ARROW lake overclocking

[ShrimpBrime]

That's why $150, not $400 like today. xD

Prices go up, sense goes down. Like in any other capitalism fiesta.

I remember paying around 680$ US (MSRP 800+$) for a single core cpu, named AMD FX-55 single core. 19 years later, paid 400$ for 14700K. That's 28 threads for several hundred US dollars cheaper.

You get a LOT more bang for your buck in computers than we did 20 years ago. Even the Quality of our manufacturing has VASTLY improved for those dollars spent. And will hopefully continue to do so as much wages go up year after year. I'd expect consumer prices to follow suit.

 

[Macro Device]

You get a LOT more bang for your buck in computers than we did 20 years ago.

I am not THAT young. I remember the prices of 20 years ago. 25, even.

Not my point. I'm talking overclocking prices, not hardware prices. Back in my day... ahem. Well, technically in mine too, OC was purely a skill issue: you could buy whatever CPU, whatever GPU, whatever motherboard and if you know what buttons to press you will get +50 or more percent performance VS stock. Provided it survives.

13 years ago: OC is locked on cheap SKUs (to some extent: was possible to gain +400 MHz if you slotted a 2nd/3rd gen non-K i5/i7 into an overclocking motherboard), you also need a specific motherboard for OC. There's almost no way you OC your stuff more than +45%.
7 years ago: OC is even more so locked; price difference between overclockables and non-overclockables became very apparent. There's totally no way you get more than +20% from OC (except for 8350K but it was so expensive you'd rather buy an i5-8400 so you have better MT performance instead).
Today: OC is dead. Unless you have some extreme cooling system. And even then, gains are single digit %.

 

[phanbuey]

I am not THAT young. I remember the prices of 20 years ago. 25, even.

Not my point. I'm talking overclocking prices, not hardware prices. Back in my day... ahem. Well, technically in mine too, OC was purely a skill issue: you could buy whatever CPU, whatever GPU, whatever motherboard and if you know what buttons to press you will get +50 or more percent performance VS stock. Provided it survives.

13 years ago: OC is locked on cheap SKUs (to some extent: was possible to gain +400 MHz if you slotted a 2nd/3rd gen non-K i5/i7 into an overclocking motherboard), you also need a specific motherboard for OC. There's almost no way you OC your stuff more than +45%.
7 years ago: OC is even more so locked; price difference between overclockables and non-overclockables became very apparent. There's totally no way you get more than +20% from OC (except for 8350K but it was so expensive you'd rather buy an i5-8400 so you have better MT performance instead).
Today: OC is dead. Unless you have some extreme cooling system. And even then, gains are single digit %.

if you buy mid-range or i5 k sku you can still OC pretty well, but you're limited to like +20% performance tops. Not like back in the day when you could buy a Celery and move a few jumpers on the mobo for a 75% bump.

I've seen some leaks of the ARL 5 series OC'd and it looks pretty nuts tbh. Might be worth picking one up just to see if it can clock 6+ghz on air.

 

[Macro Device]

if you buy mid-range or i5 k sku you can still OC pretty well, but you're limited to like +20% performance tops.

How's life in 2017? 8600K was the last i5 known to a man to have successfully OC +20% without using extraordinary cooling.
9600K has a higher baseline, virtually the same ceiling.
10600K, even higher baseline.
11600K? Laughable. Barely runs at stock.
12600K? Achieving more than +15% is really hard.
13600K? Almost maxed out at stock.
14600K? Practically maxed out at stock.
12400 + BCLK OC? Mobos are so expensive and rare it makes zero sense from the $ standpoint. Cheaper to buy a pedestrian mobo and a target performance CPU.
Ryzen CPUs, with a single exception, namely Zen 1 and 1+ series, run pretty much close to their final form at stock. You can't expect more than 7% boost, unless you purchase some extreme cooling (which makes OC financially inefficient).
GPUs, too. Probably the last GPU series you could OC +20% was... Tesla (or R700 if we talk ATi/AMD) most likely. Don't remember. But it's definitely not anything from 2010+. Edit: remembered getting +30% on an HD 7850 and 27% on an HD 7950 but that's still very dated hardware.

 

[Outback Bronze]

I've seen some leaks of the ARL 5 series OC'd and it looks pretty nuts

Got any screenies of this? If ARL clocks well, it may be its saving grace!

 

[iameatingjam]

I remember paying around 680$ US (MSRP 800+$) for a single core cpu, named AMD FX-55 single core. 19 years later, paid 400$ for 14700K. That's 28 threads for several hundred US dollars cheaper.

You get a LOT more bang for your buck in computers than we did 20 years ago. Even the Quality of our manufacturing has VASTLY improved for those dollars spent. And will hopefully continue to do so as much wages go up year after year. I'd expect consumer prices to follow suit.

Yeah. I think people expect.... prices to go down and performance to increase at the rate it once did indefinitely ( not saying BMD inparticular) but it seems to be a common thought process. And its just not true anymore. The technology is maturing and progress will slow, most likely. As it has been with many other technologies over the centuries.

 

[BoggledBeagle]

Honestly I believe that updating CPUs every single year was never needed. It was/is just a waste of resources.

 

[phanbuey]

How's life in 2017? 8600K was the last i5 known to a man to have successfully OC +20% without using extraordinary cooling.
9600K has a higher baseline, virtually the same ceiling.
10600K, even higher baseline.
11600K? Laughable. Barely runs at stock.
12600K? Achieving more than +15% is really hard.
13600K? Almost maxed out at stock.
14600K? Practically maxed out at stock.
12400 + BCLK OC? Mobos are so expensive and rare it makes zero sense from the $ standpoint. Cheaper to buy a pedestrian mobo and a target performance CPU.
Ryzen CPUs, with a single exception, namely Zen 1 and 1+ series, run pretty much close to their final form at stock. You can't expect more than 7% boost, unless you purchase some extreme cooling (which makes OC financially inefficient).
GPUs, too. Probably the last GPU series you could OC +20% was... Tesla (or R700 if we talk ATi/AMD) most likely. Don't remember. But it's definitely not anything from 2010+. Edit: remembered getting +30% on an HD 7850 and 27% on an HD 7950 but that's still very dated hardware.

12600k base clock is 4.4 you could reliably get them to 5.1, many clocked over 5.2 which was 700Mhz ~15%, PLUS ring OC and ram OC will get you that last 5% - my sample here ran at 5.4 daily, which was 22%.
13600 K can easily clock to 5.5-5.7 without melting, from a base of 4.9ghz so another 600-800 mhz plus ring + ram. 14th gen is the same chip.
10600k/8700k needed ram/ring OC but tuned were pretty fast, here it is netting roughly 20% on the lows in game.

Intel i5-10600K Cache Ratio & RAM Overclock Beats 10900K: How Much Memory Matters (youtube.com)

Here is the arrow lake i5 somone pumped and leaked online @ 6.2 Ghz EDIT: looks fake now that I revisit thanks @Outback Bronze for catching that.
Alleged Intel Arrow Lake Desktop CPU Performance Leaks Out, 20% Faster Single-Thread Versus 14900KS (wccftech.com)

 

[Outback Bronze]

Here is the arrow lake i5 somone pumped and leaked online @ 6.2 Ghz:

Sadly, inside that article reads... :Update [7/8/24] - After some internal analysis and comparing numbers with an ES chip, it looks like these numbers might have been faked.

 

[iameatingjam]

12600k base clock is 4.4 you could reliably get them to 5.1, many clocked over 5.2 which was 700Mhz ~15%, PLUS ring OC and ram OC will get you that last 5% - my sample here ran at 5.4 daily, which was 22%.
13600 K can easily clock to 5.5-5.7 without melting, from a base of 4.9ghz so another 600-800 mhz plus ring + ram..
10600k/8700k needed ram/ring OC but tuned were pretty fast, here it is netting roughly 20% on the lows in game.
View attachment 367792
Intel i5-10600K Cache Ratio & RAM Overclock Beats 10900K: How Much Memory Matters (youtube.com)

Here is the arrow lake i5 somone pumped and leaked online @ 6.2 Ghz:
Alleged Intel Arrow Lake Desktop CPU Performance Leaks Out, 20% Faster Single-Thread Versus 14900KS (wccftech.com)
View attachment 367793

I was going to say something similar, as even the TPU 14600k review says 5.7ghz was attainable, though I don't know what kind voltage was required for that. Thats a good a 10%... though probably a little less since we're kind of passed the point of diminishing returns. Nontheless, it is significant.

Though the base clock of the 12600k is 3.7ghz and its turbo is 4.9ghz...?

 

[phanbuey]

I was going to say something similar, as even the TPU 14600k review says 5.7ghz was attainable, though I don't know what kind voltage was required for that. Thats a good a 10%... though probably a little less since we're kind of passed the point of diminishing returns. Nontheless, it is significant.

I honestly think we're at the point where the cores speed itself is less important and the chips are memory/bandwith starved in games. On my 12600K a 1 GHZ boost in core clock and 22% boost in ST netted ~5% net FPS increase... As soon as I touched the ring and ram that number jumped to 15-20%. Add DDR5 to 12/13/14th gen and you would see a 10-15% boost just on that.

Same story with 7700x to 9700x -- the 9700x core on zen 5 is +15% in ST, something games supposedly care about - but nets virtually no gain in FPS. Yeet the timings on any memory kit over 5600mhz and both chips start scaling -- add X3D and theres a massive 25% increase in FPS.

I'm betting ARL is in this boat as well - it will probably OC better than 14th gen, for what thats worth (not much), but I think it's going to come down to that new memory controller and interconnect fabric oc.

 

[ShrimpBrime]

Yeah. I think people expect.... prices to go down and performance to increase at the rate it once did indefinitely ( not saying BMD inparticular) but it seems to be a common thought process. And its just not true anymore. The technology is maturing and progress will slow, most likely. As it has been with many other technologies over the centuries.

Technology for centuries? Plural.

I suppose that would be true if we where only talking about telescope technologies.

Computing technology wasn't a thing 100 years ago yet. I feel we are just beginning. Computing still in its infancy. Like I had mentioned, the X86 core is dated, but only by some decades.

 

[phanbuey]

INTEL ULTRA 7 265K vs RYZEN 7 7800X3D | Test in 6 Games (youtube.com)

almost positive this is BS but if true, those lows look pretty tasty... Hopefully this is the case.

 

[chrcoluk]

Dated tech. The X86 core it's self is dated tech. 2 threads 1 clock cycle. That's powerful and very much efficient way of computing. It costs you no extra die space. (Use AMD for an example, coming in about 1.5 decades late to the SMT game and everyone loves it.)

My 14700K has more E-cores than P-cores. Right, so will the new processors. Disappointingly enough, as I once stated already that I feel there should be more P cores than E cores. Not that we shouldn't have any E-cores at all.....

285K - SIXTEEN Area efficient cores. EIGHT Performance cores. (New!! with no HT!!)
So a different perspective there.

They should now advertise this way ""NEW 285K with Sixteen, that's right 16 E-cores!!! 40% faster than the P-core array you might rather do your harder more demanding work loads with. But no HT now, BT wait 16 area efficient cores!

My i3 14100K completely smokes an i7 1355U out of the water for an example. That's a quad vs a Deca core by the way. The U obviously a mobile part. A very higher end expensive mobile part. That was a 1200 lappy when I purchased it. 14100F like 100$ and it completely spanks the Fk out of a high end business class mobile processor. (this is how I sales builds for people. You have mobile in your pocket, how much more do you need than the Big Little cores in your hand really? This sales pitch is easy, because I have personal numbers to reflect these differences.

So Area efficient cores are totally awesome in a mobile environment, and also great for background and low end tasks. But let me ask how many threads from software that you've installed actually set into a low priority in the system? NEVER. That's part of the issue indeed is accurate control of the operating system and hardware interface. (Why they tell you to update ME firmware AND software). But how many people actually keep up on that kind of PC maintenance? Some bios gain performance, others loose it. Averages.

I think the single core IPC gains are phenomenal actually. I just think the Big Little approach should favor P-cores. Just my humble opinion. Nothing has to be factual cause they are going to be sold. Just not to me personally. I see no gains buying one. I'd rather go 16c32t AMD and call it a day.

Oh yeah there is such a thing as too many e-cores, I wasnt backing up product designs that have ridiculous amount of e-cores, but just saying I think putting extra threads on e-cores is much better than overloading p-cores as HT is basically an overload mechanism that utilises scheduling holes to process the second thread. To me, the ideal CPU is 2-8 e-cores depending on where it sits in the product stack, and 8 p-cores. A higher amount of p-cores with no e-cores I also consider ok like those bartlett chips, its just I am not a huge fan of HT.

 

[chrcoluk]

I remember paying around 680$ US (MSRP 800+$) for a single core cpu, named AMD FX-55 single core. 19 years later, paid 400$ for 14700K. That's 28 threads for several hundred US dollars cheaper.

You get a LOT more bang for your buck in computers than we did 20 years ago. Even the Quality of our manufacturing has VASTLY improved for those dollars spent. And will hopefully continue to do so as much wages go up year after year. I'd expect consumer prices to follow suit.

Different for me.

The first time I spent circa £400 on a CPU was my 13700k.
I remember paying £100 and something for my Pentium MMX chip, and about £50 for my DX4-133.

 

[ShrimpBrime]

Different for me.

The first time I spent circa £400 on a CPU was my 13700k.
I remember paying £100 and something for my Pentium MMX chip, and about £50 for my DX4-133.

Quick research puts the DX4 100 was like a 700$ MSRP. Which many processors where that expensive at launch. Because it was new.

We don't even get "extreme" series processors, which I remember Intel releasing in the 1000$ plus range.

We made vastly less income in the middle 90s.

 

[chrcoluk]

Quick research puts the DX4 100 was like a 700$ MSRP. Which many processors where that expensive at launch. Because it was new.

We don't even get "extreme" series processors, which I remember Intel releasing in the 1000$ plus range.

We made vastly less income in the middle 90s.

I certainly didnt pay anything near that for it, my entire dx2-66 PC was less then that as well and that included a monitor.

 

[ShrimpBrime]

I certainly didnt pay anything near that for it, my entire dx2-66 PC was less then that as well and that included a monitor.

I didnt pay MSRP on a lot of processors either honestly. But my point was/is, wages to cpu cost then and now is not paying more for the same thing as exclaimed by the other poster above. You get more for the same moneys as I example a couple times here already.

 

[phanbuey]

I still remember seeing guys shell out $1k for a pentium 4 extreme a few weeks before core 2 duo came out. Brutal.

 

[Bobaganoosh]

I still remember seeing guys shell out $1k for a pentium 4 extreme a few weeks before core 2 duo came out. Brutal.

"this guy I know" spent way too much money on a QX6700 and then had friends with Q6600's that overclocked almost as fast. I don't "He doesn't" want to talk about the price difference between those two. Never bought an "extreme edition" or KS part again lol. The QX6700 was definitely overclocked to it's max stable settings, but you could easily get a Q6600 to similar clocks for ~1/2 the price

 

[chrcoluk]

I didnt pay MSRP on a lot of processors either honestly. But my point was/is, wages to cpu cost then and now is not paying more for the same thing as exclaimed by the other poster above. You get more for the same moneys as I example a couple times here already.

If I remember right, at the time there was multiple vendors making these dx4 and dx5 chips, amd was one, but there was at least one other as well. I think I had one of these alternative non Intel versions. They were made as much cheaper alternatives to Intel's new Pentium chips.

 

[N/A]

almost positive this is BS but if true, those lows look pretty tasty... Hopefully this is the case.

Arrow lake needs to run with 9000 memory due to moving the memory controller off the chip. 6000 simply doesn't cut it anymore.

 

[alcoholbob]

Jays 2 cents confirmed the e-cores are overclocking monsters, not so much for the P-cores. Chiphell forums indicating their 285K samples can do maybe 200Mhz on P-core, and some of the twitter leaks are saying +/- 100Mhz on P-cores, 300+ MHz on e-cores.

Looks like this will be a generation for getting lots of multicore performance but no so much for gaming for overclockers.

 

[BoggledBeagle]

There should be already some experience with overclocking these?

I tried to add 200 MHz to both types of cores on 265K and it seems to work.

BTW I tried to install XTU, which wanted something else installed, and in spite of that something else appearing to be installed correctly, XTU still did not work.

 

[phanbuey]

Arrow lake needs to run with 9000 memory due to moving the memory controller off the chip. 6000 simply doesn't cut it anymore.

That's so sad... We're pushing 9000 mhz memory to a design so it's at parity with chips running bone stock 5600 memory.