted to. In conclusion I am still happy with what I got for the price I got it at. The card is extremely well built, the Limited Black actually comes with a cool and very solid brace/sag bracket that bolts right into the card and really makes it look like a serious weapon inside of my case. It runs super quiet and super cool (once again, this might not be the case for stock cards as it has been repasted), and its performance is above the expectations I had for it, despite the limited overclocking potential. But those limited overclocking potentials do need to be highlighted to anyone considering this card for that purpose. As they might have seen it has the XTXH chip and rushed out to get one thinking it will be an overclocking weapon, but the card ends up being the exact opposite. Just something to take note of.

I had to go down to +50mhz and a -20 offset, this is the only time that it passed every single stress test and normal testing. 12042 as the best score at R23 with maxing out at 74c. After a 10minute R23 test, the result dropped down to 11096, with a max temp of 75.4c. Passed both TimeSpy and FireStrike Passed 20minutes of Prime95 with a full "blend" load. This was by far the most demanding test, but even still, at the quiet fan and pump profile it only hit 80.8c max after 20minutes. No crashes. All of the scores were a 3-5% improvements overall compared to the stock PBO results. With these settings, the voltage fluctuates between 1.275 and 1.281v, however, what I did notice is that with these settings the core clocks are literally LOCKED to 4.7ghz, they dont budge at all, even after 10min of R23 it still hold strong at 4.7ghz. Whereas with the +100mhz and -20 settings, the clocks do spike to 4.75ghz, and stay between 4.7 and 4.75 when doing a single test, however they fluctuate quite a lot when doing a stress test, especially as temperature goes up, and most of the time they were actually below 4.7ghz, hovering around 4.65-4.68. This is reflective as to why the scores and temperatures between +50mhz and +100mhz runs were so similar, after the 10min R23 test the +100mhz run was at 11106, a 0.09% difference compared to the 10min +50mhz run. The +50mhz setting just manages to be more stable and lock the clock at 4.7ghz more often. Ill keep testing day to day and gaming use with these settings and see how it goes, but as of right now I am happy. Small boost in performance, decent drop in temperatures over stock PBO, exactly what I wanted.