*Official Benchmark Thread* - Post it here or it didn't happen :D

[Mr. Fox]

I've been watching the insane increase in DDR5 prices. I don't need any, but decided to see if I could find some spares cheap to hang onto. Just scored a Trident Z5 DDR5-8000 48GB CL40 1.400V kit with Supercool direct touch cooler on eBay for $250. That same RAM kit is now about $300 sold and shipped by NewEgg. So, basically like getting the Supercool cooler kit for free, give or take a couple of bucks.

 

[Papusan]

 

Made for throwaway. Maybe Nvidia and their AIC partners should send with spare 12VHPWR cable adapters so the cards can survive the warranty periode? Better say, let the consumers have new 12VHPWR adapters each 8 months of usage.

 

Lets say they all could ship the cards with 3 pieces of the hated tiny adapters in the package? 

 

Btw. The better and more costly PSU's often have two 12V-2*6 connectors due AI/creators... There you have it. You have one spare cable all time for your precious GPU. Useful even if you don't run 5000 series cards in pair for AI tasks. Aka you have always an backup if you prefer the tiny trash on both ends.

 

Two RTX 40 series cards left us this week, an RTX 4090 and an RTX 4080: soldered plastic connectors, graphics cards on the verge of warranty expiration.

 

They're failing faster and faster. Time waits for no one; the 12VHPWR and 12V-2x6 graphics cards should be recalled now, while there's still time, before they continue to silently burn out one after another.

 

 

 

As you can see... The engineers at nvidia is fantastic. Same pattern here as with the tiny trash... Amazing engineering skills. Fragile as hell.

 

[Mr. Fox]

Hopefully this was not a waste of sand. But, I will find out as soon as I button things up. If it's not better than what I already have it's going back for a refund.

[electrosoft]

8 hours ago, Papusan said:

 

Nice. The new Intel flagship got one boost bin upgrade of 100 MHz for TVB. But no change in all core boost clocks(Turbo oost max). Everything else is almost exactly the same. What a fantastic upgrade. Run and buy. Why bother put resources on an refresh? Plus models? LOL

 

 

Intel Core Ultra 290K, 270K and 250K Plus spec leak: “Arrow Lake Refresh” with higher clocks, more cores and faster memory support

 

Core Ultra 200K Plus (overview).

Core Ultra 9 285K➡️ 290K Plus: + 100 MHz P-Core (Max)

Core Ultra 7 265K➡️ 270K Plus: + 4 E-Cores

Core Ultra 5 245K➡️ 250K Plus: + 100 MHz P-Core (Max), + 4 E-Cores

 

7 hours ago, Mr. Fox said:

Did you mean to link to a post by @electrosoft or an article about the shiny new product that offers so little? The link leads here:

https://notebooktalk.net/topic/109-official-benchmark-thread-post-it-here-or-it-didnt-happen-d/page/824/

 

 

I'm living rent free in @Papusan's mind!  🤣

 

As for Intel, that is classic Intel offering the exact same chip with an extra bin or two or tacking on more e-cores and either calling it a new model # or appending an S on the K......

 

290k Plus = 285KS 🤣

 

-----

 

When you are shopping for a new system, really want a 5090 based desktop and score a fully kitted out 285k + 5090 CyberPC open box for $2999.....

 

Even included the KB/M and with a Gigabyte Gaming 5090 currently going for $2900 on its own....

 

(Timestamped at the "score"):

 

 

---------

3 hours ago, Mr. Fox said:

I've been watching the insane increase in DDR5 prices. I don't need any, but decided to see if I could find some spares cheap to hang onto. Just scored a Trident Z5 DDR5-8000 48GB CL40 1.400V kit with Supercool direct touch cooler on eBay for $250. That same RAM kit is now about $300 sold and shipped by NewEgg. So, basically like getting the Supercool cooler kit for free, give or take a couple of bucks.

 

 

I'm kicking myself for selling off multiple sets of redundant 6000-6400 kits back in July but then I realized they all sucked badly.

 

Right now I have the TG 8200, KS 8400 and a set of G.Skill 8200 I sniped locally. All 2x24GB kits. I think I should be good for my personal use.

 

Any build outs for friends and family, well, they are in for a world of hurt when they see the price of that memory.....

 

1 hour ago, Mr. Fox said:

Hopefully this was not a waste of sand. But, I will find out as soon as I button things up. If it's not better than what I already have it's going back for a refund.

 

Oh damn! It's here already. Definitely hoping it turns out to be a better bin. Good luck!

 

 

[Mr. Fox]

5 hours ago, electrosoft said:

Oh damn! It's here already. Definitely hoping it turns out to be a better bin. Good luck!

I still have a lot to figure out for the best settings for tuning of this new EPYC CPU but it does appear to be better than either of my good but average 9950X samples. Initial impressions are that it is a pretty decent sample. I think my memory kit could be a little better, but I was surprised it could run 6400 1:1 and FCLK 2200. I haven't spent much time with core tuning yet, just set a simple PBO, but will definitely need a good manual overclock to see any respectable Cinebench results. I will play with it more tomorrow to confirm it is worth keeping before I delid it. I always forget how crazy hot these CPUs are that have not been decapitated and run bare die. There are still many hours that need to be invested to find what works best. Requires a lot more time and effort with AMD to get things figured out.

[Mr. Fox]

One thing I find very puzzling about the Apex is that it does not matter what CPU I have installed, my Cinebench scores are always notably lower than the Strix. Every CPU that I test consistently has stronger Cinebench results when I install it in the Strix. This is my third Apex. I don't really understand why this is the case, but it is consistent. Same was true of the AORUS Master... consistently higher Cinebench scores than the Apex using any given CPU. This really bugs the crap out of me and makes me wish I had saved $300 and just purchased another Strix X870E-E or just kept the AORUS Master.

 

I think I am spinning my wheels with the EPYC CPU. It seems like a much better sample overall, but I am not going to be able to really confirm that until it is delidded and bare die. Absolutely ludicrous thermals with the factory solder and IHS. It's unfortunate that most CPUs suck with the factory solder and IHS. It's true of both Intel and AMD, but seems much more so with AMD for some reason. They seem a lot more sensitive to high temperatures. 

 

Jury is still out whether I like 3DvCache. The 9950X3D that I RMA'd was such a pathetic silicon sample that it wasn't fair to draw any conclusions based on what a piece of garbage it was. I'm leaning toward either not liking it or viewing it as an irrelevant selling point/gimmick. (I'm a casual gamer, not a gaming enthusiast, and it seems to provide no tangible benefit for an overclocking enthusiasts.) I do see higher FPS in the few games I tested. 3DMark doesn't seem to benefit.

[Papusan]

 

Double up is the way to go😎 Sure.

 

 

 

[Papusan]

On 11/7/2025 at 8:19 PM, Mr. Fox said:

If AMD doesn't keep their prices in check they'll have nothing to sell to budget-conscious shoppers that are willing to sacrifice performance and good thermals to save money.

 

Is there something that is called budget gaming nowadays? AMD, walk their own way... Maybe they even beat Nvidia increasing prices first? Better than letting Nvidia decide the price of Radeon cards?🤔

 

AMD rumored to raise GPU prices just as Radeon RX 9070 XT finally reaches MSRP

 

AMD & NVIDIA Could Kill Off Budget GPUs as Memory Shortages Drive Costs Up, Leaving Entry-Level Gamers With Little Options

 

 

And used will also cost you more... US has now become Norway. Tax hell. Everything I buy from abroad is taxed very hard. Used or new doesn't matter. 

 

Vintage PC parts are getting hit with huge tariffs, even when they're worth almost nothingA $355 box of retro Apple boards ended up with a $684 tariff

 

 

[electrosoft]

On 11/2/2025 at 7:44 PM, electrosoft said:

 

This really makes you think with everything done right, and oodles of power slammed through the connection, it is just fine.

 

I would say it is human error coupled with potential QC, but I also believe after the 4090, OEMs/AIBs definitely test their connectors under high stress conditions including MSI. 

 

I also still stand by the fact lighter colored connectors will show scorch marks much easier than black connectors and to see if a black connector is degrading you will need to either get it to a point of melting and/or have the actual pins scorched too.

 

I would posit that there are probably equal parts black connectors scorched out there as there is lighter color connectors. You just can't detect it as easily with the naked eye.

 

 

I do not subscribe to the "yellow tip disaster" theory quite yet.

 

We are talking anecdotal meets the potentially lowest level of confidence. It just doesn't fly.

 

I do subscribe to the "a very small segment of connectors in general burn but we  just quite don't know why yet" theory which applies to both Nvidia and AMD cards that have adopted this connector and includes MSIs cables along with other ones.

 

In that line of questioning and wondering about cable/tentacle QC, how many founders edition 5090s / RTX 6000 pros have burnt up using the supplied Nvidia adapter cable? You would think their adapters would be the gold standard in all of this....

 

 

Following up on my post a few weeks ago about burned 5090 connectors using Nvidia's official adapter.....

 

I've been doing deep dives into finding any Nvidia 5090 Founders using the supplied Nvidia adapter that have burned and I haven't found anything substantial yet. It is always third party adapters / PSU cables that burn. Asking Google AI and it came up with the same results:

 

 

 

Of course we know while using the card w/ the supplied cable isn't a guaranteed solution, it does make it easier to get the card repaired/replaced from the AIB since there are no third party cables involved if using the supplied AIB cable or same brand PSU.

 

 

---------------------------

 

9070xt prices are as good as they are close to getting. Even Sapphire's Nitro is now officially $100 below MSRP along with the Taichi 9070xt $50 below MSRP and the Taichi has black friday protection in place at NE for all Asrock 9070xt models which are already all priced below MSRP so you know you're going to get some coin back after BF.

 

Plenty of $999 priced 5080s and $749 5070ti's with a few models below MSRP. Zotac 5080 has black friday protection even with its $999 pricing.

 

Same goes for 5070's with several models below MSRP.

 

I'm expecting some serious deals on black friday at this rate as there is a glut of GPUs and especially since it seems the Super series has been pushed way back with memory pricing uncertainties there's no need to wait now if you need a primary or even secondary GPU.

 

I'm leaning towards picking up a Nitro+ 9070xt as a backup/secondary/SFF model especially as the recent batch of Taichi's are using Samsung memory....

 

If anyone is in the market for a card not named 5090, now is the time to buy up to and through black friday.

 

Cheapest actually available 5090 is the $2800 Zotac AIO model in the US.......

 

Bonus: Extended return windows are now in effect at da egg and amazon....

 

 

 

[electrosoft]

On 11/17/2025 at 8:11 AM, Mr. Fox said:

One thing I find very puzzling about the Apex is that it does not matter what CPU I have installed, my Cinebench scores are always notably lower than the Strix. Every CPU that I test consistently has stronger Cinebench results when I install it in the Strix. This is my third Apex. I don't really understand why this is the case, but it is consistent. Same was true of the AORUS Master... consistently higher Cinebench scores than the Apex using any given CPU. This really bugs the crap out of me and makes me wish I had saved $300 and just purchased another Strix X870E-E or just kept the AORUS Master.

 

I think I am spinning my wheels with the EPYC CPU. It seems like a much better sample overall, but I am not going to be able to really confirm that until it is delidded and bare die. Absolutely ludicrous thermals with the factory solder and IHS. It's unfortunate that most CPUs suck with the factory solder and IHS. It's true of both Intel and AMD, but seems much more so with AMD for some reason. They seem a lot more sensitive to high temperatures. 

 

Jury is still out whether I like 3DvCache. The 9950X3D that I RMA'd was such a pathetic silicon sample that it wasn't fair to draw any conclusions based on what a piece of garbage it was. I'm leaning toward either not liking it or viewing it as an irrelevant selling point/gimmick. (I'm a casual gamer, not a gaming enthusiast, and it seems to provide no tangible benefit for an overclocking enthusiasts.) I do see higher FPS in the few games I tested. 3DMark doesn't seem to benefit.

 

That is very odd. How much lower? Margin of error lower or substantially lower?

 

Do you delid your own AM5 CPUs or send them off?

 

X3D is very much hit or miss and it really comes down to over saturating it and watching performance become uneven. It is a one trick pony aimed at gaming. Perfect example really is Fallout 76. Older engine with such small assets and it just destroys Intel even at 4k but take something like the modernized WoW engine or even Starfield at 4k and you can get those drops when the cache is being over saturated. Overabundance of player data in WoW can render X3D moot many times and heavy populated areas or raids with a ton of player FX popping off will tank just about any CPU in the face of a 5090 even at 4k. You will see GPU utilization all over the map from 99% down to the 50's depending on what's going on.

 

tl;dr when it works, oh mama! when it doesn't, ugh.

 

 

[Mr. Fox]

1 hour ago, electrosoft said:

 

That is very odd. How much lower? Margin of error lower or substantially lower?

 

Do you delid your own AM5 CPUs or send them off?

 

X3D is very much hit or miss and it really comes down to over saturating it and watching performance become uneven. It is a one trick pony aimed at gaming. Perfect example really is Fallout 76. Older engine with such small assets and it just destroys Intel even at 4k but take something like the modernized WoW engine or even Starfield at 4k and you can get those drops when the cache is being over saturated. Overabundance of player data in WoW can render X3D moot many times and heavy populated areas or raids with a ton of player FX popping off will tank just about any CPU in the face of a 5090 even at 4k. You will see GPU utilization all over the map from 99% down to the 50's depending on what's going on.

 

tl;dr when it works, oh mama! when it doesn't, ugh.

 

 

I delid my own with the der8auer tool. Works fantastic. You just have to spend about 30 minutes moving it back and forth with the tool until it is loose. Tedious but foolproof as long as you keep going until it falls off and don't try to lift the IHS before it is loose.

 

Usually in the range of 500-1000 Cinebench R23 points lower with the Apex. AMD CPUs are inconsistent between Cinebench runs and can have a wide variance between runs with the same CPU. There are too many artificial algorithms in play with Ryzen. But, I can't get within 500 to 1000 points of my best scores with the Strix or AORUS Master using the Apex with the same CPU installed. It almost has to be something in the firmware. It would not seem logical for something with the motherboard hardware to affect it. Also worth mentioning that it is with the same Windows installation. To rule out an OS issue I create an image from one, restore it using Macrium Reflect on the other and reactivate. So all of the same software, tweaks, drivers and OS tuning on both machines. This makes no sense to me. There is probably something I need to do different that I don't know about. Maybe an obscure BIOS setting I need to turn on or off on the Apex. This should be an almost entirely mathematical outcome based on core count, IPC and clock speed as long as something is not interfering with the Cinebench run.

 

This EPYC CPU seems like a really good sample and I am pretty sure I am going to keep it. It seems better than average looking at the V/F curve and voltage requirements in practice. Getting it to function like an X3D in a scenario where it should has not worked for me yet. When I try to configure it to use the 3DvCache CCD it is defaulting to the non v-cache CCD no matter what I have tried so far. Configuring it to function like an ordinary CPU it seems fine, so I suspect this is a firmware configuration problem, an OS/game/chipset driver bug issue, or even an ignorant noob issue on my part (relating to firmware configuration). I did a clean OS install with the latest chipset drivers and set everything that is supposed to be how it is done that I know of and it still parks the wrong CCD and activates the one without v-cache, LOL. I don't care about the v-cache necessarily, but if I can use it as leverage for 3DMark scores I certainly want to take advantage of it. I don't expect it to matter much at all with something like Cinebench, wPrime, Pifast,  or Y-Cruncher.

[Papusan]

1 hour ago, electrosoft said:

Asking Google AI and it came up with the same results:

 

 

This ain't AI overwiev. This is real. The winter is here to stay for next +6 months. And I hate it even more than last year. Already come above one meter of this white beauty trash the last few days. Some love it, but it won't be me. Wish this was a fake snowy picture from my phone. But nope, it's equal real as the melted 12VHPWR connectors that pop up in pictures. 

 

Taken from the window in the living room.

 

 

 

In God we have trust, LOL

 

[Mr. Fox]

7 minutes ago, Papusan said:

 

This ain't AI overwiev. This is real. The winter is here to stay for next +6 months. And I hate it even more than last year. Already come above one meter of this white beauty trash the last few days. Some love it, but it won't be me. Wish this was a fake snowy picture from my phone. But nope. 

 

You hate cold and I hate heat. My body has adjusted to the hot weather here, but too hot is a health risk that is equally hazardous as freezing to death. Extreme heat and extreme cold are both an issue. Temperature aside, I love living in the desert because I do not like precipitation in any form. I hate rain and I hate snow. I like it to be very dry, clear skies and low/no humidity whether it is hot or cold. I find it easier to get warm if I am too cold. I can bundle up and add more blankets to the bed. If I am too hot it is harder to fix that problem. You can be wet and naked and still be too hot, LOL. Both can kill you.