Do you know what the approximate selling price is, thanks.?

So it increases utilization of GPU's? Especially during training part? By better scheduling stuff.

You will soon be covering CES. I look forward to see it.

accelerator accelerator? for accelerating your accelerator?

Nope, Power10 is not the first one with its own logo...

nice, but what will cost this card?

I got so nervous just watching him fling that thing around.

Thanks for introducing us to that product! Attempt at constructive criticism: when showing a product in your hands, maybe try and stay slow with your movement if any is needed at all, and deliberate, highlighting a point you make

Editor's note: I said 1.2 teraflops MXINT8 for a single Corsair package, that is incorrect, I meant 1.2 Petaflops of MXINT8 performance per Corsair package. Sorry for the mistake!

How does power compare to Zen in something like y-cruncher? What does latency mean in this context? I thought epyc had made huge improvements in something to do with memory?

Pretty interesting. But I'm curious about something; Could an OMI slot potentially be used for other types of components, like another processor? This would weaken the flexibility advantage that CXL has over it.

DDR6 will definitely change for servers. CAMM2 will likely be the only way forward for clients

finally someone speaking sense in terms of memory. thanks for this informative interview.

FYI: His name is Bill Starke

Thanks for uploading. Keep them coming! :D

If just portable teleprompters are around and you could bring one to SC... I mean, you look in pain while reading out all those El Cap stats 😂

What's the target architecture for a piece of software to be compiled to run on these things? And what do those chips slot into?

I was about who ask who the hell is "rebellions" glad you explained it

From competitive gambling to the AI bubble. That company sure is on a journey.

Thanks for suggesting "Computer Architecture: A Quantitative approach". I'm reading six edition and it's great. I do have to say that topic that I was trying to understand - Cache Associativity wasn't initially very easy to understand. Took me a couple of days. And now that I finally understood it (I think) Cache Coherency and things like snooping are also a bit "tough".

Thank god for 1.25x playback for all these good KZbinrs and their lisps

🍷🥱

I want to see OpenCL on this thing :P

🧀

Yay Cheese!

KZbin is having a stroke with those video timestamps.

Dr. Ian Cutress casually walking by at 1:32

so very true that it's an underserved market probably the segment that can best provide a start-up with the breathing space it will need to grow just look at the continuing demand for 2017 sparc on 20 nm

This was great. Very concise and informative

Would we want a Cloud500? Where Azure competes with Alibaba and AWS (or any other non A providers) on who can allocate the largest instance to run linpack?

Intel & AMD need more improvement. Intel Price is also too much for this spec 👎 Mac M4 is rocking the field with Les Cost. Can buy a Mac mini M4 at a VGA cost.

When will your review of the 285K be ready?

A lot of it is likely software issues. I have a 10th gen Intel and have a weird issue where sometimes I don't get a video post from my dedicated GPU. I have to plug into the integrated chip (which is disabled) and toggle it on and off again. I chalk it up to an issue with Dell.

Stability was one of the few "wins" Intel still had over AMD. Even though AMD was managing to shorten the time between launch and full stability with each generation.

Ooh, this is like the moment when a mother calls her kid by his full legal name.... Intel just has to be #1. They saw how AMD messed up the Zen5 launch, and had to mess up harder, to be #1 at messed up launches.

I’m waiting for that heat

Hope Intel can fix things by the end of this year and if not, it will be very bad for Intel and for us customers.

Gelsinger destroyed Intel 😢

I don't get how tf Arrowlake pushed off its Mem controller into another tile when they just moved it on same tile in LNL.

I mean, ARL is out of stock anyway. Also LNL has been pushed early and works?

Looks like Zen5 dodged a bullet. While definitely an underwhelming launch with definite issues, it seems to be a perfectly functional product. It's not really much if any better than Zen4 for gaming, though it does seem better in some workloads especially under Linux. The thing is that it doesn't have to be a huge step forward if it's competing with something that's not even as good as it's predecessor and isn't reliable either.

Many People still dont understand currently trying to compete with likes of TSMC and get more patnership to buy their nodes they building new fab designing new nodes so their target is not only for 1-2year but long term and i hope they succed because currently TSMC is market leader used by AMD,Nvidia,Apple etc, also if intel just went AMD route and outsourcing semiconductor from beginning they will do just fine, but because they stuck in 14nm for so long it kinda bite them because their funding issues they fumbling down along the way. They clearly sacrifice their end customer with the arrow lake release by rush then also planning with only just one gen for the 1851 because they just to appease investor and clearly they didnt want relies too much in TSMC itself.

I hope at least this approach drives the prices down, especially of the motherboards. Prices of Z890 Thunderbolt 5 boards dropped considerably after launch in my country. They con iron out the quirks after I got my system; please not earlier.

Appreciate the thoughts. Looking forward to Chips and Cheese deep dive into Arrow Lake when you guys get your hands on it

i want to know who is getting fired for these not ready to launch rush to market products???

Bro is that Noodler's Ink on your hand?

This video shall go down in history as the great 'Intel's Smurf Hand's Truth Bomb.'

I don't understand the conclusion that this is Intel's fault. I imagine the fixes will arrive as BIOS updates, placing the blame there.

Chips and Cheese not recieving a review sample is all you need to know about the current state of affairs.

It's really tough to give any guarantees on completely new design. When you incrementing you can include new features in a design as you verify them, otherwise you need to reserve more time just to get sure. Intel doesn't have that time.