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

Overclockers are a niche market, not a market driver. That means what we want would be the tail wagging the dog, not the other way around.

Also, 25-30% when Intel, under LN2 only gets 40% (literally, 5GHz for the 9900KS which then gets 6.9-7.1GHz under LN2 is only 40%) shows they are not as different as you state.

Now, I stand by my statement boost is killing our hobby. It is. But, I also don't stand for mischaracterizing a product, even if hyperbole to stress a point being made, if that hyperbolic statement may mislead others.

For AMD being closer to 4GHz, 200MHz OC is a 5% increase. For Intel clocked at 5GHz, 250MHz, or a 5.2-5.3GHz OC all core, is the same 5%. Please stop acting like they are all that different. Intel trying to leave less on the table to compete or keep some form of lead has reduced them to giving roughly similar overclocking. On HEDT, they still leave a lot of room due to AVX2 and AVX512 workloads being more common, forcing a downclock, or because long rendering or scientific tasks would eventually raise the temps so high it would kill a chip in a situation, whereas the boost would automatically adjust the frequency on AMD, negating needing an offset for AVX workloads, we are seeing the industry head a certain way. We don't have to like it, but we also don't need to act as though we are self-important or that one company is not reacting to another competitor, reducing that OC headroom.

Now, I do wish they would build in the steps so we could better fine tune the hardware ourselves manually on AMD, because I'm betting me spending hours getting the voltages dialed in would give more performance than their algo. But that is trying to work within the system we have (similar to manually adjusting the curve on nvidias cards since Pascal).

Once again, don't have to like it (I don't because there goes my hobby), but it's something that is understood.

 

Be happy your best overclock being all cores boost at best case. Of course you can overclock Ryzen chips and that clock speed will scale with voltage and cold, but you'll need exotic cooling. Maybe we finally will see LN2 in all homes where you find (OC) enthusiasts who want eak out more than stock boost clocks on all cores Chilled water works bro @Mr. Fox. But I'm not so sure you would be happy with barely +stock boost on all cores

 

Well, that's why OCing is dying. Do you think Intel isn't working on a boost algo now? Nvidia has had boost since Maxwell or maybe it was Kepler (can't remember). Took them the better part of a decade to make it where OC doesn't do much. AMD took like 3 generations. Intel will do it with rocket lake, possibly, then will have it good within a gen or so of that (not good as in it is a good thing, but good as in it functions well).

For Ryzen, a 360, 420, or 480 rad likely is where diminishing returns will happen on an open loop. You should have a good 50-100MHz over CLC users. But hoping for a good bin will matter more. This is why the hobby is dying.

But, even though frequency isn't going to be going up, and signs point to lower frequency with future smaller nodes, IPC will become a bigger battle ground, along with adding in ASICs for specific tasks being included as a chiplet alongside normal core chiplets. And that is where Intel truly is ahead of AMD, having bought FPGA IP, hiring AMD's AI engineers along with others from the industry, etc.

And, yes, for competing on OCing, a water chiller will become the price of entry. Welcome to a brave new world.

 

Boost algo doesn't necessarily means Intel will close the door in same way as AMD. And none know how newer Intels chips will be.

Intel also takes overclocking very seriously. It has a dedicated OC lab where technicians track silicon quality, voltage scaling and long-term stress testing results. All this information is used to develop better processors for regular consumers. The company is very supportive of extreme overclockers, because can showcase these processors in a environment without the normal limits of heat and cooling. We can give them an idea of what internal voltages are scaling and how much is too much, which can help them to improve future products. Tomshardware, by Allen 'Splave' Golibersuch



One thing for sure... 1700X. 2700X and now 3xxxX sucks for overclocking. Or shall we say Ryzen?

Silicon Lottery Overclocks Most AMD Ryzen 9 3950X Samples to 4.1 GHz tomshardware.com | today

Isn't 3950X 4.7GHz on single core?

 

I think @ajc9988 is viewing it like it is. We don't have to like it when it comes to overclocking, but it does seem logical. I'll still be carefully considering cooling solutions for future builds though as any of the boosting is still gonna be related to that, and no doubt I would try manual overclocking vs using any automatic boost algorithms to see if there's any performance gains vs increased power consumption/temperatures...then I'd make my decisions on settings to use. It makes the process less exciting because there's less excess performance on the table for enthusiasts, but I can see why it's becoming like that.

One thing that is exciting though is the 'fact' that NVidia (&AMD) are gonna be using multiple chips on their GPUs in the future - and I hope they create a product line where they go nuts for enthusiasts by incorporating 'many' dies and many many cores that can allow for some massive power consumption & performance if cooled properly.....then I hope they sell this product line at reduced stock clocks to fit within a competent air cooling solution.....but then the enthusiasts (& 3rd party sellers) can take those same cards and slap some custom water cooling on there to allow the silicon to be taken to their maximum limits of silicon stability (like how we know Pascal is often 2050-2100Mhz limited when it comes to the pure ability of the silicon)......so essentially they'd be selling them underclocked so as to just about fit within normal cooling, but then they would be able to thrive under enthusiast cooling.......I'd also be hoping that this massive GPU enthusiast line wouldn't be too expensive as producing products using multiple chips is cheaper than producing one large silicon chip.

 

Funny, you say the company that misleads the 5GHz 3175 chip supports and prefers overclockers. Lmao.

And yes, an algorithm, by definition, means they leave less on the table. They already showed with the 9900KS they reduced average overclocked to 5%, as I already showed. So you think if they are willing to do that now with pegged speeds they won't use the algo to automatically Max performance, which is what their software does for auto OC, which will wind up being moved onto the CPU and firmware? Are you kidding me?

Also, guess what? You have LN2 OCers saying Ryzen is fun to OC. Look at Joe S.

They chase numbers. They don't care which company they come from. They don't always care if it is a lower frequency if you are getting world records.

So this bull on "must love Intel" and "Intel is for OCers" ignores changing industry trends. You want to OC, buy a chiller. That is where this is going.

Not everything is about clocks, and due to IPC differences, a 5.3GHz AMD clock is like an Intel 5.8 clock, depending on the task. Go pour over the world records. AMD has more than they possibly ever had.

Further, guess what? Nvidia with boost uses OCers, along with EVGA. AMD does too. They regularly sample lucky noob (Alva). So why are you acting like Intel is unique in that regard?

Edit: and on silicon lottery, two things. First, that is all core. Second, they test with AVX2 workloads and you darn well know what frequency they sell it at is not the top frequency. Instead that is the frequency and voltage they found stable within their cooling solution. So stop the crap!

 

If you read the link.... Not mine words. The authors. The same one as this link https://www.tomshardware.com/news/intel-core-i9-9900ks-world-records

The new and shiny 8 cores from AMD. Without the IPC gains from newest arch this would look ugly https://hwbot.org/submission/4212447_hardware_numb3rs_cinebench___r15_ryzen_7_3800x_2907_cb

Of corse I know you can eak out more from SL chips with better cooling

A few results can be find in my link above. Not that I say this is SL chips. But you can see the direction

 

Let's put this in perspective: less than 10% of people that buy K series chips overclock them. Period. Then, less than 1% of those do LN2. So you are literally pointing to only one LN2 OCer and acting like it represents the market. You also are ignoring the fact that most on AMD are getting 5-10% on OC whereas Intel's new chips have 5% for the average person OCing the 9900KS.

If you care to be oblivious that most don't have chillers or cascade or LN2, that's on you. If you can't recognize how few people care, so be it. All Intel has at this moment is OEM and system integrators. And when Intel moves to chiplets, there goes the benefit of ring bus, but they will use active interposers which is better than MCM.

I'm done trying to explain it to you. Look at DIY sales numbers from different retailers. Intel is literally selling only around 20%. Yet they still can't produce enough chips, generally, and are resurrecting 22nm Haswell now. This is causing OEMs to start looking into AMD, which simply underestimated demand. Intel's shortage is expected to H2 2020. AMD should have it alleviated before then and have Zen 3 by then. Comet lake is still a skylake rehash due in April, while rocket lake is supposedly a port of Willow cove to 14nm++++.

And AMD announced Zen 4 is on track for 2021, TSMC says 5nm is already over 50% yields with 1.8X density over 7nm and 3nm is on track for 2022 volume production.

Where's Intel's 7nm news and what yields are they getting? Intel's 7nm is about equal to TSMC 5nm. Are they at 50% already and ready for volume in 2020 like TSMC? Doubtful considering they won't even tell yields in 10nm, which was at 12% a year ago and they say they have improvements on it, but won't give yields data, so likely high teens to twenties on yields.

You are so blinded with your arguments, you are missing that Intel is stagnant on process. Look at the speed of ice lake, which uses more power than even their 14nm comet lake parts, which on desktop would likely have frequencies roughly in line with AMD, which that low of frequency eats up all IPC improvement.

So enjoy what you are saying now, because nothing I've seen says this will last.

 

It is totally evident that we agree on more than we disagree. I'm just not liking the direction that things might be turning (ultimately remains to be seen) and if I can't have things the way I like them I will simply find something else to enjoy. It's all voluntary and all about doing something you find enjoyable. It's difficult for me to get excited about overclocking using new tech in a scenario where the cost/effort to benefit ratio sucks. It's really more about how they have designed things and decisions that are being made by people that are occupying positions that afford them the autonomy to make those decisions based on what they deem is the best approach.

I handle things like this the same way I handle politics. I exercise my right to vote no with my wallet the same way I use a ballot. If my vote doesn't produce an outcome I want it to, I am still under no obligation to embrace the result. I am free to reject the outcome and continue to stand in opposition to it, and there is a sense of satisfaction in doing so. And, doing that deprives those who are being resisted (or rejecting the products they are offering) some measure of success and satisfaction in the process. The continued resistance inflicts a degree of harm financially and/or psychologically. So, it's still a partial win in that respect. They can say it doesn't matter, but to an extent it actually does matter because they are not enjoying the same level of success they would have if everyone was drinking their Kool-Aid. You know it irks them whether they will admit it or not. Nobody enjoys being viewed as inferior or repulsive, even if it is on a small enough scale that they are able to survive or thrive in spite of it.

This is not a small margin though. Not sure I can be content with the concept of needing LN2 to hit 30-40%.

https://hwbot.org/submission/4157385_mr._fox_3dmark11___performance_geforce_rtx_2080_ti_45041_marks

 

i guess that depends on what your baseline is. if base clock then sure, its +90%. but if its stock all core boost at 3.6 ghz for the 7960x, then ure at +47%.

Sent from my Huawei Mate 20 X EVR-AL00 using Tapatalk

 

Never said I'm happy about this. I've been coming to terms with losing my hobby for awhile now. It sucks! And if I'm right on Intel with boost, with Willow cove on rocket lake getting around 18-26% IPC (estimate based on sunny cove cores giving 18% IPC on skylake and Intel saying Willow does not focus on ST performance, meaning likely a lower IPC gain), being built on 14nm++++ (so high clocks, but lower core count than comet lake), but competing with Zen 3 with 10-17% more IPC (current rumor is 10% integer, up to 50% floating point, but only 17% in mixed workloads) or Zen 4, Intel may have an FX moment for one year. That scares me because no company should deal with that! (It should be equal with Zen 3, but lower than Zen 4, so depends on when rocket lake would get released versus AMDs release schedule).

In fact, in having to come to terms with deciding to buy something for my use cases instead of overpowering them because of oc dying and boost.

Intel may never get those high speeds again and future shrinks may take the speed lower. TSMC barely figured out how to get the speed they did for Zen 2 and ice lake speeds SUCK right now and are 1GHz lower than 14nm.

I will still tune my hardware, but competing might dwindle for me just because there won't be much for me to so until I go exotic. I was hoping Zen 2 would drop in and be like a last hoorah before things get so optimized there was nothing to do (AMD killed that hope by requiring a new MB).

I do think beyond 2022 Intel will become more competitive, just hope they solve their process issues, otherwise we go down to two major fabs: Samsung and TSMC. And then that means less variety.

And I do apologize if slightly prickish about breaking this news. I also am losing my hobby and it is rough. If I saw the industry headed a different way, you know I'd say so. So I do apologize if coming off as crass here. Consider it grieving.

 

No, I know you're not pleased with this outcome and I did not think it was prickish on your part. This is deflating and discouraging to everyone. We're both pragmatic and opinionated even though it might surface from somewhat different angles.

AMD has made amazing progress that surprised everyone beyond anything imagined. I think everyone would love to see the results enhanced by more elevated clock speeds, and it would be a joy to see the rivalry with Intel continue to escalate to an adrenaline-overdosed frenzy that carries both companies to heights never imagined.

It's almost like watching a boxing match between two competent prize fighters; each with a strong fan base, one being an underdog expected to go down with a TKO making a strong comeback surge in the 10th round. Both are landing some bone-crushing body blows. Both are bleeding and badly bruised, eyelids are swollen... but the battering continues into an 11th round, then a 12th round. The pounding seems to be escalating. The adrenaline is flowing freely and the shameless exhibition of testosterone is in overdrive. The crowd goes wild. Then, suddenly, both just stop trading blows and go sit down in their corners. They're still giving each other dirty looks and making obscene gestures. They think that what just happened should be good enough to keep the fans happy. When it's not good enough. it's the fault of the fans that paid for tickets and sat through 12 rounds up to that point for having unrealistic expectations.

 

3970X benching coming soon. Build is in process... many of the water looping parts are in route so I have to wait, plus the Holidays. Busy busy...

Very interesting read and I get both sides.

When it comes to competition among the mainstream brands that are being discussed, there are a lot of things to be considered, which makes it hard to compare apples to apples whether it involves positive things about the company or negative things. ie....

So looking at things neutrally from outside of the box;

Intel has been flying solo in many of the sectors for years and years, dominating and eating its cake LOL'ing all the way to the bank... so the big question is even with AMD's success right now in the respected sectors will that be enough to make up for all the years lost?

Is AMD's success that we're witnessing right now enough to catch them up?

In the consumer PC realm, it has been a Blue world for a very long time, all the while AMD has been hanging on with workstations, all its contracts with game consoles etc...

Some will argue that it's all about the nm process. But is it really? Intel with milking the 14nm process to the +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++'th degree already proved that newer doesn't always mean that the current or even the previous nm process is dead.

What about Moore's Law? We're already hearing news about 3nm process... how far (small) can it or will it go, before running into situations that can not be overcome?

I get that we've gone from computers being the size of a large room to now being a small device in your hand, but at some point that relativity will come to an end unless some quantum miracle happens that has yet to be discovered.

My 2 cents on the whole Intel and AMD thing:

Generally, because Intel has always had a much wider range of scalability than AMD, it made it more fun for enthusiasts and XOC'ers.

Overclocking has grown to be an expectation to some degree due to Intel allowing that to happen. This isn't to say that AMD's can't overclock, which as we know is a major misconception, however, it is true that AMD isn't the one that transformed overclocking into the culture that we see today.

Intel right now has a lot to prove, because it has a much longer, "credit history" on the table with a more established history left open to be scrutinized.

Whereas, AMD was always just getting by trying to make ends meet so it never was in the spot light, nor did they achieve even the ability to have the scalablity that intel does, hence why AMD isn't facing the same issues due to the expectation being absent compared to intel.

Ex. Let's look at the 9900K vs the 3900K at the ~$500 price point.

Intel 9900K - 8 Cores / 16 Threads / 14nm / Base: 3.6GHz / Max Turbo Frequency: 5.0GHz

AMD 3900X - 12 Cores / 24 Threads / 7nm / Base: 3.8GHz / Max Boost Clock: 4.6GHz

Aside to the obvious differences with the core count etc... at this price point when looking at purely the ability to overclock, the 9900K is the better overclocker. Most would agree that it's also the more interesting one due to the ability to OC ALL cores that scales much farther than the 3900x.

Note: Let's make something clear... real enthusiasts* generally don't look at a 1 or 2 core OC at 5.0GHz for example, being a true overclock*. In my example here, a true overclock* is an OC on ALL Cores at 5.0GHz and most would agree... The difference between an all core OC vs a single core OC is far to vast to even consider them to be of the same.

Therefore, between the two chips above there's a reason why the 9900K is still the king of $500 gaming CPU despite having fewer cores. For now most games are still single-quad core optimized and although that will change in the future......

....unless AMD can achieve higher frequencies, they'll continue chasing Intel in the frequency war, while winning at the core count war.

When games/devs do start utilizing multi core and assuming that things remain the same, it'll be interesting to see how well the 9900K @ 5.0GHz/5.3GHz on ALL 8-cores handle the games vs the 12-core 3900x with a single core @ 4.6GHz ...

So the question becomes, what CPU is more fun for an overclocker? Definitely not the AMD. (IMO)

All the while, with AMD mopping the floor with their 3000 series CPU's, it's absolutely not surprising at all, because most people / gamers are just regular users who don't even know what temp their system is running at.

For me personally, I find way more joy in my 9900K/KF/KS than the 3900x, which honestly was a very boring chip. To be fair, I went into it already not expecting much from the 3900x for the reasons that I've mentioned above. That OC expectation just wasn't naturally there, because it really never existed in my books with AMD chips. (Scalability)

When it comes to work station CPU's I still really like the 7980XE for what it can do and it'll always be a classic in my library, but even with the new AMD Threadripper CPU's, it's the core count that excites me over its boost frequency. Relatively speaking the same factors apply here between let's say the 7980XE vs the 3960x TR.

7980XE 18-Cores all at 46x-53x -OR- a few cores at 4.7GHz with the 24-Core 3960x

For those who actually utilize the cores, they will still favor the 3960x with having 6 more cores, despite them being at a lower frequency. This is why retailers can't even keep the new Threadrippers on the shelf. It's also the reason why I'm in the process of a 3970x build.

Didn't mean to write a book, but with all that being said, Intel isn't dead necessarily, it's just that they were caught off guard (being lazy really) and now the clock is ticking for them to answer back. The loss for Intel with AMD's current success can potentially be so great that it'd make all the years Intel had flying solo, less and less irrelevant. The gap is closing....and for the benchers here we'll go with the whoever has a better overclocker to get points in that category / class.

 

Great post! I totally agree with just about all of it. We are living in some interesting days.

 

I've been using older CPUs even today.

Intel Core i7 3920XM @ 4.7Ghz ~31% OC
Intel Xeon E5-1660 v3 @ 4.4Ghz (for the moment) ~33% OC
Intel Core i7 4800MQ @ 3.9Ghz (not unlocked) ~12% [my 4930MX was around ~25%]

Only exception is the Intel Core i9-9900K @ 5.3Ghz (4.7Ghz daily for now) ~13%

They all clocked decently (~30% all core for Ivy/Haswell) and ~14% for newer CPUs

Some older CPUs I had recently

i7 6700K @ 4.6Ghz ~15%
i7 7700K @ 5.0Ghz ~ 14%
5820k @ 4.5Ghz ~ 32%

To be fair 6700K, 7700K, 9900K were in laptops but the overall trend does seem to point out less OC over the years. Newer CPUs are already clocked much higher than before.

All mine are on air though so I'm sure people have pushed all these CPUs even further.

I'm worried what my next computer will be able to OC.

I've seen the same with GPUs.

My 400 series GPU clocked very high and with voltage.

600 series was less
900 even less
1000 was locked on laptops and 2000 has 2.1Ghz clockblock so yeah

 

Thanks Brother and yes we are indeed. Let's hope that all the competition right now will work in our favor for Blue and Red to bring us something even better than what we're seeing now. I think after AMD launches their 64-Core 3990X, the core battle will mellow out and hopefully they'll focus on bringing better version chips in the sub $1000 line. In the meantime, it'll be even more interesting on how Intel fights back. Either way it'll be a win-win situation for all consumers and perhaps very favorable for us enthusiasts if Intel's keeps getting forced to drop prices.

 

You miss the plot by focusing on frequency without looking at IPC. IPC*frequency=IPS (instructions per second), which is the true metric of performance. For example:

1*5.0=5.0
1.1*4.4=4.84

In this example, you have a CPU with 5GHz compared to a CPU with a 10% IPC advantage running 4.4GHz. The 5GHz CPU is still the victor in this scenario. As you approach 1.2, you get 5.28 on IPS, making the slower CPU faster. This is what I refer to as the "FX scenario," referencing how the FX CPUs could hit higher frequencies, but it didn't matter because their IPC sucked, so Intel's slower CPUs were a better buy.

As I said, who cares so long as the points are there.

That is also why I gave the analysis of rocket lake getting a large IPC boost while having the frequency advantage. Zen 2 does have a solid lead on IPC, but the frequency deficit makes them only about equal with Intel's 14nm designs. Intel getting the IPC boost from a back port, while AMD gets an IPC boost on Zen 3 makes that an interesting match up. But Intel is releasing cascade around April, meaning rocket may or may not come out in 2020. If it doesn't, then it is coming out against Zen 4, which will carry another large IPC boost. If that is the case, then Intel's frequency and IPC may not be able to overcome the IPC and lower frequency of Zen 4, creating an "FX scenario."

This isn't necessarily about process and smaller being better. Intel being forced to optimize 14nm and their results from doing so are simply amazing. But, there does come a point where being on too old of a node is a problem. Would you buy a 32nm CPU today and expect it to keep up with a 14nm or 7nm CPU? I doubt it.

I also pointed out previously Nvidia doesn't need the cutting edge note to own the market either.

But there does come an inflection point. At that point, not being able to go smaller matters. Where that is can be different for each node.

Now, you point out there are limits. This, in part for nodes smaller than 5nm Intel and 3nm TSMC, has to do with electron bleed. Many assumed using a cobalt casing would prevent the electromigration, but Intel tried this on their 10nm and the cobalt, in part, was too brittle, compromising the integrity of the transistors. Silicon, overall, is coming to an end, which is why III-V materials and 2D materials are now being considered as replacements or in transitions from.

We also may have to move away from finFETs at that point. Gate-all-Around nanowire/nanosheet transistors and silicon on insulator (like sapphire, which makes me wonder if Intel will go that route, considering IBM/GF/Samsung worked on GAA and have patents for their own processes) seem to be where this is headed.

Controlling cooling due to the die shrinks and heat densities will be an issue moving forward as well.

So it is a bit more complex than your presentation suggests. And yes, more cores depends on having an ability to use those cores by the software. Considering that is now a thing, we are seeing more software do those optimizations, as well as APIs.

So just wanted to put that in context.

 

Not true with that statement. AMD pretty much owned frequency, and performance crowns for around 6 years or so beginning in the late 90's with Athlon, T-bird, and Athlon 64 lines. Even their K6 chip smoked Intels Pentium II. This is when Intel used underhanded tactics to deny AMD market share, which was settled with a lawsuit as we all know. Nearly all enthusiasts where on the AMD platform during that time.

 

No I'm well aware of the plot as I was just emphasizing a general point that catered to the view point of most consumers, because they see things differently than most of us here. I think perhaps you missed the point of what I was trying to say and not seeing this.

Your example about the IPC and the result of IPS is correct and something we've known all along between Intel and AMD. When you break down the technicality of how the wheels turn under the hood as stated above, it only goes to compliment the points that I was trying to make.

Ex. "Zen 2 does have a solid lead on IPC, but the frequency deficit makes them only about equal with Intel's 14nm designs. Intel getting the IPC boost from a back port, while AMD gets an IPC boost on Zen 3 makes that an interesting match up."

I was purposely explaining the real simple reality of the fact that the 8-Core 9900K @ an ALL core 5.0GHz/5.3GHz with less cores overall and other minor handicaps -vs- the 12-core 3900X with a FEW cores at 4.6GHz with handicaps of its own; still makes the 9900K the better gaming chip and IMO the more interesting chip for overclockers.

As for the complexity of what all goes into the smaller nm process and all the questions surrounding that, it has already been heavily discussed in threads about a year ago. I'm not sure if you were in the discussion, but it pretty much talked about all the things that you've listed.

My purpose wasn't to shed light on any of the technical details, rather the goal was to just give a birds eye view of the very simple realities that I think often gets overlooked with people overthinking things too much.

To sum it up, this is why I personally give a lot of attention in respecting the different classes of CPU's, which is why I do run the 9900K OC'ed for my main gaming rig and not the 3900X and also why I'm building a 3970X work station rig and not using my 7980XE.

The simple reality IMO is that when it comes to core count, AMD has that in the bag and with results to prove it, eating Intel for lunch all day. However, when it comes to overclocking, scalability and overall fun factor, Intel still is the go too choice.

Unfortunately, the group of people who understand and enjoy the art of overclocking is very small compared to the majority that can't even tell what the temp on their computer is. - It's not good for us OC'ers/Benchers at all and yes it totally sucks.

I think this is the main point that @Mr. Fox was trying to make in that we're all being doomed due to the complacency and ignorance of the majority of the consumer base, which would naturally give the companies a silver platter to take advantage of that in doing what they want...

Overall, I agree with most of what you're saying and I think we all share the same frustration of where the market is headed in the overclocking realm. The danger here is with Intel in likely following AMD's footsteps in releasing cpu's that run at the upper end of of the chips potential, leaving little left for non XOC overclocking. That would kill the scalability and OC culture that Intel is famous for.

However, the direction of where Intel goes from here could be interesting as their only weapon might be to milk the nm+++++++ process again and offering the very chips like the 9900KS just to get by....

We shall see... until then Happy Holidays & Overclocking...

 

You're going way way back into the days my friend. I used to be all team Red back then. Fun era for sure...

The statement is true if you don't go back that far in time lol. I had thought this would be something that would be obvious due to the conversations right now talking about chips in this era, but apparently not. I'll try to be more specific to save any confusion, but really shouldn't have to.

So to be clear, we're talking about the recent era...

 

I was the one leading that conversation, and now IEDM was dominated with that discussion this year! That means we are much closer to market, which is exciting.

But, you also are not accurately representing the market. The majority buy at the $200 mark for CPUs or lower, not the 9900K/F/C/S level. So although you focus on that, and that is Intel's best selling level currently for DIY segment, it is NOT where consumers are or what they want.

I pointed out that only less than 10% who buy a K series SKU overclock. So we are now in subsection of a subsection. Because of this, people do, in part, ignore some overclocking numbers. Instead, if it can boost and leave less performance on the table, they like it (like MCE or PBO). To them, it is a simple switch and gives what they want.

So even you may be mischaracterizing the market, focused on high end gamers and enthusiasts, not your average consumer. Companies rely on those individuals to have the goodwill rub off to lower skus in the stack. That is where Intel is trailing behind AMD except in older games heavily single threaded. But even that is changing with new optimizations and newer titles.

What we need is a new API that allows the simplicity of optimization of DX11, but where with a switch they can deeper dive into optimizations like DX12 or Vulkan, while also having better use of multicore CPUs.

So it is time for software to catch up.

Also, AMD had a problem with M$ not fixing their antiquated OS in many ways. So they started putting the fixes in hardware to guarantee it works as intended. This tend will continue. So it is interesting times.

It's only 15-20 years. Although that may literally be a lifetime for some forum users here. Lulz.

 

My references to the 9900K vs the 3900x generally still apply to the lower cpu's as well and the reason I used the ~$500 CPU was because it is the pinnacle of both teams popular gaming CPU. There was no intention of mischaracterizing anything as the examples that I was giving were based between the popular 9900K and 3900X at the ~$500 price point. The majority of gamers that opt for the best selling high end cpu opt for those two options.

As Devs further optimize games to utilize more cores beyond the six~eight (depends on the game as some have no difference after six cores w/ HT etc...), I think Intel with their 9900K will still have the lead if the core utilization doubles, jumping to 8. In this scenario, it would still give the victory to the 9900K.

However, if and when it goes beyond utilizing 8 cores, then things can change a bit... Intel at this current time has the ALL core OC on lock (leading/good grip), where AMD lacks despite having typically more cores. --

Then again, even beyond 8 cores for example the 7980XE can OC higher on ALL cores where the AMD 3950x, 3960x and 3970x can not with a much lower oc if on all cores. There are of course advantages that the newer AMD chips have, but how that will translate to games has yet to be seen.

As you know there are a lot of moving parts that are responsible for the overall outcome...ie it's not all about the cores frequency, but it does play a big role in the grand scheme of the over all fps and performance in games. (currently)

With overclocking it's still a lost cause for AMD as they are relying on pure core count, but in some ways it's working to their advantage.. It's two different approaches so it's hard to compare apples to apples based on the "engine displacement." lol

AMD is like the naturally aspirated big V8 -vs- Intel with their Twin-Turbo Charged V6.

They both have torque and can do some major pulling, but on the top end that Twin-Turbo Charged V6 will walk on the V8 (typically).

So for actual users who utilize the core count, this is obviously where it makes sense to go AMD.

For most OC'ers/Benchers, we still favor Intel, but who knows... that can all change depending on if AMD cares enough to focus past the core count, which I think they will be forced to after the 3990X? Hmm ... because even if they were to revisit the high end consumer line CPU like the 3900x, due to it already running near its max ceiling, I wonder how much they could actually improve from an already near maxed CPU....

I really wish Vulkan gets more attention...oh and yes you're right eh.. 15-20 years can literally be within some users entire lifetime lol.

 

So lots of misinformation and misunderstandings here on why Intel has gaming still.

Back during skylake and Kaby, games already scaled to 6 cores, some utilizing SMT/HT, some not, with the rare gem using 8 cores, which was toward the upper end of Intel's core offerings on HEDT when the underlying game engines were created. We are only now moving beyond those game engines. That is why many newer game engines show scaling to 8C/16T today. But until all the older ones, like far cry is based on, get retired, Intel will hold the crown as that is the top current scaling for those engines.

Next, there are three parts to why AMD lags in games: 1) frequency, 2) latency, and 3) memory (related to latency). You focus on frequency, and that is why the single core boost of the higher quality chiplets of the 3900X and 3950X close the gap so well. But, using chiplets introduces latency between the cores off of CCX. If you could keep a quad core game on one CCX, it would help with performance. But, then you get heat build up. This will be helped by Zen 3, which dumps the CCX making the entire core chiplet where each core can talk to each other and have access to the massive 32MB of L3 cache, instead of just the 16MB of L3 on the CCX, then requiring going to the I/O die and back to the other CCX on the same die. You also have latency to go to the I/O die, then memory and round trip. The L3 helps combat this, but it isn't enough to fully overcome the current deficit. Also, with the infinity fabric, it can only run at a Max speed which is lower than what modern ram is capable of. This acts as a bottleneck.

Taken all together, if AMD was using an active interposer (which is expensive), they would already be in the lead on gaming. It is only current inefficiencies in the chiplet design that is allowing Intel that edge in those use cases, and as the industry moves forward, more of those inefficiencies will be removed, while Intel will be adding them as they move to chiplets. So although you have a point, it is much more complicated than you present.

Does that better fill in what point you were making?

 

Intel: We Aren't Stepping Back to 22nm Haswell; We Never Left

By Paul Alcorn
Intel says it never intended to discontinue the 22nm Haswell chip.

For each year I have seen AMD push the release a few months forward vs. last year. Will Zen 4 be on track for 2021 2022? Will they even have the best chips for sale spring 2022? And same for Ryzen 4000... Spring 2021? This years modern tech words... Paper launch Hence push release further and further forward for each gen chips.

Ryzen 4000 and X670 scheduled for late 2020
By Hilbert Hagedoorn on: 12/06/2019 08:40

I swear it, each year it's the same of course. Regardlessly, website Mydrivers.com shares some new information about AMD's Ryzen 4000 CPUs as well as the X670 motherboards.

Read more

 

Again, from a birds eye view without the technicalities the point that I was making is very simple and straight forward. There are no misinformation or misunderstandings. I'm just stating a simple point while you're over stating things, but thank you. Just two different approaches.

Edit: I said 1-4 cores when my mind was stuck on the past lol...corrected.

Infinity fabric wise, that's why I opt for 3600 low CL RAM. 3600/3733 being the sweet spot. Cache wise, AMD shines in that area and we're seeing where it's paying off with the newer cpu's.

I've stated multiple times that there are a lot of moving parts... not once did I say that the technical side of what makes the wheels turn is not complicated lol.

The focus on frequency was to make a point on the examples given, which ultimately results into the overall gaming performance and fps. Simple simple. I get what you're saying with the details, but at the end of the day AMD still loses here in this department.

I think we're really close to breaking the current barrier of core utilization beyond what we have now. Before Destiny 2 launched it was rumored that it would be able to use utilize a lot more cores, but that wasn't true as the statement at the time was confused with optimizing the game engine. We're not too far away to see a break through in the Dev realm with games... exciting times.

 

I thought my explanation was a simplification. Lol. I'm great at hardware, but need to learn software more

 

All good. BTW how were your timings on the G.Skill kit at 4133? Are the sticks forgiving? I assume those are a 3600 kit originally?

I've been tuning RAM lately so it's something that's fresh on my mind.

 

Those are the 4134 G.skill kits with 19-21-21 timings. I have 2 2x8GB kits. I've run them with CL14 3600 on my 1950X and at CL16@4000 on my skylake 6700K or 14-17-17@3733.

 

Great stuff. I have the G.Skill 4266 kit on the Z390 Master and so far have it dialed in to 4266 @ 17-17-17 ... as this is my gaming rig, I'm happy with it at least running at advertised speeds with lower timings.

 

Yeah, I'm sure if I had a different platform I could have had them run faster. But the Z170 with skylake men controller was hard to get 4 sticks running at 4000 on the maximus viii extreme. People were binning MB to get those speeds and I did it on 4 sticks.

Now, I can't wait for DDR5. Whole new challenge with a new memory generation.

 

Impressive for the Z170 board. Yea DDR5 is going to be great and I have a feeling that it'll be pretty strong, based on what we're seeing now with Samsung memory on components. It's pretty mind blowing to think that we'll have ram like at 5GHz that's faster than most CPU frequencies lol. Just wild...

 

Wait for when HBM is integrated on package! 330MT up to 1TB/s type bandwidth, depending on type used and how much.

 

Optimized 4.3Ghz run...gotta break 4k mark



4.4Ghz CB15

 

You've got this bro... Good luck!

 

Here are some preliminary runs with my msi 2080Ti on the stock blower, before I install the water block... not too shabby.

9900KF w/ 2080Ti on air.

Fire Strike: 31175

https://www.3dmark.com/3dm/41930971

Fire Strike Graphics: 40410 on stock blower.

https://www.3dmark.com/3dm/41930968

Fire Strike Extreme: 18495

https://www.3dmark.com/3dm/41930979

Fire Strike Ultra: 9580

https://www.3dmark.com/3dm/41930985

Time Spy: 13927

https://www.3dmark.com/3dm/41930991

I can't wait to get this on water.

 

2130Mhz max clock then on your stock blower, and that's overclocked of course. Water will allow for a higher stable boost clock even at the same overclock, and it might be interesting to know if you can achieve a higher stable overclock offset than you can using just the stock blower. Have you found your max stable overclock on the stock blower then? If so, then once you get it under water will be interesting if you can increase the offset some more and have it still be stable? Will also be interesting to compare your max stabalised core clock during long runs (say 10mins+) of max overclock on blower vs max overclock on water.....that way you're basically showing the difference that water has given you in real gaming scenarios. (The peak frequencies are less important, I think it's important to compare real measured core frequencies after say 10mins+ of running Firestrike Graphics Test 1 on a loop...for example).

 

That's the plan...just showing benching data here, which is different to what I consider stable gaming data. The main point was breaking the 40K graphics score barrier in fire strike on the stock blower lol.

 

I have to retrain myself on TR overclocking. Second attempt, allowing the system to do whatever it feels with the CPU.

https://www.3dmark.com/spy/9692520

 

Bearded hardware is streaming overclocking the 3970X right now!

 

https://hwbot.org/submission/4303975_

Cine20- 14191CB with the 3960X. A bit of a monster.

 

Spent some time dialing in the 7980XE last night paired with the 2080 Super.

The msi 2080Ti goes in tonight for some benching...should be pretty exciting.

Fire Strike: 29559 (7980XE / 2080 Super)

https://www.3dmark.com/3dm/41966883



I'm hoping to get 5.2GHz to pass at least some benches...work in progress..

* I'll validate it later as I didn't have the test bench connected to the internet.

 

Cool, well I'd be interested in seeing the comparisons from you that I talked about if you do decide to gather that information...because I think that will show how much of a benefit you get from putting it under water...but you won't know if you don't do the same testing before & after. I'm interested in the max stabalised core frequency comparison during long 10min+ runs of looped Firestrike Extreme GT1 (during max stable overclock of both configs).....ha I know that's very specific and I'm not actually autistic (not that I'm prejudiced), but I think that's about the harshest and most realistic & controlled representation of long gaming sessions.

 

It's getting there. I'm pretty impressed with with this base model non oc version of the msi 2080Ti that I got for under $1k during BF. Huh...who would have thought eh?

7980XE on water (non chiller, at least not yet) + MSI 2080Ti Stock Blower (AIR)

Fire Strike: 34737

https://www.3dmark.com/fs/21271256

The numbers are all there in respect to the set up.

Physics: 39189 @ 5.1GHz All Core
(This might be the highest physics in its class at 51x? (non XOC) @Papusan you're the search wizard, perhaps you could double check. takk! )

Graphics: 39625 @ 2160MHz Core
(Once it cracks the 40K barrier it'll put the overall Fire Strike score over 35K for sure.)

Next, in goes the 2080Ti FTW3. I was able to reach a PR on the graphics of 41636 with this card w/ the 9900K, however, with the 7980XE it'll take some work as the environment is a whole different playing field with the CPU.

https://www.3dmark.com/fs/19298675

Work in progress...hope everyone is having a great time during the Holidays. I'm trying to fit in as much benching as I can in between life stuff lol.

@Mr. Fox @Rage Set @Raiderman @JoeT44 @Talon @ssj92 @electrosoft @ajc9988

 

The graphics card ain't so important in 3DM FS Physics Still good scores

https://www.3dmark.com/fs/14222780

 

Ah yes 7uly1 really nice guy from S.Korea. I'll message him about that run, but I'm pretty sure he mainly uses a chiller. That 32855 run was likely during his warm up, but amazing physics for 51x. I'll try to break the 40K barrier.

As for the gpu being non related, that's somewhat true, but the 9900K vs a Skylake X chip do bring in different variables to the overall picture in general. It's definitely not 1:1 the same.

Ex. RAM can be another variable that could throw things way off. With the Skylake X you're dealing with mesh and a lot of other balancing acts to try an get things working nicely together.

Thanks for checking +rep!

 

When are you gonna put the 2080ti under water? Don't forget to do that max OC 10mins+ Firestrike Extreme GT1 loop to determine long term high load stabilised core frequency as a means of comparing your air vs water advantage....if you still wanna collect that info.

 

The 3DM physics scores don't scale well with more and more cores. Hence you see differences between mainstream and Hedt.

 

Absolutely. Two very different animals.

By the way so for the chilled class on hwbot this guy ( https://www.3dmark.com/3dm/38934943) has 35669, but with LOD settings mod... which is....

The next guy (2nd on hwbot) https://www.3dmark.com/fs/18048001 , is only at 33696 ...

Hmm am I seeing this right?

Update: The search criteria is acting weird on hwbot lol. Anyhow, glad to have made it in the ranks of the Top 20 in the respected class.



Happy OC'ing!

 

After I finish testing and gathering my data. I don't like using quick disconnects although I do have them as I find them a bit too restrictive, so swapping stuff out isn't much of a luxury with my set up. The 2080Ti FTW3 is my main card in my gaming rig so it's still in there lol.

I'm also having a lot of fun stretching these cards with their stock blower... It's quite impressive and I find it intriguing to get nice numbers in this form factor, before putting the water block on it.

When I game I don't really apply a heavy OC on the GPU as it isn't really worth it, but yea ... I'll run some graphics tests np. Super position is also another nice test along with Unigin Heaven, all of which I use anyway so as far as data goes...there will be a lot...

 

UL Announcing end of support for 3DMark 11, PCMark 7, Powermark and some 3DMark benchmarks
By Hilbert Hagedoorn on: 12/19/2019 02:19 PM

You may have heard that Microsoft will end support for Windows 7 on January 14, 2020. Benchmarks also have a natural lifespan that ends when they no longer provide meaningful results on modern hardware. When old benchmarks are used with new hardware, the results can be skewed or limited in ways that reduce their accuracy and relevance.

From January 14, 2020, they will no longer offer updates or support for 3DMark 11, PCMark 7, Powermark, 3DMark Cloud Gate, and 3DMark Ice Storm benchmarks. These benchmarks, all of which were released between 2011 and 2013, no longer provide useful, comparable results with modern hardware.

After January 14, 2020, these unsupported benchmarks will:

• No longer be sold by UL, Steam, or other app stores.
• No longer receive updates.
• No longer be guaranteed to work with our online services.
• No longer be eligible for customer support.

As you can see, Futurmark UL follow Micro$oft like hand in glove(same date, same target and goal).