3rd gen Threadripper, big fail and missed opportunity for AMD.

So then it sounds like you do get more from TR3 does it not?

I want more cores AND more performance, more performance is the price and the justification of the time needed to release a new product.
time equals to devaluation, so to mantain the same value you need to add value.

a better product after a year is the minimum requirements to launch a new product and ask the same money.
to ask more money you need to add even more value.

probably we are used bad with Intel acting like a shit in recent years.
 
Last edited:
Same here, 2950X was a well balanced CPU, but


why should I thank AMD for?
they don't acted like good competition, they simply get the intel crown.

intel had their flaship CPU at 2K, now AMD have their mid tier threadripper at 2K with more to come that will easily surpass 3K.

I get it. You want something for nothing. Holy cow.
 
I want more cores AND more performance, more performance is the price and the justification of the time needed to release a new product.
time equals to devaluation, so to mantain the same value you need to add value.

a better product after a year is the minimum requirements to launch a new product and ask the same money.
to ask more money you need to add even more value.

probably we are used bad with Intel acting like a shit in recent years.

TR3 is brand new. No devaluation. New processes tons more cache. So you do want more for the same at best but when factoring in the real
Cost of developing these processors you want more for less. Good luck with your endeavors and your charity case.
 
Loose your hopes, Trx40 will never see ddr5

Way to misunderstand what I was saying. Yes I am not fully versed on DDR5, although thankfully others have supplied better information than mine.

The point was, DDR5 or lack there of is not enough to destroy a socket.

Advantages
- twice the channels in half the dimms, so to run a quad channel socket you need 2. (twice as dense)
- chop and burst length change to16, so it can operate at higher frequencies.
Disadvantages
- higher frequencies are needed to provide equal latency.
- not out

The same stupid argument was made in the past 2 transitions of memory format. Neither had the outcome you are proselytizing.
 
I want more cores AND more performance, more performance is the price and the justification of the time needed to release a new product.
time equals to devaluation, so to mantain the same value you need to add value.

a better product after a year is the minimum requirements to launch a new product and ask the same money.
to ask more money you need to add even more value.

probably we are used bad with Intel acting like a shit in recent years.

Pretty sure troll, or idiot, but probably troll. Regardless, this should be easy. Compare 3rd gen ryzen to second gen, and extrapolate. There's a large increase in IPC and frequency between the two, plus the improved memory access on 3rd gen TR.

Not sure what more you want, you've got improvements across the board. Launch price for 2nd gen 32core TR was $1800, inflation alone would add $40 to the cost vs last year.
 
Intel will come in the forst half of 2020 with a 10nm HEDT that will surpass Threadripper line and use DDR5 and PCIe 5.0. AMD needs to get the money fast.
 
Intel will come in the forst half of 2020 with a 10nm HEDT that will surpass Threadripper line and use DDR5 and PCIe 5.0. AMD needs to get the money fast.
LoL

inside info?
 
Last edited:
Intel will come in the forst half of 2020 with a 10nm HEDT that will surpass Threadripper line and use DDR5 and PCIe 5.0. AMD needs to get the money fast.

Bullshit, next hedt intel will be 10980XE that brings nothing of that.
But it's sure, Trx40 will live very little.

It's a non sense in buying a $900 motherboard now for most enthusiasts. This is the Hollywood only platform not the enthusiasts one.
 
Way to misunderstand what I was saying. Yes I am not fully versed on DDR5, although thankfully others have supplied better information than mine.

The point was, DDR5 or lack there of is not enough to destroy a socket.

Advantages
- twice the channels in half the dimms, so to run a quad channel socket you need 2. (twice as dense)
- chop and burst length change to16, so it can operate at higher frequencies.
Disadvantages
- higher frequencies are needed to provide equal latency.
- not out

The same stupid argument was made in the past 2 transitions of memory format. Neither had the outcome you are proselytizing.

interesting, didn't know that about the channels per dimm with DDR5.
 
Intel will come in the forst half of 2020 with a 10nm HEDT that will surpass Threadripper line and use DDR5 and PCIe 5.0. AMD needs to get the money fast.
mber-for-having-listened-to-it-i-award-you-no-points-and-may-god-have-mercy-on-your-soul%2Fimage.png
 
Obviously not, when ddr5 arrive this socket is dead

I can afford the CPU but I wont. Stupid prices for a platform that will not last more than 2 years.

Ok this is becoming FUD.

AMD
Code:
754 939 940          2003 - 4
S1 AM2  FL  AM2+     2006 - 7
AM3 G34 C32 AM3+     2009 - 11
(dubba dubba)
FM2 AM1              2014
AM4 SP3 TR4          2017 - xx

There are obvious compatibility gaps within the groups.

Most notably: Sorry Bristol Ridge you can't fit in an 128 Mb (mega-bit, 16 megabytes) with the rest of the cool kids.

AMD extends sockets like a Hobo sewing up socks. Really it's 3 years with an obvious AM3 double cast.

intel

Code:
479 775                                   2003 - 4
M 771 P                                   2006 - 7
1366 B G1 1156                            2008 - 9
1156 H 1248 LS 1567 1155 2011 R 988B G2   2010 - 11
1356 1150 649 647 G3                      2012 - 13
(the missing year)
1151 H4 3647                              2015 - 16
2066 R4 R4                                2017 - xx

It's a bit hard to group releases. Intel puts out a lot more sockets and segments. Except for a few minor gaps it's basically 2 years. Yet the latest crowd has extended that metric a few.

One could say that intel is now consolidating things and current sockets will have less turnover. If that is the case and you want a long lasting intel socket, next years socket could be that socket.

I for one would not bet the farm or predict how long any one technology stays on the market.

I also wouldn't throw out the bathwater for a new memory speed. Designers design systems to balance the needs of each component.
 
Threadripper 3rd gen is a missed opportunity for AMD to crash intel.
Performance per dollar is not bad but not that good.

With the upcoming 10980XE 18 cores at $1000, $1400 for the AMD is pretty hard to justify even considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

Consider the fact that this platform will not have a long life this is pretty a bad deal for enthusiasts.

Ddr5 and usb4 will soon obsolete this platform, in two years we will not have any upgrade path even considering that higher end 3rd gen processor will require a different chipset like trx80.

3960X is pretty disappointing, it costs 100% more than 3950X while offering only 50% more cores, and then add the crazy costs of the Trx40 mobos.

This threadripper is clearly a huge opportunity for AMD but its prices and it's upgrade path destroys it in pretty every front.
 
Threadripper 3rd gen is a missed opportunity for AMD to crash intel.
Performance per dollar is not bad but not that good.

With the upcoming 10980XE 18 cores at $1000, $1400 for the AMD is pretty hard to justify even considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

Consider the fact that this platform will not have a long life this is pretty a bad deal for enthusiasts.

Ddr5 and usb4 will soon obsolete this platform, in two years we will not have any upgrade path even considering that higher end 3rd gen processor will require a different chipset like trx80.

3960X is pretty disappointing, it costs 100% more than 3950X while offering only 50% more cores, and then add the crazy costs of the Trx40 mobos.

This threadripper is clearly a huge opportunity for AMD but its prices and it's upgrade path destroys it in pretty every front.

ok zoomer

cost yeah, didnt we go over that a couple times now..
 
Last edited:
... considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

Clearly you have privileged information re: X299X motherboard pricing in order to make a claim like this. Care to share it with the rest of us?
 
Threadripper 3rd gen is a missed opportunity for AMD to crash intel.
Performance per dollar is not bad but not that good.

With the upcoming 10980XE 18 cores at $1000, $1400 for the AMD is pretty hard to justify even considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

They're

Tom's Hardware
The new TRX40 Designare apparently commands $646.83, which is up to 58.7% more expensive than the previous model that retails for $407.62

72 PCIe4 lanes should justify that $100-150 cost alone. There are lower end boards in the $400 range.

Consider the fact that this platform will not have a long life this is pretty a bad deal for enthusiasts.

Ddr5 and usb4 will soon obsolete this platform, in two years we will not have any upgrade path even considering that higher end 3rd gen processor will require a different chipset like trx80.

Rather than just repeat your FUD. Why not argue it? I gave you plenty of information a few posts above. As have others.

3960X is pretty disappointing, it costs 100% more than 3950X while offering only 50% more cores, and then add the crazy costs of the Trx40 mobos.

This threadripper is clearly a huge opportunity for AMD but its prices and it's upgrade path destroys it in pretty every front.

Put the needle on the broken record.

Again wanting something for less than its value. I'm starting to think you really can't afford it and even if you did somehow buy either platform. You would never justify or make use of the extra IO these high end boards have. One of the main reasons to invest in them.

But never mind that DD5 is years away that's a total deal breaker.
 
You all all making me want to do an upgrade to am AMD setup. Still running an overclocked 4930k here and have a XEON E5-1650 v3 (5930k equivalent) CPU/board/RAM waiting to go into my system.

For gaming though I don't really see the need yet.
 
1. Talk of DDR5 as it relates to a dead socket. Intel doesn't have this yet ether why the hell is this even a talking point. Both will need new sockets when both decide to even use it.
2. Comparing 10980XE to 3960X. The $750 3950X beats that part. (just one source: https://www.notebookcheck.net/AMD-R...and-Core-i9-9980XE-in-Geekbench.442122.0.html )

I think it better to compare threadripper to Intel's W-3175X at the cost of $3000

I personally think it's (tr) overpriced too, but I paid too much for my 2080ti, and i'll pay too much for this too. :D
 
You all all making me want to do an upgrade to am AMD setup. Still running an overclocked 4930k here and have a XEON E5-1650 v3 (5930k equivalent) CPU/board/RAM waiting to go into my system.

For gaming though I don't really see the need yet.

I had handed down my x79 setups to my kids. But given how much they game, I could see the power spikes on my tesla panels diagnostics. I know when they be playing lmao cuz the whole house power draw spiked a couple K. Just saying...
 
Threadripper 3rd gen is a missed opportunity for AMD to crash intel.
Performance per dollar is not bad but not that good.

With the upcoming 10980XE 18 cores at $1000, $1400 for the AMD is pretty hard to justify even considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

Consider the fact that this platform will not have a long life this is pretty a bad deal for enthusiasts.

Ddr5 and usb4 will soon obsolete this platform, in two years we will not have any upgrade path even considering that higher end 3rd gen processor will require a different chipset like trx80.

3960X is pretty disappointing, it costs 100% more than 3950X while offering only 50% more cores, and then add the crazy costs of the Trx40 mobos.

This threadripper is clearly a huge opportunity for AMD but its prices and it's upgrade path destroys it in pretty every front.

Your total fixation on TR3's core count for the sake of core counts, while completely disregarding the MASSIVE leap in pretty much every other front (Base clock, boost CLock, L3 Cache, infinity fabric architecture improvements, the list goes on) just means that basically, you, sblantipodi, are an idiot and a tool.

You refuse to acknowledge the absolute massive and overwhelming evidence that your argument is flawed, and you are just. plain. wrong.

This is my last interaction with you. I'm gonna go click the "ignore" link on your profile, and figure out how to turn off notifications on this thread. We can not be friends.
 
I think the only market not being served by this are people who want the extra I/O from an HEDT/workstation platform but for whatever reason don’t want a big processor to go with it. Those three dudes are out of luck.

Yeahhhhhh, I'm one of those three: our FEA work benefits from crazy fast cores with monster memory bandwidth, but doesn't scale cores all that hot, so 8-16 is more than enough. So if I could buy the absolute fastest cores at a lower total count and get the benefits of the platform, I'd be a happier man.

But these processors are still great. This thread is indeed hilarious.
 
Yeahhhhhh, I'm one of those three: our FEA work benefits from crazy fast cores with monster memory bandwidth, but doesn't scale cores all that hot, so 8-16 is more than enough. So if I could buy the absolute fastest cores at a lower total count and get the benefits of the platform, I'd be a happier man.

But these processors are still great. This thread is indeed hilarious.

I've not messed with Ryzen anything. Is it possible to disable some cores and then clock the other ones up more than they could if they were all enabled?
 
I've not messed with Ryzen anything. Is it possible to disable some cores and then clock the other ones up more than they could if they were all enabled?

Not sure if you can granulate it down to the core, but I know you can overclock individual CCX's
 
I've not messed with Ryzen anything. Is it possible to disable some cores and then clock the other ones up more than they could if they were all enabled?

Thanks for asking, maybe someone will be able to answer this, as I don't know! I have a 1600 at home that's treating me well and I haven't done this form of heavy lifting on it.

For FEA, it'd be a work computer and if I cannot configure it through a major vendor, no soup for me (womp womp). We have the software on a solid enthusiast Intel system from a couple years ago (there was some confusion about the spec I really wanted as it bubbled its way through IT/purchasing, namely quad channel memory bus got lost in translation). It's not worth my time fighting it, really.

I'll live in my fantasy of a 4.8+ GHz boost 8-16 core Threadripper for a little while, don't mind me. :) I just thought it was funny that deruberhanyok basically nailed my edge case.
 
OPs thread title is edgy, but I am just thankful he did not post it in Hardware Tech News sub forum. As others have said, it would be really nice if AMD had a 16 core HEDT part. The quad channel ram will allow you to get maximum performance while using cheaper low latency ram at lower speeds.

Platform life is a non-issue. USB 4.0 will not change anyone's life. DDR5 is still about a year away and the first wave will have minimum performance benefits of some of the top tier DDR4 today.
 
I've not messed with Ryzen anything. Is it possible to disable some cores and then clock the other ones up more than they could if they were all enabled?

Yes. No. In that order.

You can disable cores and you can overclock a specific CCX and potentially increase your head room over an all core overclock. There are two things to consider with this. The first is that you don't need to disable cores to do this. Doing so won't net you any gains. The cores don't do anything if they aren't in use. The way the Ryzen CPU's are laid out varies based on whether or not its a one or two chiplet part, but lets take a 3700X as an example. This is a dual chiplet part with a single CCD and one I/O die. The CCD contains two CCX's. The problem is that one CCX will always be better than the other one. On any given Ryzen CPU, there are only one or two cores that are capable of hitting the advertised boost clocks in the first place. All your other cores will fall short of that. You can gain a slight amount of performance overclocking a specific CCX, but the value of this varies based on a few factors.

Typically, you'll have one core that can boost the full amount advertised. This is what we call a single-core overclock. Using the 3700X, this would be a 4.4GHz boost clock. When using a multi-threaded workload we get a typical overclock of 4.0-4.1GHz boost clocks on all cores. Manually, you typically won't see anything over 4.3GHz across the entire CPU. Many retail CPU's don't even do that well. One CCX is generally what holds you back. You can overclock specific CCX's, but not all the cores in a given CCX are equal. Again, only on or two of them can achieve the full boost clock. The others in the good CCX are still likely to drag your overclock down.

On the eight core and lower count CPU's, a manual overclock or raising the clocks of a single CCX could be beneficial. However, on CPU's like the 3900X, you might sacrifice your higher single-core boost clocks which sometimes works against you. Essentially, the value of this type of overclocking is dependent on your applications and where your best core is. In heavily multi-threaded work loads, going for a maximum all core overclock and then taking your best CCX further is certainly worth doing. Unfortunately, on a higher boost clocking CPU like the 3900X, your single-threaded performance will likely suffer. However, CPU's like the 3700X you probably won't have to sacrifice much if anything at all.
 
I suppose I have only a few points to make:

HEDT isn't for everyone. It's expensive. The market position, for consumers, is halo. Relative to the other available gear, it will always be expensive.

"Expensive" is, itself, relative. If you're putting your machine to work, the cost/benefit of specs changes dramatically.

This is a very important point: it's ok to not be a HEDT-level consumer. It's a Ferrari. There are lots of very formidable - and more economic - choices. Don't sweat it, man. Just do you.
 
AMD didn’t want to release a 16 core version. Everyone says they want it but it doesn’t sell. The 1900x and 2920x didn’t sell. That’s why they were massively discounted. People that need/want the extra PCI-E lanes want the extra cores. The pricing makes sense, it’s $100 more than TR2 but with massive improvements. IPC, performance per watt, PCI-4. 20% more PCI lanes, better memory controller, and more L3 cache. The benchmarks aren’t out yet but my guess would be better value for TR3 vs TR2 from a dollar per performance metric. And you’re comparing this to what? A CPU architecture that has the same IPC from 4 generations ago? One can argue that it’s barely got better IPC from Sandy Bridge. Honestly if all you’re doing is comparing cost per core count and not actual performance, it’s ridiculous and not even worth discussing.
 
Threadripper 3rd gen is a missed opportunity for AMD to crash intel.
Performance per dollar is not bad but not that good.

With the upcoming 10980XE 18 cores at $1000, $1400 for the AMD is pretty hard to justify even considering the whopping price of the Trx40 mobos that costs nearly double the price of an intel mobo.

Consider the fact that this platform will not have a long life this is pretty a bad deal for enthusiasts.

Ddr5 and usb4 will soon obsolete this platform, in two years we will not have any upgrade path even considering that higher end 3rd gen processor will require a different chipset like trx80.

3960X is pretty disappointing, it costs 100% more than 3950X while offering only 50% more cores, and then add the crazy costs of the Trx40 mobos.

This threadripper is clearly a huge opportunity for AMD but its prices and it's upgrade path destroys it in pretty every front.

If you want it all, then TR3 is the only option and of course there's going to be a premium for that. AMD selling it much cheaper would just cannibalize their 3950X sales. So you should blame Intel for not releasing a more competitive part.

Also, the high end of the market isn't what gets these companies marketshare. They are crushing Intel in the mainstream sector of this market, there's not much need for them to do it in HEDT.

I don't see why TR2 isn't an option for you btw, is NUMA that detrimental to your workloads? Only the 24+ core TR2 had issues with NUMA in Windows. The 16 core is fine. NUMA can hurt gaming performance somewhat, but then again, so does the mesh architecture Intel uses on their HEDT parts in comparison to ringbus. I imagine the TR2 parts will start getting discounted a lot once TR3 hits.

I'm still on a TR1 1950X and the gaming performance is fine for 4K at least, don't find myself CPU limited very often. Higher refresh rates likely a different story. I don't think I'll upgrade to TR3 simply because everything is still running so well on my current system, it is tempting though.
 
I do support on the side for a few Drafting and Animation firms in the area and if half of the claims made by the AMD partners in their promotional videos are true then the ROI on the TR3 systems are less than 3 months for a some of them.... That is a huge deal and they certainly don't consider it too expensive, a high initial cost yes but they see it as a solid investment in growing their business. We are waiting for 3'rd party benchmarks and reports for sure but they are already budgeting accordingly for the upgrades.
 
  • Like
Reactions: N4CR
like this
I do support on the side for a few Drafting and Animation firms in the area and if half of the claims made by the AMD partners in their promotional videos are true then the ROI on the TR3 systems are less than 3 months for a some of them.... That is a huge deal and they certainly don't consider it too expensive, a high initial cost yes but they see it as a solid investment in growing their business. We are waiting for 3'rd party benchmarks and reports for sure but they are already budgeting accordingly for the upgrades.

Allegedly 3rd gen fixes the Premier Pro problem.
 
  • Like
Reactions: N4CR
like this
Back
Top