*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Even a "dud" 5950x will smash everything Intel has to offer right now. Your 5900x might just be super magical.

 

In single threaded performance yes. As for multithreaded rendering or any heavy loads my 7980XE can out pace the 5950X (Fairly well)
Not without some serious downsides though!
But still, this cpu will be 4 years old next month so
I suppose it’s still pretty amazing.


It’s funny how this type of performance hasn’t really made any sense until lately. Especially with Games like Death Stranding, and Cyberpunk 2077, which can use every single core and thread on the bigger chips and maintain that steady GPU usage when the action gets going.

Anyways, I might just go grab me a 5950X too!! Especially now that the awesome X570 Dark motherboard is coming out.

I would seriously love to try out the new Ryzen platform my self.

 

See, the fact they didn't throw a 16 core onto the TRX40 until the non-overclockable TR Pro is part of what made me decide not to get the platform. That is more to do with the 5950X's lack of I/O. I mean, after having 40 lanes on the 7980XE and the 64 lanes on the TR platform, going back to the mainstream is infuriating (plus mem bandwidth for some things). But, if that is not your only platform, then having both Intel HEDT and AMD's chip can make sense, just use whichever computer has your extra cards in it or that is meant for your purposes.

I said when AMD pushed the 16-core onto mainstream and had nothing on HEDT until 24-cores, that was the biggest saving grace AMD could have given to Intel. That's because they kept the X299 platform relevant. If they truly killed it by offering a 16-core on mainstream and a 16-core on HEDT, AMD would have cut against Intel's value on the CPU and MB, as well as offering mainstream a way to step up, while you had HEDT for the I/O and mem bandwidth.

Because of that, I'm sticking to my word in not buying AMD until Zen 4 and/or the next TR platform (they are creating a workstation platform with 8-channels like the server chips, but that line will probably be the TR Pro line moving forward, which makes sense; no idea if 8-channels will be brought to consumer channels for HEDT). But, with doubling the memory bandwidth on DDR5 for Zen 4, it will then be equivalent to quad lanes on DDR4, which I grew accustomed to with the HEDT platforms for bandwidth per core. And that means Zen 4 DDR5 TR will have the same memory bandwidth per core that Epyc had all this time!

But I'll probably buy a next gen GPU next year (as soon as released or getting on a list) before upgrading platforms. Unless cards from cryptofarms crash the GPU market and I buy a used current gen card. Time will tell.

Either way though, we are now having a lot of new tech hit the market quickly. Let's just hope TSMC can keep their schedule on opening their mega-fabs!

 

Old 14nm +++++ will struggle on the bot. And it doesn't matter if its in multi threaded rendering or 3d benchmarks. As it is, not much Intel can compete with. Maybe when they start offer phone cores and Alder lake... For power effiency. And mayby gaming and single threaded workflow. Bur don't forget a 2022 chips from Intel compete with an 2020 chips from AMD.

https://hwbot.org/benchmark/cineben...d=processor_6347&cores=16#start=0#interval=50

https://hwbot.org/benchmark/cineben...d=processor_5458&cores=18#start=0#interval=20

Sorry, but Intel has dreamed that they could ride at 14nm ++++ forever/in eternity. It will may change with the sub 10nm small phone cores big.Little as you know. OEMs will probably be damn happy they finally get some power efficient cores.

Edit. The older Ryzen 3950X is from 2019. Not the best gaming chips but even with this chips Intel HEDT still struggle to beat it in CBR20 multi threaded rendering https://hwbot.org/benchmark/cineben...d=processor_5887&cores=16#start=0#interval=20

 

Yes I'm taking it easy bro. I'm spending most of my time sleeping with a fever and headache. Anything that I put in my mouth, including water, is nauseating.

What area in CTR tells you that you have a golden chip? Mine also has idle temps between 38 and 41°C at 4.7GHz. My 10900k idle temps were 5°C lower at 5.2GHz.

It's really unfortunate that this can't be delidded and run bare die. I think it would be so much better if it could.

I also consider the AM4 socket design to be inferior. It's unfortunate that AMD stuck with that obsolete model to pacify people that wanted to have a perpetually upgradable system. Intel's design having the pins on the motherboard and the ILM retention mechanism is far more intelligent and less prone to accidental damage.

 

In CTR 2.1 RC5, go to the Tuner page. Click Diagnostic (I keep CB20 testing enabled) and let it do its thing at stock settings. At the very end of the diagnostic test, in the Info pane, it will give you a rating. I, now, have the 5950X, 5900X, 5800X and 5600X. All but the 5600X are golden samples. The 5600X is a bronze.

According to the manual "The category can vary depending on several factors: DRAM stability, AGESA version, UEFI settings, processor temperature, or VRM temperature"

I run the diagnostic test a minimal of five times at stock settings since the grade could possibly change due to those factors mentioned above.

 

Thank you bro. When I get past this COVID mess I'll have to check that out and see what it says about my CPU.

 

Oh man I would consider that much more complicated than my issues when looking at it like a check list. throw in a confused and angry little infant and suddenly it doesnt matter how long that list is lol I would say your challenge required much more endurance to overcome than my own, but here we are on a laptop forum parsing details on computers, performance and if said products are justifiable under typical or even untypical consumer use cases under current or future market forces.

It took me a long time to trust anyone enough to learn from them, My 3rd grade teacher told me I would never succeed in life if I didnt learn to write in cursive, lovely lady Mrs. Carrillo. Then a slew of hmm family eroding issues that started when I was 4 and didnt end until I was an adult. But yes, luckily my 5th grade teacher, Mr. Woods, refused to let me continue making my mistakes. Simple things like increasing my indent progressively on each line, ain't vs aren't / can't, shutting down altogether. The biggest leap was his endless encouragement to find a book to read, any book as long as it was a 100+ pages and tell him what happened in the book. Which is one of those requests people hate to do, because explaining something in your own words is also a rewiring of sorts that is beneficial as well.

All that being said, I feel bad for those teachers lol, I was not a cooperative student.

 

If a 4 year old 7980XE cpu on (2) thin 360’s, can compete with (top 10 hwbot) 3950X on LN2, then I am totally satisfied!! @Papusan

Pretty sure I am sub 12K in R20. Maybe like 11,800-11,900. Just a normal daily stable settings run.

Yes It’s 14nm, and yes it’s old lol. Nothing I can say about that one haha. As long as it performs good, and doesn’t bottleneck my GPU is all that matters to me.

 

I'd be curious to see out of the box 7980xe vs out of the box 5950x numbers. Like you said, single thread is no contest at this point, but 5950x @ 4.7ghz all core vs 7980xe 4.7 all core....that would be fun to see how those 18/36 cores stand up to 16/32 cores.
When @Mr. Fox is feeling better and dials in his 5950x and chiller maybe a nice little shoot out.

Of course if you're pitting your OC'd 5.0 all core against a stock 5950x, that's the same argument I use with the 11900k vs 5800x too.

I'm going to wait for the refresh with better performance and dive right back in. I had no complaints about the 5800x. The Tomahawk on the other hand......in hindsight I wish I had ponied up the $$$ for at least an Asus.

As it stands now, with the x570s Dark coming out and refresh on the horizon, that will be my next potential move unless Intel suddenly dazzles us (unlikely) with Alderlake.

 

@Rage Set I cringed when he plugged in the GPU:

 

I wouldn’t under estimate single threaded on them either. With proper memory it’s practically a 10900K at 5Ghz. I can manage 230 R15 single. That’s around 30% faster single thread performance than a stock 7980XE.

@Mr. Fox already knows. He had his 7980XE screaming faster than mine. Their just power hogs. The 5950X makes way more sense. This is a newer and more efficient cpu, it’s practically doing what I’m doing, only easier. And the 5950X can put out some good numbers in physics test’s and things like that too.

At 4.7Ghz with a 7980XE vs a 4.7Ghz 5950X? Its very comparable in straight multithreaded loads. R15,R20,R23 etc. Seem to be very similar performance at the same frequency.


As for single thread performance, it’s around the same in R15 as a stock box 10900K’s (Without optimization or tuning). They boost a single core at 5Ghz right? So I’m right at 230 in R15 single thread at 5Ghz..

They are both 14nm. Intel really, and actually, didn’t do anything for IPC at all lol.


The 7980XE is a 3.4Ghz cpu all core stock. So when you over clock it by 45% it also becomes 45% faster too.

 

Who is that?

 

I have a serious case of "punch someone in the face" syndrome from that video. I don't understand how someone with access to that card doesn't have more knowledge of computing. "graphic card power cord extenders"......

 

A documented train wreck.

 

That was hard to watch start to finish but plugging it in with power delivery still engaged was the topper for myself.

 

I only corrected your statement Yep, 7980XE is still a great Cpu No need to upgrade before you see something much better or can help on your workflow.

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

It seems the crypto-farms got all the love from AMD and Nvidia with their AIB partners. Much easier sell graphics cards in bulk with the " agreed correct" prices. A very smart way to keep prices 1.5-3 times the MSRP.

Cryptomining Firm Purchases Site With 223,000 GPUs tomshardware.com | Aug 12, 2021

What we do know is that 223,000 GPUs is about 100 times as many GPUs as what all of Micro Center received for the Radeon RX 6600 XT launch. And this was just one mining firm selling off some of its hardware. Based on the Ethereum network hashrate, it would take five million GeForce RTX 3090 cards to power the network; ten million GPUs like the RTX 3070 and RTX 3060 Ti, or RX 6800/RX 6800 XT/RX 6900 XT; or 20 million GPUs like the RX 6600 XT and RX 580 8GB. Small wonder GPU prices remain inflated.

 

Zen 4 / AM5 are a land grid array (LGA). So that complaint disappeared next gen. The ran for PGA wasn't for long life of compatibility. They could have done an LGA and still had compatibility, just like Intel could have but instead added a pin or changed the pin assignment just to force buying a new motherboard.

 

This still grinds my gears......

 

What the hell did I just watch?

 

I am glad to know that I am not the only person that holds them in contempt. Stupidity at that magnitude should be illegal.

 

I wasn’t gonna be the first to say anything but yeah, what the hell did I just watch too? . That’s pretty cringe worthy. Not only that, but why are beginners testing a 6900XT on a X299 Dark platform?

This guy just said he doesn’t know much about overclocking, so I can only assume this platform is not properly optimized at all.

Which it would certainly need full optimization to demonstrate competitive performance with other reviewers of 6900XT’s on the internet.

That’s gonna make that 6900XT look pretty rough lol.


Worst review ever! Lol

 

Well, if CTR's rating is believable, looks like I have a crappy "silver" CPU sample. I am not sure that it is accurate, but I am running Legacy BIOS. Maybe it won't accurately rate it since I am not running UEFI.

 

I think I am going to sell the Z490 Dark Kingpin mobo, my sweet 5.5GHz 10900K, bare die setup with OptimusPC Signature block as a package deal. If I am feeling better tomorrow, I will list it in the NBR Marketplace.

 

Check out this massive Fujipoly thermal pad. This pad covers the entire rear of the RTX3090.

it showed up in the mail today. It is a 3MM thick full coverage. It’s a medium quality Fujipoly pad, which is perfect, because I need the 3mm to squish down some.


Gonna slap this on the back of my 3090 Kingpin underneath the backplate.

This thermal pad is for the FTW3 Optimus waterblock.

 

Looking forward to the results.

 

This guy's build videos are next level and you can see how they've just gotten better and better over the last few. I can watch this stuff all day.

 

Very fancy build. The guy is very talented.

But, I don't understand how that build would cost $15K. The cost seems excessive in light of the specs.

 

Labor?

Hope you're feeling better @Mr. Fox

 

The one he released today on youtube said 30k build. It has a :1600 gpu, $300, mb, $550 cpu, antec 1k crappy psu and a $350 case. Couple rads, pump and line still nowhere near 30k

 

Thats a pretty budget build there, good to showcase these things so consumers can follow in their footsteps.

I myself tend to never be able to keep my budget below 100k, I dont know why but it always works out that way

/s

 

Honestly if someone were to spend money like this. They’d be better off getting an 8 pack build or something.


He has really nice systems that are well tuned with binned silicon too. The guy is known for overclocking. And apparently he tunes each system him self. His build are expensive, but when I look at the builds, and the specs, I feel like I’d be getting at least a little something for that huge chunk thousands lol.


https://www.overclockers.co.uk/pc-systems/prestige-pc/8pack

 

FS: COMBO: Z490 Dark K|NGP|N, 10900K, Bare Die Kit, OptimusPC Signature V2 Block

 

Yeah, looks like I won't be buying any more Gigabyte products in the future.

 

Yep, I remember when Gamers Nexus first started exploring this. What a pathetic response to an obviously faulty AND dangerous product both for hardware and human life. The only thing they will respond to is people voting with their wallets.

Sent from my SM-G970U using Tapatalk

 

The sum up...
PSU Expert Aris Mpitziopoulos Responds to Gigabyte's Exploding PSU Problem tomshardware.com

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

EVGA Z590 Dark Motherboard Review: Improved Aesthetics, Premium Hardware tomshardware.com | Today



A full-featured overclocking flagship, with an improved appearance and some unique features

Final Thoughts
The EVGA Z590 Dark arrived on the scene later than most Z590 competitors, but I feel the wait was worth it. Like the Z590 FTW WIFI, the Dark’s appearance is much improved, along with the hardware. While the looks have certainly improved, the company didn’t forget about premium hardware or the overclocker either.

Overall, we like how EVGA updated the EVGA Z590 Dark. You get a much-improved and unique aesthetic, robust power delivery and plenty of storage options, including a rare-on-consumer-motherboards U.2 port. If you require Wi-Fi 6E or want the latest and greatest audio, you’ll have to look elsewhere or use an add-in card, but those are hardly deal-breakers. If you have $600 to spend on an overclocking-focused build, EVGA’s flagship Z590 Dark deserves to be on the short list.

 

They totally missed the mark on this review. It is like me reviewing a track car and I lower the score because it lacks heated leather seats. Smh.

 

What are talking about? Everybody knows you do track season in the winter, I've been waiting on heated seats for forever now!

/s

But hey at least the aestetics were improved on my overclocking focused motherboard. A rather odd line of thought

 

Nevermind went a different route

 

It seems like Intel really really want the "single-threaded" crown, from this rumors More cores as you can see isn't always needed to be the king.


Rumor | Full list of Intel 13th gen Raptor Lake SKUs leak, Intel apparently looking to hit a world record 5.5 GHz turbo and retain single-threaded dominance with the Core i9-13900K notebookcheck.net | Today

Bro @Mr. Fox Hope you feel better

 

To be fair, single core performance is what matters the most in the overwhelming majority of use cases, both for normal users and power users. The more single core performance we get, the better.

I don't think this will change anytime soon, even with the increasing number of programs that are well multithreaded (including games). Some operations are serial by nature, so it'll probably be a while before we figure out a way to parallelize them or pseudo-parallelize them using some clever tricks.

If you want insane throughput, the GPU is the better processor for your needs anyway.

I do however applaud AMD for its efforts. The competition is forcing Intel to actually innovate, which means no matter who your favorite team is, you win as the consumer.

 

Cores are great if you run VM’s and some other tasks like video editing to a degree and some other productivity applications that the majority of enthusiasts don’t even use their computer for. I know I run a lot of vm’s but I’ve learned over the years heavy overlooks will show their ugly side running VM’s, so I generally don’t overlook when running vm’s.

 

Intel Architecture Day 2021: Alder Lake, Golden Cove, and Gracemont Detailed

Bro @Papusan if you want to get latest Intel technology of big little bs, make sure you run Windows 11 ONLY. Forget running Windows 7 on this franken CPU design, RIP. I assume LTSC 20H2 won't be getting either if Win10 doesn't get. M$ is going to make a ton of userbase and cash from this move, worst OS, worst CPU topology with all the drama of cores, DRAM, PCIe guinea pig perfect recipe.

ADL Architecture is out. Anandtech covered it, and the key points are - PCIe system (5.0 and 4.0), Memory System (LPDDR, DDR4, DDR5), Interconnect fabrics expect DRAM controller to be like RKL Gearing crap and the biggest of all Intel Thread Director, which will leverage Windows 11 scheduler AND Intel's Hardware Scheduler. Also their new Golden Cove design (P cores) having a 19% IPC with very wide design vs Cypress Cove inside Rocket Lake system. Even RKL had that boost of IPC and it was all lost in latency and Gear system, I wonder how this will scale. Also their Skylake class / Gracemont (E cores).

Here's the real deal about those phone cores and important takeaways, there's more stuff like super wide and ton of ports etc etc CPU engineering which pretty much flows over many minds.

Memory and Bus and the PCIe systems, lot of drama.. But notice the chipset link speed 4.0x4 which is same as RKL PCIe3.0x8, why is this ? Why not a Gen5.0 DMI....

Oct 2021 is the month for benchmarks and reviews. I certainly have no interest in this platform as a buyer because it introduces too much OS related changes, HW changes to the fundamentals and brings a lot of new things which Intel is trying out for the first time, also it screams more of a Mobile oriented design than a performance which is why Gracemont cores are more optimized for voltage as Intel themselves say plus the big cores are locked at 8C16T max, they should be better than 10900K and 11900K and 5800X in ALL workloads from Gaming to Productivity if not then it's going to be a huge disaster. Plus there's a lot of segmentation here on PCIe / DRAM system which will cause a lot of pricing variations, I would simply avoid this. Whoever owns any Ryzen 5000, Intel CML, RKL should also avoid this. There's no reason unless you want to bludgeon yourself with all these new changes and market dynamics with software adaptation. And Intel has to prove it. Remember there's that Zen 3D V-Cache refresh also coming, which AMD definitely showed in confidence vs this ADL.

Also these Atom cores are non existent on the Sapphire Rapids Xeon platform (maxed out at 56C112T)which will have AVX512 and full Golden Cove Design no phone cores nonsense, which must compete with Milan / EPYC 7763 which is the king of the Enterprise right now probably that's where Intel's new HEDT will debut, maybe 2022. Note Genoa leaks are out, Zen 4 will have full AVX512 and it will be a 96C192T monster, a true successor to the EPYC lineup.

 

Hi and Thanks. Intel have created a monster. But not in a good way. Worst of the worst design for maximum performance. Not sure about the OS yet. I hope some will come up with an solution. Intel even screw Linux users with this disgusting design/direction. With all the problems on the bot, I can't see this will make it any-better.



A new way to address CPUs (Cores/Threads) on HWBot

 

Since the W-1390P ( 11900K ) launched in Q2 2021, you'd probably want the 12900K ( W-1490P ) at launch ( since they'll bin the good ones for the W-1490P ( 12900K ) before Q2 2022 ).

 

Many people just simply guess that AMD is have to go this route eventually. I never got understand that statement because, AMD Zen is a big x86 design which scales upto 64Cores right now in production, but the speed and power limits so they cap at 16C on the mainstream socket. Now with the Big little direction, Zen 4 TSMC 5nm from Gigabyte leaked data suggests there's nothing like that, Zen 5 (TSMC 3nm ?) there was a leak but that was on the APU/BGA side. Zen 6 is way long in the future we are talking so many gens ahead, assuming that's on TSMC but which node again ? Too much assumptions from these folks...

Intel Sapphire Rapids which is the new Golden Cove Xeon peaks at 56Cores max, that is using a new Mesh System from Ice Lake where Intel finally fixed the power wastage vs Infinity Fabric. We all know Ring bus is the fastest of all from Intel for High Clocks, Gaming and only Skylake was able to scale past 8C. I think this is the primary reason why Intel had to do the Rocket Lake backport to clock higher and this Skylake shrinked into this design then add them to the Ring Bus system and pair with Golden Cove a Frankenstein design. They simply do not have any route except this to fight against AMD on Mainstream socket.

But AMD on the hand does not have this limitation since their designs already use the Infinity fabric which scales a lot in both Mainstream, HEDT and Enterprise now 64C Zen3, future is 96C on Zen4. Also going this route means more BGA nonsensical market sadly which is the peak revenue generator probably this is the primary reason why Intel is heralding this efficient design.

Then there's that whole Intel 10SF / Intel 7 now TSMC nodes etc. I just hope AMD doesn't go this nonsensical route on the desktop platform and just give it all. If they do, going Zen1 or 2+ cores on a Zen 5/6 design sounds like jank and a copy of Intel (which is a copy of ARM) so far AMD did not copy anything from Intel based on their Zen series and Threadripper / EPYC series. Better they up the ante to next level and implement Xilinx specialized FPGA onto the CPU design and aim for stars. I just really wish that happens.

 

Yes, but keep in mind frequency is not everything. Instructions per second and instructions per cycle are better measures. Why? IPC is the number of instructions processed every cycle. Frequency is the number of cycles in a second. So, if you multiply the IPC by the frequency, you get the number of instructions in a set unit of time, which arguably is what matters.

I still need to review the announced architecture changes from this morning and the perceived and Intel started gains from those changes. But, AMD is making moves on increasing more IPC.

An example of higher frequency not being better is AMD's FX processors that you could overclock the snot out of, but still getting beat by Intel's lower frequency, higher IPC chips.

So Intel announcing a faster single core speed, without me having looked at release date, architectural changes, etc., means very little to me. 300MHz on a single core is 6% over the 5GHz boost clock they have made common for their single core. Although impressive to increase the speed of single core 6%, many workloads or even the OS working alongside can cause two cores to have load and exit that state.

Just wanted to note frequency, by itself, isn't innovation (unless revolutionary changes on process). It is iteration, which is still progress.

If you think Intel's doing this to compete with AMD, you miss the point. Intel is doing this to compete with ARM. Apple dropping Intel, talking about how crappy their silicon was, and then creating the ARM M1 heavy hitter with awesome power consumption numbers with good performance, then rumors of the M1X and M2 chips, with some variant having 40 cores, Intel got scared and needed to show you cannot do what X86 does in fears other companies would start developing their own ARM and RISC-V chips that are specialized for their purposes instead of a generalized CPU architecture.

This fear is also on Nvidia's bid to buy ARM, meaning Nvidia would develop an entire platform where you could buy all team green and jump the boat off of x86 entirely. Nvidia could make smaller, more power efficient systems which also would be powerful and have brand loyalty to Nvidia. We all know Nvidia can engineer good products. Imagine if you reduce licensing costs what putting out arm platforms could earn them. And with Intel the predominant player in mobile and small form factor, it would be eating them from below while AMD eats them from above on server and HEDT.

Then is AMD on this front. AMD will include the small cores on their own chiplet. So you will not hear much about it. With an I/O die, this controls the routing. And AMD has been rumored to have developed 2 different I/O dies. Even Intel has to develop a special router similar to an i/o site routing system to say go to small or large cores. AMD could develop the I/O die to route to small and large cores, develop a small core chiplet, and release it without much being heard on it.

Further, that allows AMD to keep their server lineup pure, same with HEDT. They could put solely large cores on their flagship desktop, then lower skus with a chiplet of each big and small cores. Then move the mobile lineup to chiplets and do more small cores, if they can solve the power draw of infinity fabric.

But that's just how I'd do it if AMD.

 

With Skylake, Intel kept pushing higher and higher frequency with no change in IPC. (6th gen to 10th gen)
Unlike AMD, they aren't sacrificing IPC for increased frequency. Their new node is just that good for high drive current/frequency and it clocks even higher than the FX chips to boot.

 

"Why even develop desktop Alder Lake chips with efficiency cores,? “You’re absolutely right—battery life doesn’t matter in desktop,” said Stephen Robinson, a CPU architect and Intel fellow. “But the thermals do. Fans, cooling power—at some point, you have a limit.”

And that’s why Intel designed its new efficiency cores.

There you have it. Power efficiency over pure performance. Intel still struggle with High power consumption. The great fix is the added small cores

https://www.pcworld.com/article/3629502/intels-alder-lake-what-you-need-to-know.html

 

I feel Intel is probably going a similar direction like Apple

I have a M1 MacBook Air which is on 5nm process, and that thing is a beast. It’s practically a stock 8700K in multithreaded performance. And faster than Ryzen 5000 series single thread (Only while running about 3Ghz) so single threaded performance is extremely fast. And it does all of this while running in a non-native format through a passthrough called Rosetta 2. (Which results on a performance loss)

The efficiency cores will more than likely handle constant windows background activity going on. These little resources that are always running, and always going back and forward.

The efficiency cores do a good job managing things like this, and back ground activity. And you’ll just sit there with a solid 0% CPU activity while idling in the OS, even though you can clearly see stuff is happening in the back ground. It’s until you start running legitimate apps or games, you then see your larger cores get utilized.


Now, reading about these new Intel chips sounds familiar. And it sounds similar to
how an M1 architecture operates to me.


I’m not an expert on how architectures operate and stuff like that. Just a thought.