New Threadripper CPUs Start to be mentioned in software

I have a Threadripper 1920X. You will not be disappointed...especially if it includes BAR by hardware.

 

I still want Threadripper "Zen 3" in my old x399

Too bad they went with TRX40

 

TRX80 now with socket sWRX8 (same as TR Pro)

 

trx80 is for epyc/threadripper pro there's like 3 motherboards, this is a product for a 10 people in the world market
BAR is already here for the 3xxx threadripper gen on some motherboards like the MSI TRX40 creator
7C59v164(Beta version)
Release Date
2021-01-11
Description
- Support Re-size BAR function to enhance GPU performance including AMD Radeon RX 6000 series and others.

 

I'm honestly a little surprised why AMD is even bothering when you consider the chip shortages. They can't produce enough Epycs, and they sell for a lot more.

 

Maybe slightly lower bins than epyc? We know that the consumer line ryzen gets mostly bottom of the barrel bins

 

-We are talking about TR 5000, you are talking about previous gen
-Epyc doesn't share it's socket with TR Pro at all (and neither with TR of course) despite they are close derivative from SP3.
-I have writen TRX80 not sWRX80 (normaly the socket will remain)
-for everything else you can look the pro page from AMD itself.

 

Well that's the thing - I think they're actually higher bins, because they clock so much higher.

 

Well, the shortages are going to be resolved soon--I hope!--so I might someday be able to buy a 6800XT online from the AMD store at reference MSRP! Man, sure didn't imagine that in April 2021 I'd still be trying to buy an RX-6800XT! I agree they are shortages--not demand caused--these are all supply-caused, imo. It's really pathetic. I'd love to know what has been eating up their TSMC capacity and why the AMD store in the states isn't bursting at the seams with CPU and GPU product! Here we are literally months past the official availability dates for RX6k, 6.8KXT, and 6.9kXT--and they are harder to buy than hen's teeth! AMD has been completely silent on the issue thus far. The line about "unprecedented GPU demand for $650-$1000 GPUs" just doesn't hold water--those are minimal markets compared to the gigantic $200-$300 discrete GPU markets, traditionally the sweet spot, and the volume markets for discrete GPUs. Fascinating for me to note that nVidia and AMD are both staying out of the ~<$300 GPU markets so far! Actually, I have no objection to nVidia and AMD making some money at their respective MSRPs--none at all. What I object to is the total unavailability the next-gen GPUs from both companies at those MSRPs. With AMD--it's not just the GPUs--it's also their Zen3 CPUs (and even some Zen2 CPUs these days) that are restricted in quantity.

 

They probably oversold their capacity due to PS5/Xbox SX launch. Bad planning on their part to launch those, Zen3 and next-gen GPUs all in the same window. You can easily buy an M1 Macbook TODAY so I don't think its a TSM problem.

 

I have a 1920x but not used for primary gaming rig.
I use it for game streaming to my surface 3 in bed.

 

You mean that processor that is on 5nm not 7nm? That processor that is used for only apple products vs a wide range of products?

Not sure how you can compare the availability of an effectively niche (on a world scale) product in a different fab to availability of zen3 processors.

 

are they ? they might be clocking higher but can they go as power efficient as epyc ? i mean they might be able to do ((random example )) on1.4 v 4ghz ..... but they can not do on 1.15v 3ghz ... not sure thought the same epyc chips might be able to do 4ghz on 1.35v , also i thought a bit about it too i think they just release few pieces so people do not forget that threadripper exists too ...or to pretty much confirm that threadrippers are not canceled all together ...well that's my speculation as always XD

 

Depends on your definition of power efficiency. I look at efficiency in terms of performance-per-clock. Generally speaking, regardless of binning, higher clocks mean worse performance-per-clock.
Epycs are purpose-built to yield a sweet spot of performance, wattage, and reliability. I'm sure you could do as much as a 40% overclock on any random one you pick, if it were permitted. Remember, cooler chips are more stable, so you don't need something pristine to remain efficient. You do, however, need a high-quality chip to achieve high clock speeds under extended workloads.

 

@schmidtbag yeah i see what you are saying , my main idea was that as some chips clock higher with lower voltage .... The same happens when you go on the opposite side of things.but if a chip can do low voltage on "sweet spot frequency ... Does that automatically mean it is a golden sample overclocker too ? I literally do not know .

 

I wouldn't say so, because stable clock speeds, thermals, and voltage do not scale proportionately. Server CPUs are meant to be mass-producible. All CPUs are made with extra voltage headroom in order to retain stability, whether that be because your VRMs are bad or the chip quality isn't perfect. That's why it's usually pretty easy to undervolt processors without really losing anything. However, you've probably noticed that flagship CPUs and GPUs don't tend to have much room to undervolt, because most of the time they're not at a point of being overvolted anymore.

This is actually how Celerons came to be: Intel had CPUs where all the transistors were functional, but not good enough quality to be at their full speed. So, they underclocked the CPUs, sold them for a lower price, and marketed them as efficient models.

I'm not saying that Epycs are low quality, but rather, I don't think it's economically possible for them to all be golden sample bins. Threadrippers, meanwhile, are pushing the limits of the technology and are niche products, so in my eyes it makes sense why they are the best samples.

 

Apple is Niche product? LMAO. If there was a real "semiconductor shortage" then Apple would suffer as well. They aren't. Neither is Intel. In fact, only is AMD with their poor planning. That's why their stock price has stagnated lately. They miscalculated the demand. There is NO GOOD REASON I can't buy a 5900X like 5 months after launch, it is 100% AMD's fault.

 

yeah that make sense especially if you consider that threadripper does not move even remotely the volumes epyc does !

 

There is a semiconductor shortage. Car manufacturers have temporarity shut down factories because they can't finish the cars due to missing chips. Governments have tried to pressure TSMC to do something about it. TSMC hasn't really made any promises because it can't.

Intel has its own fabs, while all the myriad others (AMD, Nvidia, Sony, MS, etc) are trying to buy capacity mainly from TSMC and then Samsung. Apple handled its affairs better. It's not such a giant for no reason. Money talks, no doubt. They could afford it since Apple customers would pay anything to get their next shiny piece.

AMD's fault was to fail at business back in the day and thus be forced to let go of GF. Consequently GF didn't care anymore and dropped out of the nm race. If GF had a fully functional, perfected 7nm process and was finishing its work on 5nm, the current situation would look totally different. The availability of AMD products would be astronomically better.

Somehow you also have forgotten Nvidia. You see a lot of their cards around, huh? It's not only AMD that failed.