Comments for The Next Platform

Comment on TACC Fires Up “Vista” Bridge To Future “Horizon” Supercomputer by HuMo

HuMo — Wed, 04 Sep 2024 20:56:46 +0000

At 16,000 feet, the Alps’ Mont Blanc will remain I’m sure roughly twice as high as Texas’ 8,800 feet Guadalupe Peak, and 20 times Austin’s Bonnell Mountain. But, yes, who could blame TACC’s Stanzione for seeking to build new Horizons that can indeed elevate Austin TX from its Faulty Balconies (or is it Balcones Fault? And if not, whose fault is it?) and into the stratospheric heights of the Venado (13,000 ft, 130 PF/s peak), and Alps (16,000 ft, 353 PF/s peak) of this world!

The most important question to be answered at this time, by this newly installed Vista onto the vertiginous 388 PF/s peak Horizon of upcoming high performance computational oomph, is clearly not “what can those redlegged, bigheaded, high plains gracehoppers do for you?”, nor is it, interestingly enough, “what can you do for those redlegged, bigheaded, high plains gracehoppers?” (obviously), but rather: which local specialty of food is it that must be associated with this here newfangled machinery? It’s swiss cheese for the Alps, and blue corn tortillas for the Venado, but what of the Horizon? “Inquisition minds” … 8^b

Comment on The First AI Benchmarks Pitting AMD Against Nvidia by Calamity Jim

Calamity Jim — Wed, 04 Sep 2024 19:27:44 +0000

In reply to Slim Albert.

Alrighty then, I’ll bite … so, from that Token-to-PeakFP16 ratio:
1) H100-SXM*8 to H200-SXM*8 gives 4.14/2.73 = 1.5x advantage to H200 from 76% more HBM in its saddlebags;
2) H200-SXM*1 to B200-SXM*1 is a 4.78/4.25 = 1.1x advantage for taming the Blackwell bronco (with 28% more HBM), and;
3) MI300X*8 to H100-SXM*8 means a 2.73/2.11 = 1.3x advantage for TensorRT+CUDA vs PyTorch+ROCm, which seams believable at this here outside leg juncture of the barrel racing cloverleaf …
And if that’s a sensible enough hold up for you, I reckon horses ain’t quite left the barn yet, AMD’s got its work cut-out for itself, it’s in that there BBQuda secret sauce, just got to ROCm’it and SOCm’it, real good now!

Comment on The First AI Benchmarks Pitting AMD Against Nvidia by Slim Albert

Slim Albert — Wed, 04 Sep 2024 16:05:51 +0000

Magnificent analysis! The ratio of Server-Tokens-per-second (perf on workload) to peak-FP16-Teraflops (max theoretical perf) is a wonderful tool to use here. As you astutely note, it gives a rating of how well-balanced the architecture is (eg. memory-wise, as in H200 vs H100), and how well the software stack manages to take advantage of that arch (eg. PyTorch+ROCm vs TensorRT+CUDA). Additionally, in Blackwell, the chip (and/or model weights, and/or software) have the possibility of FP4 computation that can boost perf by 2x vs FP8 (possibly 4x vs FP16), and this is not available in either MI300X/325X nor H100/200. The FP4 and FP6 in next year’s MI350X should give it the wherewithal to address that B200 feature.

It’s great to see AMD finally coming out of the MLPerf closet with results for the MI300 family. Beyond pricing for an AI perf similar to the H100/200 competition, I think that their value proposition also includes the possibility of running cloud-HPC FP64 workloads at the highest performance, which can help hedge one’s bets, if desired.

Comment on The First AI Benchmarks Pitting AMD Against Nvidia by Patrice

Patrice — Wed, 04 Sep 2024 02:14:13 +0000

I expect the MI325X to surpass the H200 and the B100 (at worst on par?). The B200 having higher wattage consumption will be hard to beat.
Like you said, if AMD is pricing correctly the MI325X (around $30K?) it can be a great $/perf card and fit perfectly in the market.
I expect both AMD and NVIDIA to be supply constraints though, so little relief on $/perf for the H100/H200.
The Silo and ZT acquisition will probably help AMD a lot on the software side, software optimization is one thing that AMD has been doing really well recently if given some time.
Lisa said in a previous interview that she has a lot of supply from the manufacturer; we will see soon if that is true.
Last one, I hope AMD will launch a 1000W MI350X (or watercooled one), otherwise, NVIDIA will own the higher end of this market until Q1 2026 at least.

Comment on Deep Dive Into Intel’s “Ice Lake” Xeon SP Architecture by baldis basics

baldis basics — Tue, 03 Sep 2024 04:23:07 +0000

Great insights on Intel’s Ice Lake architecture! I’m particularly interested in the improvements in AI workloads and how they compare to prior generations. Looking forward to seeing how these enhancements play out in real-world applications!

Comment on So Who Is Building That 100,000 GPU Cluster For xAI? by LP

LP — Mon, 02 Sep 2024 17:28:33 +0000

And today the 100k H100 cluster was supposedly completed…

Comment on Nvidia Rolls Out Blueprints For The Next Wave Of Generative AI by Doctor Zin

Doctor Zin — Fri, 30 Aug 2024 02:41:53 +0000

In reply to HuMo.

Well, that’s one pretty fly interpretation, high as a tethered heavier-than-air rectangular-ish craft IMHO.

I’m partial instead to the more grounded functional analogy between the mechanical flywheel and the electrical inductor, or in-doctor, as we commonly say here in Appalachia. If the doctor’s in, indeed, then he/she must necessarily investigate the development of new drugs, and if NIM can help with that, then so much the better!

Denver’s nice, for a Rockies town, but do beware of mixing the thin air of high altitudes, with the thick air of the magic smoke, as you’ll need to breathe simultaneously faster, and slower. I think it’s easier here in Appalachia, where the grass is green and doctors are pretty!

Comment on IBM Shows Off Next-Gen AI Acceleration, On Chip DPU For Big Iron by Calamity Jim

Calamity Jim — Fri, 30 Aug 2024 01:32:13 +0000

In reply to HuMo. Must be some of that youthful enthusiasm of RedHat perfusing slowly through that there entreprise matrix (or something in the water), like a rejuvinating ray of KC's Sunshine Band ... Oh, shake shake shake, shake shake shake, shake your hardware offerings, oh yeah! You can do it, do it! Oh, don't fight the feeling! (eh-eh-eh!)

Comment on Interview: Post-Earnings Insight With Nvidia CFO Colette Kress by Slim Albert

Slim Albert — Fri, 30 Aug 2024 01:27:23 +0000

In reply to Tom Schmo. Wow! I second that motion ... outstanding interview!

Comment on IBM Shows Off Next-Gen AI Acceleration, On Chip DPU For Big Iron by HuMo

HuMo — Thu, 29 Aug 2024 19:49:35 +0000

Who’d ever have thought mainframes could be so hip!? That z17, zesty and seventeen, my little rock’n’roll CPU, Wo-oh, wo-oh, Telum II I like your style … you go real wild!

Can’t wait for that elusive Spinal Tap Power11 to come a knockin’ as well, the way I’ve been waiting, it sure won’t need to “come back tomorrow night and try again, whoa!”!

And even the Spyre is inspiring, in a fresh, jazz fusion, Spyro Gyra Morning Dance kind of way … Groovy! Gyrate your IBM to the RYTM I say! 8^p