9900K/9700K Overclock performance review

[TechguyMaxC 516]

30 minutes ago, Rob Ainscough said:

Got one on order … as pointed out CO2 would probably be a better option, but dang that's a nice loop.  LN2 tanks are about $4000, didn't see what they are using to recover … where's the compressor and the like?  Cost of LN2 is pretty cheap anywhere from $0.50 to $2.00 gallon … but even in the best containment/recovery setups you lose about 2% per day.

Great idea, now if they could just make it more "for the masses".

Cheers, Rob.

The best you can get at home is a multi-stage phase-change setup, which cost well over $1000.  Many years ago OCZ was working on a product by the name of "Cryo-Z" and had even shown off prototypes but decided to cease development when they ran into logistic issues.  Namely, you can't ship refrigant (e.g. R134) via air, and since manufacturing facilities would be located in Asia and most customers in North America and Europe it would either take a very long time to ship, or they (OCZ) would've needed to create/contract expensive manufacturing facilities in those locations as well.

I know all of the above because I desperately wanted this solution and followed the ordeal closely.  This all happened about a decade ago.

The best you can do nowadays (in a pre-packaged consumer product) is this: http://www.ldcooling.com/shop/ld-pc-v2-115v-usa/192-ld-pc-v2-115v-phase-change-black-xl-suction.html

It's rated for 320W heat dissipation with an evap head temperature of -30C.  CPU temps should be right around 0 then.  Better than ambient, but probably not good for more than an extra 100MHz or 2 nowadays.  

Edited October 25, 2018 by TechguyMaxC

[fghdgdfdfgfgf 111]

2 hours ago, w6kd said:

@FunknNastySo what's special/different in Slamm's VRM config?

Hi Bob-

I’ll post a picture of the digi page when I get back to the house (4to5 hours or wife permitting).

 

[martin-w 3,238]

1 hour ago, TechguyMaxC said:

Novec isn't sub-zero and thus doesn't really factor into this discussion, it just has a low (above-ambient) boiling point.  

 

I'm aware of that Max. I know what Novec is. I'm not bothered if it factors into this discussion or not. 

As I said, I'm wondering how he's doing with his Novec AIO. 

You can continue discussing what you like and ignore me 😉

Edited October 25, 2018 by martin-w

[Guest]

Back onto the 9900K, sounds like they are indeed going to be delayed until 2019 … and the 10nm CPU lineup has been cancelled or put on indefinite hold?  Hmmm … what is really going on at Intel?  Did AMD's 7nm CPU change Intel's "focus"?  Of course Intel immediately denied … so the Truth probably lies somewhere between the two extremes.

I've always maintained that competition is good, lets hope AMD's 7nm CPUs don't go on Pre-Order a year or two before they actually ship.

I must admit, things do seem a bit "desperate" on the CPU front … the performance wall hasn't moved much, small increments.

If it's going to be 2019, then I'll probably just go ahead and chill my 7900X and go wild with voltage.

Cheers, Rob.

[macwino 14]

28 minutes ago, Rob Ainscough said:

Back onto the 9900K, sounds like they are indeed going to be delayed until 2019

Rob, what's the basis for this conclusion?

If so, looks like I'll have to try and return all the parts I've been gathering up for this build.

[jabloomf1230 2,445]

33 minutes ago, Rob Ainscough said:

I must admit, things do seem a bit "desperate" on the CPU front … the performance wall hasn't moved much, small increments.

In my opinion, it's a combination of two factors. First, gaming is what drives the high end of the consumer PC market and most of it is on consoles and phones. Second, on the PC, games are heavily optimized for GPU usage. Multithreading on the CPU is really not a top priority for game developers. Else, we'd be seeing a consumer version of the 28 core Xeon W-3175X.

Edited October 25, 2018 by jabloomf1230

[TechguyMaxC 516]

1 hour ago, Rob Ainscough said:

Back onto the 9900K, sounds like they are indeed going to be delayed until 2019 … and the 10nm CPU lineup has been cancelled or put on indefinite hold?  Hmmm … what is really going on at Intel?  Did AMD's 7nm CPU change Intel's "focus"?  Of course Intel immediately denied … so the Truth probably lies somewhere between the two extremes.

I've always maintained that competition is good, lets hope AMD's 7nm CPUs don't go on Pre-Order a year or two before they actually ship.

I must admit, things do seem a bit "desperate" on the CPU front … the performance wall hasn't moved much, small increments.

If it's going to be 2019, then I'll probably just go ahead and chill my 7900X and go wild with voltage.

Cheers, Rob.

Intel publicly denied those rumors.  I have a bit of history with the author, having worked for him for a short time writing product reviews for his website.  I happen to think he is wrong in this case.  Their public outright denial of the rumor puts them in a tough situation if they're lying as that would expose them to lawsuits from investors so there's that...

[Guest]

21 minutes ago, macwino said:

Rob, what's the basis for this conclusion?

It's not a "conclusion", just information I read … a few articles one from PC World … BUT, I've also read Newegg saying Oct 26th they'll be back in stock and other sources saying Oct 31st.  But when I check my sources, Newegg (called and chat), Amazon, local Silicon Valley retailers and NONE have any ETA because the distributors haven't been provided any ETA either.

Maybe I need to take a drive down to Intel and say "what's up" (they're only an hour away) 😉

Cheers, Rob.

 

[Guest]

19 minutes ago, TechguyMaxC said:

Their public outright denial of the rumor puts them in a tough situation

Not really, they denied they've "cancelled it", they didn't say anything about delaying it again or "shifting strategy" or whatever spin they need to make the investors happy … only that work continues on the 10nm.  It sounds very similar to Intel denial of the failed Larrabee GPU on initial rumors it was cancelled, which they then ultimately did cancel as their product after several denials ... just wasn't performance competitive.  If the 10nm CPUs are harder to make and not really performing any better than 14nm, then it could very well be time to skip a shrink and go to 5nm which was supposed to be out in 2019? 

Going back to some old article about Intel's road map and 10nm was supposed to be in 2015, 7nm in 2017, and 5nm in 2019 … we still haven't seen any 10nm and Intel seem to be having a hard time getting 14nm's out (which were originally scheduled for 2013, 5 years ago).  This presents a solid case of "is it worth it" to continue with 10nm … high costs to fab and very low yields = expensive CPU with very marginal performance gains.

Intel need a re-think, a technology that can provide impressive improvements rather than just chasing the ever more difficult scaling process. Time to look outside the box for other ways to improve performance that isn't just adding more idle cores.

Cheers, Rob.

[TechguyMaxC 516]

Just now, Rob Ainscough said:

Not really, they denied they've "cancelled it", they didn't say anything about delaying it again or "shifting strategy" or whatever spin they need to make the investors happy … only that work continues on the 10nm.  It sounds very similar to Intel denial of the failed Larrabee GPU on initial rumors it was cancelled, which they then ultimately did cancel as their product after several denials ... just wasn't performance competitive.  If the 10nm CPUs are harder to make and not really performing any better than 14nm, then it could very well be time to skip a shrink and go to 5nm which was supposed to be out in 2019? 

Going back to some old article about Intel's road map and 10nm was supposed to be in 2015, 7nm in 2017, and 5nm in 2019 … we still haven't seen any 10nm and Intel seem to be having a hard time getting 14nm's out (which were originally scheduled for 2013, 5 years ago).  This presents a solid case of "is it worth it" to continue with 10nm … high costs to fab and very low yields = expensive CPU with very marginal performance gains.

Intel need a re-think, a technology that can provide impressive improvements rather than just chasing the ever more difficult scaling process. Time to look outside the box for other ways to improve performance that isn't just adding more idle cores.

Cheers, Rob.

This is inaccurate.  

When publicly traded company:

a) publishes timeline for 10nm volume production in quarterly earnings report

b) makes very public statement disavowing rumors about 10nm cancelation and reiterates the timeline outlined in a)

You have a recipe for disaster if they're lying.

Leads me to believe they're not lying.  If you want to believe otherwise, feel free.

 

 

[Guest]

4 minutes ago, TechguyMaxC said:

If you want to believe otherwise, feel free.

Like I said, no conclusions, I believe nothing until I physically have one, until then nothing is "accurate" from my perspective … since I physically don't have a 14nm yet let alone a 10nm that is 5 years late … scales of creditability aren't with Intel.  I do think they have real issues or why else "release" a 14nm CPU and then have very little supply?  Intel aren't providing any information as to why … this is following all the marks of the standard business practice of obfuscation.  If all was "good" at Intel, they'd not be in a position of having to make denials nor releasing products on short supply.  It sure smells like "investor" feeding … but time will tell I guess.

Cheers, Rob.

[HiFlyer 9,363]

6 hours ago, martin-w said:

I'm just wondering How de8auer is doing with his AIO Novec cooler. 

All this craziness..... What ever happened to Peltier coolers?

[jabloomf1230 2,445]

2 hours ago, TechguyMaxC said:

Intel publicly denied those rumors.  I have a bit of history with the author, having worked for him for a short time writing product reviews for his website.  I happen to think he is wrong in this case.  Their public outright denial of the rumor puts them in a tough situation if they're lying as that would expose them to lawsuits from investors so there's that...

I'd tend to agree with Intel since everything I've read is that there is only a minor performance gain going from 10 nm to a 7 nm mode. What's going to be a bigger problem is whether there is enough fab capacity at TSMC to produce enough 7 nm AMD CPUS for PCs. Globalfoundries is not going to be making 7 nm  for AMD and TSMC may be busy with PS5 production during 2019. The new Playstation is rumored to have a 7 nm AMD CPU and if and when Sony green lights production, that will tie up resources.

Edited October 26, 2018 by jabloomf1230

[fghdgdfdfgfgf 111]

My little cpu did this Passmark single core score a few weeks after I put it together earlier this year...

https://imgur.com/UOV31Ye

5.4 - 1.55v - not delidded and cooled by h100 v2.  oh, and the little chip did it all with class and dignity at 92c. 🙂

 

I believe the world mark in the capture is Westman's 8700K running 5.5 .

 

 

Edited October 26, 2018 by FunknNasty

[LRW 42]

4 hours ago, Rob Ainscough said:

Like I said, no conclusions, I believe nothing until I physically have one, until then nothing is "accurate" from my perspective … since I physically don't have a 14nm yet let alone a 10nm that is 5 years late … scales of creditability aren't with Intel.  I do think they have real issues or why else "release" a 14nm CPU and then have very little supply?  Intel aren't providing any information as to why … this is following all the marks of the standard business practice of obfuscation.  If all was "good" at Intel, they'd not be in a position of having to make denials nor releasing products on short supply.  It sure smells like "investor" feeding … but time will tell I guess.

Cheers, Rob.

Nothing specific, but there was this from the Q&A on the Intel Q3 call today (10/25):

John William Pitzer - Credit Suisse Securities (USA) LLC

Good afternoon, Bob, and congratulations on the strong results. Bob, I wonder if you can just comment a little bit on when do you think the supply constraints will be over relative to your capacity addition plans. And if I caught it right, it sounds like in the calendar fourth quarter you're choosing to shortchange the IoT Group. Does that mean that you're already caught up in the PC market, or how should we think about that dynamic?

Robert Holmes Swan - Intel Corp.

First, John, as we mentioned earlier, we were caught off guard a little bit this year by explosive growth well ahead of what our expectations were back in the beginning of the year, and that growth came from all different segments of the business. It put us in the unfortunate situation of constraining some of the demand signals that we were seeing from the market and our customer base.

In conjunction with our customers, our teams have done an outstanding job in the third quarter and we project into the fourth quarter, and that has enabled us to increase our revenue outlook for the year by $1.7 billion. But I think as we go into the fourth quarter, given the demand signals we continue to see across the business, we in fact will be constraining growth.

Our focus has been prioritizing, in conjunction with our customers, Xeon and Core processors. And therefore by definition, the lower end of PC and the IoT business is being constrained. So we are in a constraint scenario into the fourth quarter both at low-end PC and IoT.

As we go into next year and the timing, we've put a lot of capital to work this year. It's a record year for CapEx for us at $15.5 billion. It's $1.5 billion higher than what we expected entering the year, and we have taken some of our 10-nanometer equipment and tools and began to blow that back to meet the increased demand for 14-nanometer. So we're working extremely hard to get back on track in 2019. But at this stage of the game, given the demand signals we see in the fourth quarter, we're going to be constrained a little bit. And we're trying to prioritize best we can with our customers.