No sign of "standard" Threadripper.

They should just can Threadripper and unlock TR Pro.

Now all we need is a March release of a Radeon 6950XT.

windwhirlNo sign of "standard" Threadripper.

There might not be one.
HEDT is kind of dead, even if it hasn't been officially announced.

windwhirlNo sign of "standard" Threadripper.

Higher Margins on the pro part sadly.

Hi,
HEDT is being reinvented

TheLostSwedeThere might not be one.

Yeah, with next socket coming soon it likely doesn't make much sense to push non-PRO TRs. Might as well save the wafers for Zen 4.

TheLostSwedeHEDT is kind of dead, even if it hasn't been officially announced.

Yeah, at least on AMD, since Ryzen 7 and 9 parts can do the job rather fine. There are even PRO variants if you want, e.g. ECC, but don't want to go all in with a TR PRO.

AirisomThey should just can Threadripper and unlock TR Pro.

Considering all the TR 3000 series CPUs have a TDP of 280 W, you'd likely need at least the Intel chiller for OC.

ThrashZoneHi,
HEDT is being reinvented

Yeah, I could agree with that.

Hi,
Yeah just look at 12900k thread count and it's a little bitty chip/ socket than 2066 is for 11980xe :laugh:

what do you mean can't buy? They sell WRX80 boards now and Threadripper pro cpu's. These will be available to buy for consumers, just like the pro 3,000 cpu's are.

Real slap in the face for trx40 buyers. I personally won't forget that AMD lied to me.
Amd/comments/dsy4kw

Nephilim666Real slap in the face for trx40 buyers. I personally won't forget that AMD lied to me.
Amd/comments/dsy4kw

Damn straight! Still using my new(er) 2950X in the original TR X399 Zenith Extreme board, which had the chip moved to a cheaper X399 board. Both systems still in use.

I can't get onboard the desktop chips - I have 3x NVMe drives all connected at full PCI-E rate, along with multiple x16 PCI-E ports which don't become bastardised when in use.

I want a decent successor.

I do wonder if AMD is planning to fold TR into the TR Pro line and just have that as the standard for both public and private businesses and prosumers. They've sort of fallen through on their promise of keeping TR in step with AM4, but TR as a whole still is a pretty massive moneymaker, between the Enterprise level EPYCs and Consumer level Ryzens. As well, TR's development is tied more heavily to that of EPYC than Ryzen which makes it harder to keep in step with Ryzen. And lastly, I doubt AMD is likely to completely neglect the new market they had targeted with Threadrippers, like the animation studio behind Terminator Dark Fate for example. Thus, I assume that AMD has the numbers, and that their numbers probably show that TR Pro is probably the way to go.

Assuming that's the case though, I really do wish AMD would actually add more PCIe lanes to consumer Ryzen. TR was great with the 8 Core option as a means to get in massive amounts of lanes and a relatively high-speed CPU, but AMD has really neglected those that need more lanes than CPU, even if those add-on lanes are just PCIe 3.0 (or 4.0 once PCIe 5.0 arrives).

TechLurkerI do wonder if AMD is planning to fold TR into the TR Pro line and just have that as the standard for both public and private businesses and prosumers. They've sort of fallen through on their promise of keeping TR in step with AM4, but TR as a whole still is a pretty massive moneymaker, between the Enterprise level EPYCs and Consumer level Ryzens. As well, TR's development is tied more heavily to that of EPYC than Ryzen which makes it harder to keep in step with Ryzen. And lastly, I doubt AMD is likely to completely neglect the new market they had targeted with Threadrippers, like the animation studio behind Terminator Dark Fate for example. Thus, I assume that AMD has the numbers, and that their numbers probably show that TR Pro is probably the way to go.

Assuming that's the case though, I really do wish AMD would actually add more PCIe lanes to consumer Ryzen. TR was great with the 8 Core option as a means to get in massive amounts of lanes and a relatively high-speed CPU, but AMD has really neglected those that need more lanes than CPU, even if those add-on lanes are just PCIe 3.0 (or 4.0 once PCIe 5.0 arrives).

TR and Epyc should be merged into 1 Socket.

As one of those unfortunates running a WRX80 and a TR pro 3000 CPU right now, the "OEM-only" is a let down and I hope they release retail packages...

Nephilim666Real slap in the face for trx40 buyers. I personally won't forget that AMD lied to me.
Amd/comments/dsy4kw

I agree, they need more transparency on this and an apology if they truly intend to leave it without a successor.

I understand the changing market and the shortage that could not be anticipated at that time (Nov 7, 2019)... But the lack of transparency is extremely damaging.

People keep forgetting AMD is supply constrained. Threadripper and Epyc are the same device, so for every Threadripper sold, they leave thousands of dollars of potential revenue on the table....
If YOU were AMD which would you prefer to sell?
I have Threadripper myself, and I would LOVE an upgrade path... BUT!... I've been building my own computers for 30 years now, and only ONCE did I upgrade the CPU (and then immediately put together another with the old one for my dad).

In reality the upgrade market is so small, it doesn't matter. It's just a nice selling point for AMD.

When SP5 comes out for Epyc Genoa, it will be physically too big for even for an E-ATX motherboard (unless you want just 2 memory channels and maybe 3 PCI-E slots), so no more Epyc/Threadrippers anyway.

But Genoa has a SMALL socket version too...:)

NigglePeople keep forgetting AMD is supply constrained. Threadripper and Epyc are the same device, so for every Threadripper sold, they leave thousands of dollars of potential revenue on the table....

They're not the same, core-wise. The main issue is binning. EPYC bins for efficiency, and Ryzen bins for performance. Threadripper has been stated to skim top bins destined for Ryzen, as they're intended to clock higher than EPYCs but attempt to be more efficient than Ryzens on a per-CCX level. This can be somewhat seen in the notable efficiency difference between the 5800X and the 5950X, where some reviewers disabled the second CCX on the 5950X, but found it to be a fair bit more efficient than the 5800X. One can say that the 5950X is effectively making use of bins that could have otherwise gone to TR.

The problem is that AMD sort of created a self-inflicted wound, as Ryzens scaled very well into high core count arenas that TR was intended to bridge, and were mostly seen as good enough for the vast majority who only need up to 16c/32t due to licensing or software limitations. And on the EPYC side, there's been an increased demand in lower core but high clocking EPYCs, which is starting to pull from the TR/Ryzen bins, trading efficiency for performance. And with both ends selling well, Ryzen's top end and EPYC in general, there's very little bins left to spare to TR. Now the only major thing TR has going for it is greater core count and more PCIe lanes at the prosumer level, and for the tech savvy; better cooling thanks to larger heatspreader and better optimization, but it's not impossible that AMD could eventually increase PCIe lanes on Ryzen in future revisions, further narrowing the differences.

That said, AMD can't just abandon Threadripper, not when they were going all-in selling it to development studios and businesses, so at some point they're going to need to reset TR to be more in line with EPYC releases, rather than Ryzen, as the other issue with TR is that its development is almost lockstep with EPYC given similar I/O and packaging. Personally, I'm thinking Threadripper might be folded into EPYC as their "Performance" line, available in both cheaper prosumer variants, and enterprise variants, with some microcode differences (prosumer variants don't get enterprise features but are cheaper, and enterprise variants have the enterprise features and markup).

• Something like EPYC Threadripper, ranging from high-clocking 8 to 32 cores, subdivided between Pro and Datacenter/Enterprise variants (which is further subdivided into 1P and 2P variants), and regular EPYC, ranging from high-efficiency 16 to 64 cores, squarely focused on the datacenter/enterprise. Then there would be no more regular prosumer TR; all future prosumer TRs would just be Pro variants with the embedded security co-processor, saving some costs and reducing some fragmentation.