IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
In any event, if AMD is confident enough in Zen2's Ryzen to dispense with TR, I think they could do it if its done right, by granting formerly TR level features to Ryzen mainstream...at least on high end boards. For instance, if the rumor is true of X580 will run on PCI-E 4.0 (or even 5.0?) spec and more lanes (more than earlier Ryzen but less than TR? Plus lanes on the chipset?) that will handle both bandwidth and lanes. Now, what I don't know if they can port is the option for quad channel memory (and possibly full ECC support for those who want it). Ideally it would be neat if they enable quad channel for full power, but also have something equating to a dual channel implementation as well. If the highest Ryzen has 16 cores/ 32 threads and high clocks (made higher by OCing) , with the aforementioned upgrades, then maybe they really don't need a separate HEDT platform?
I'll start by saying that I agree with your post above overall, I just didn't want to quote the whole thing to continue the conversation .
You've more or less nailed HEDT as it stands- access to more cores with more PCIe lanes and more memory channels. And the weaknesses are there too as well; relative to 'enthusiast' consumer platforms, minimum memory access latency goes up (which can be mitigated), cooling can be difficult, boards can be 'too big' i.e. EATX, power can be picky, you likely want to populate all memory channels for best performance, etc. To that, AMD adds 'unofficial' ECC support.
Relative to an eight-core consumer desktop CPU, there's a lot of flexibility that's gained by HEDT (plus ECC with AMD), however, those benefits are mostly focused on either specific professional workloads or workloads that will just take slightly less time but still take a long time. When time is money that makes sense, but that's in a rare niche where a user desires both desktop functionality and access to significantly more compute (to include needing massive amounts of memory installed) / storage / networking resources.
But back to the 'doing it right' part.
A hypothetical R9 3000 with 16 cores and the X5xx chipset would bring the same number of cores as a 2950X with higher per-core performance in an accessible form-factor (the size of the board / power supply requirements / cooling ergonomics). The new chipset will not allow direct connection of more PCIe lanes, but it in theory would be like a massive PCIe switch, which even at PCIe 3.0 speeds would be a significant improvement.
For 'power users', this will be the same kind of overkill that they'd get from HEDT. The difference would be largely meaningless outside of benchmarks. You want high speed networking? Cool. You want a fistful of NVMe, perhaps with striping for even more bandwidth? Cool. You want to toss in a few GPUs for whichever application? No worries. Got other stuff that benefits from a direct PCIe link? Roll out!
Where you'd really run into limitations is if using such a CPU for work that would better be suited for a server. Sure, HEDT makes for a great cheap server platform, and with AMD you get ECC (and I'm still a bit miffed you can't get a bare AMD APU with ECC support...), but that's really a very narrow market for a product line. It's very hard to suggest that AMD keep doing this when they can put sixteen cores into a consumer socket that can also take ECC, and with a chipset that can support a significant number of connections natively.