Are you the type of person that is always running low on threads, PCIe lanes and VRAM? Then Chillblast has a system for you! Inside is a watercooled Threadripper 1950X, 64GB of DDR4-2666, two watercooled ASUS GTX 1080 Ti's, a 1TB Samsung 960 PRO M.2 for your systems and a wee little 10TB Seagate Barracuda HDD for storage. Sure it will run you about £7,500, or just a hair over $10,000USD, but if you want to play all the things in 4k you need to spend a bit of cash. Check out Kitguru's full review here.
"The new Threadripper processors from AMD already push the envelope when it comes to multi-threaded performance, and the Chillblast Fusion Centauri Ryzen Threadripper Ultimate takes the red team’s fantastic hardware and pairs it with a stupendous, eye-catching build"
Here are some more Systems articles from around the web:
Systems
- Overclockers UK Cobalt Overclocked Gaming PC Review – 8350K & GTX 1060 @ KitGuru
- iMac Pro review: Hard to upgrade, but holy Jony Ive it’s fast @ Ars Technica
- Axiom TS Mini PC with Raven Ridge @ Modders-Inc
- PCSpecialist Vulcan Pro @ Kitguru
- Budget Overclocking Guide: Ryzen 3 2200G @ TechSpot
- Intel Atom C3950 + Tyan Tempest S3227 @ Phoronix
- ECS Liva Q Pocket Size PC @ Guru of 3D
- ECS LIVA Q @ TechPowerUp
What the Hell, they are
What the Hell, they are charging above RX Vega 64 prices for those GTX 1080Ti’s and memory prices are not that high even at inflated DRAM costs either on DDR4 2666.
But hey look some OEMs, besides Dell/Alineware, have 16 core TR based system now. But did they get their pricing carts from from Intel somehow mixed up with AMD’s.
If they could source some Dual Vega FEs then maybe the potential suckers could mine on that for a while and relieve some of that cost pain. With Nvidia those Shader processor counts don’t favor mining but the ROP counts on those TIs sure favor some good gaming Frame Flinging!
Hey, remember when folks where making fun of Vega 64/56 for having all those excess shader processor counts and a bit higher power usage on gaming workloads. Well AMD is laughng all the way to the bank on that lately.
Crap case
Crap case
So this is a £5000 build with
So this is a £5000 build with ~£750 in watercooling parts so they are charging £1750 for building it and the 5 year warranty.
If it was me I would cool the cpu and gpus with AiOs and put the savings towards a really nice monitor but I guess that people who are spending this much on a computer are not exactly short on money.
I wonder how much it would
I wonder how much it would cost just to go Epyc.
It would be less costly but
It would be less costly but OEM’s never pass on the savings. If you build it youself Epyc is an even better value on a feature per dollar basis if you do a price/benefit analysis of the Epyc/Epyc SP3 motherboard where the price per PCIe lane and price per memory channel comes out less for an Epyc SP3/MB because it offers 128 PCIe lanes and 8 to 16(2P dual socket MB) memory channels.
Unless you game in addition to other usage no one should even consider Threadripper as the Epyc 7401P’s MSRP comes in at only $1075 ad that’s a great deal consedring the 128 PCIe lanes and 8 to 16 memory channels. That MZ31-AR0 (rev. 1.0) Epyc/SP3 single socket MB is the graphics workstation choice with those 4 PCIe 3.0 x16/x16 slots and 1 PCIe x16/x8(electrical) slots and 2 PCIe 3.0 x8/x8 slots.
The third PCIe X16 slot in from the MB’s edge also supports
“AMD’s latest Radeon Instinct™ accelerator cards. Combining AMD CPUs with AMD GPUs allows for synergies and optimization through high-speed network fabric.
For the end user, this brings:
•Lower latency
•Peer to Peer communications” (1)
So That “high-speed network fabric”(1) interface has to be something faster than PCIe 3.0 coming from that slot. And stragely enough there are Professional Graphics filters in Photoshop/other software that can make use of AI for graphics effects so there is a use case for AI and professional graphics workloads for AMD’s AI oriented GPUs as well as graphics GPUs and the Instinct MI25 is also based on the Vega 10 base die tapeout. Why AMD/Gigabyte does not use a workstation branding for the Gigabyte MZ31-AR0 (rev. 1.0) MB is confusing but it’s still the best SP3 MB on the market for features 5 PCIe x16 slots(with one of them being PCIe x8 electrical).
AI filtering is getting good enough that even Green Screen backgrounds(2) may not be not needed so for some workloads that going to be a big money saver.
So that GB motherboard retails for between $625-$680 and up on some online retailers but the MB’s price per PCIe lanes and Price Per memory channel is the same as a Threadripper MB that costs half as much, owing to the fact that the Epyc/Epyc SP3 MB supports 128 PCIe 3.0 lanes and 8 memory Channles. AMD with DRAM being so costly having 16 memory slots means that more of the lower priced lower capacity DRAM DIMMs can be used for even more savings than usual with even non ECC memory costing so much lately.
Those Threadripper MB are not really Vetted/Certified and tested for ECC memory usage and the Epyc SP3 MBs are all ECC Ceritifed and Tested with the ECC support included in the warranty, something the Threadripper MBs do not not mention in their warranties. The Epyc MB’s also get the 5+ years firmware/hardware support from both AMD and the MB makers. And that extended Firmware/Hardware support is not something that happens for the consumer CPU/MB variants that are EOLed sooner than the professional variants.
(1)
“AMD EPYC™ MZ31-AR0 (rev. 1.0)”
http://b2b.gigabyte.com/Server-Motherboard/MZ31-AR0-rev-10#ov
(2)
“Adobe develops AI-driven approach that could end the age of the ‘green screen’ in movies and VR”
https://thestack.com/world/2017/03/15/adobe-develops-ai-driven-approach-that-could-end-the-age-of-the-green-screen-in-movies-and-vr/
It’s all done with GPUs
It’s all done with GPUs and/or Tensor Cores that bothe can make use of the TensorFlow libraries and even Folks using the AI on cell phones and such(1). So AI and Graphics will be a thing. And eventually even for AA and AO/Other things done using AI for graphics. Maybe AI can make Real Time Ray Tracing possible also as the AI can be trained to understand materials and their reflective/refractive properties. Ray Tracing and Shadow/AO using Rays is the most realsitic way but is also computationally intensive and if AI training has done one thing it’s that AI ability to do more with less computation required once the AI training is completed. So that implies better gaming on phones and consoles in adition to diecrete GPUs.
(1)
“YouTube uses AI to replace video background without green screen”
https://www.slashgear.com/youtube-uses-ai-to-replace-video-background-without-green-screen-04521989/