28 - Core Intel Demo Questioned on Cooling

It seems odd that Intel would have been so blindsided by this, they should have had at least an inkling what AMD was going to announce, but this was a pretty funny "I am taking my ball and going home!" response.
 
That has too be the tackiest setup from a "professional" company I've ever seen. It reminds me of back in the day one of the guys at our LAN party tried too tie 2 celeron together with wire then water cool it without proper tube wrapping. It worked and ran as a dual CPU setup in Windows NT but turned out too be a hot wet mess.
 
I will bet someone at Intel will be looking for work later this week. This is not just embarrassing for them, but HUMILIATING! They would have been better off just announcing and showing non-functional demos than trying to show this! Even if they needed external refrigeration, couldn't they have bought something....nice? There are PLENTY of people making compressor based refrigeration chillers for watercooling rigs...if they had anyone on the team with any experience in building chill rigs, they COULD have shown something that, while not mainstream, would have given the impression that there was some actual engineering involved in the creation of their prototype rig.

AMD is going to capitalize on this horrible mistake, and in a BIG way!
https://www.hardocp.com/news/2018/06/06/siri_google_alexa_dont_care_if_you_kill_yourself
 
Could be worse, that could have been a huge Dewar flask with the first commercial LN2 cooled server.
 
They should have gone with a mineral oil cooling tank while they were at it !
 
haha, not to sound as if I am defending intel , but that 28 core must be running on pretty regular cooling in a server set up. They had to overclock the piss out of this thing to achieve the desired cinebench #s. Now I really want the 2nd gen Threadripper. I'm just not sure yet how I will sell it to the wife. Maybe she will not notice....:D
 
Intel has to play catch up to AMD. This happened before. Then Intel came out with the core2duo.
Well.. as far as i remember Core processors were repackaged Pentium M (witch in turn were Pentium 3 mobile parts as Pentium 4's were too hot for mobile), now Intel has Atom in that space so they probably wont pull same trick this time...
 
Well.. as far as i remember Core processors were repackaged Pentium M (witch in turn were Pentium 3 mobile parts as Pentium 4's were too hot for mobile), now Intel has Atom in that space so they probably wont pull same trick this time...

Core is long in the tooth. Now Adam since Silvermount has used OOE execution, and if you look at the average IPC, it's not that far off from Core parts clock for clock. They are incredibly efficient running at 10 Watts max now with 4 cores. But you are right, Atom/Gold Mount +, etc aren't an effective replacement unless they can really scale the MHz.

Intel said a few months back that "We need to take more chances" I think they were referring to "trying out new architectures." instead of sitting on their fat rich tail.
 
Don't take this the wrong way, I'm extremely impressed with AMD's showing and love the competition, but we haven't seen any performance numbers on anything (other than what we are fed) and have no idea what that Frankenstein was other than a disaster of PR... Sure does seem like Intel was working on something but got caught with their pants down and pie on their face - and that's assuming they caught wind of AMD's pretty monumental announcement and rushed to show anything. If not, well we got ourselves a shift finally after over a decade. We all also need be praising Keller right now, this arch seems to be 100% legit like his last venture at AMD (and now he is working at intel so maybe he loves competition too ;) )..., but let's remember all this is at a trade show.
 
I think it's giving Intel too much credit to call this a rush job. Hasn't a 32-core Threadripper been rumored for quite a while now?
 
Maybe this is just the best they could do in that time? It's not super likely that they ever expected having to clock up these high core count cpus. It takes time to go back to the drawing board when you can't expect die shrinks to help solve your various scaling problems.
 
That has too be the tackiest setup from a "professional" company I've ever seen. It reminds me of back in the day one of the guys at our LAN party tried too tie 2 celeron together with wire then water cool it without proper tube wrapping. It worked and ran as a dual CPU setup in Windows NT but turned out too be a hot wet mess.

Back in the day I overclocked my Athlon Tbird 1.4 to 1.73 GHz via several big gulp cups filled with ice, an open case, and a fucking box fan - of course, using the pencil trick on the CPU itself to adjust the multiplier. I'd have to refill the big gulp cups periodically, but hey... a couple extra mhz. Intel's stunt kind of reminds me of that. But I'm loving this new core war, man. Intel and AMD are going balls out. Good for us!

When Intel is going so all out they pull out extreme overclocks and 32 phase VRMs... we know competition is back, baby. They never did this kind of shit when AMD sucked.
 
Wow that was the cooler? Man those pics are fucking priceless Kyle.....What is going on with Intel?
 
Wow that was the cooler? Man those pics are fucking priceless Kyle.....What is going on with Intel?

knee jerk reactions, they did this a year ago too when AMD released cpu's that caught them off guard (i9 re branded xeons). I hope Nvidia is caught with their pants down in the same manner in the near future when it comes to sitting on old GPU tech like they are right now.
 
The cooler they used, according to the pictures over at Tom's, is a Hailea HC-1000, probably B or BH variant, which is rated at 1HP - that's almost 750 watts.

That is insane.

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9EL1kvNzc2ODA2L29yaWdpbmFsLzIwMTgwNjA2XzA4NTYzNC5qcGc=.jpg
 
Imagine if AMD did this. Brought in a 32 core Threadripper engineering sample, slapped a massive sub-ambient cooling system on it, stuck four fans on a massive VRM heatsink, clocked it to 5GHz, got it to post a single Cinebench result, and then said "zOMG we have 5GHz 32 core CPUs, suck it Intel!"

Everybody and their mother would think AMD's marketing moonies had gone completely insane.
 
Imagine if AMD did this. Brought in a 32 core Threadripper engineering sample, slapped a massive sub-ambient cooling system on it, stuck four fans on a massive VRM heatsink, clocked it to 5GHz, got it to post a single Cinebench result, and then said "zOMG we have 5GHz 32 core CPUs, suck it Intel!"

Everybody and their mother would think AMD's marketing moonies had gone completely insane.
Even Raja wouldn't try to pull something like that.
 
Whatever happened to #[H]nanogate? - better management at AMD? ;) (sorry Kyle had to throw it out there)

I could quote my previous post, but trade show, MASSIVE fuck up by Intel who looks to have panicked by finding out AMD had a superior product to present, yet nothing is "proven" from any of this. Let's wait and see some real world benchies guys. This type of rhetoric has bitten some of the fanboi's in the ass claiming the crown before seeing numbers.

(Full disclosure in my previous post I did state AMD looks to have 100% a winner here with the Zen arch in general)
 
does this mean 28core intel is vapourware?

It's a real server part: https://ark.intel.com/products/120496/Intel-Xeon-Platinum-8180-Processor-38_5M-Cache-2_50-GHz

Just being "Presented" as some new future desktop part at 5Ghz. Which is absurd. And look at the recommended retail price for the server part, $10007...

Even a die shrink will not get that to 5Ghz. It's already at 14nm, 2.5Ghz. Die shrinks don't increase clock speeds below 14nm, but do improve the power usage and thermal profile. So maybe a modest clock speed bump if they could get this to 7nm in a few months. But that isn't going to get it to 5Ghz.

That's why everyone has jumped on what they were doing, pointing and laughing. :)
 
Back
Top