Anyone regret getting threadripper after seeing 7980xe leaks?

T

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
9,743
I am curious if any fellow Threadripper owners have any regrets adopting the platform and why, given that the leaks are purporting to show that the I9-7980xe is blistering fast.

I am not claiming to be regretful in anyway. Only starting a public discussion to see what everyone thinks?

I respectfully request that Intel and AMD fanboy's please use regard in your opinions as not everyone will see it as die hard as you concentrate your vitriol towards your opposite sides.
 
I don't trust the leaks especially when looking at the 10core OCing potential. Once we get real reviews then maybe you can revisit this discussion, but for now it is too premature.
 
JustReason said:
I don't trust the leaks especially when looking at the 10core OCing potential. Once we get real reviews then maybe you can revisit this discussion, but for now it is too premature.
Click to expand...
Wasn't the 10 core decent enough? It was just a huge heat load. No problems for a custom loop.
 
KazeoHin said:
Wasn't the 10 core decent enough? It was just a huge heat load. No problems for a custom loop.
Click to expand...
Well 4.4 was hard enough for most. +8 cores does not make it easier, hence the 4.6 leaks are likely not very accurate without chilling, if even that is enough.
 
  • Like
Reactions: N4CR
like this
JustReason said:
I don't trust the leaks especially when looking at the 10core OCing potential. Once we get real reviews then maybe you can revisit this discussion, but for now it is too premature.
Click to expand...

Different die plus better binning for the 18C?
 
I expected Intel would one up AMD when buying my 1950x, but at double the cost its a non starter to me.

Besides that, I think as a workstation platform TR is better due to more pci-e lanes and ECC support. For 2K I'd really want the best of everything, and it doesn't seem like you get that with the Intel - there are still compromises.
 
Lots of suicide benches on the Intel side lately. The Intel chips can clock up but the thermals wont allow them to run at those speeds continuously. You also have to factor in the cost difference as well so I doubt that people that went with threadripper were interested in such a expensive Intel chip. The reviews at [H] should be proof just how hard it is to keep a processor like that cool and stable enough to run benchmarks in a loop for hours. So I doubt people will have regrets about buying a threadripper.
 
juanrga said:
Different die plus better binning for the 18C?
Click to expand...

Not going to change voltage needs and the thermals it will create, also the 10c created VRM overheating. There is no way the 18c will be able to run 24/7 at the same speed as the 10c. If it was coming out on the 10nm process then I would say maybe but no matter how many pluses you slap on to that 14nm node wont change the power draw.
 
I find it telling the base clock is 2.6ghz. i would expect that base to be higher if it had as much thermal and clock headroom as is being suggested right now. i think a solid 3.5-4.0ghz range for all cores might be doable, but i highly doubt you'll find that much more for daily usage, unless you start cutting cores or AVX.
 
  • Like
Reactions: N4CR
like this
I don't own a Threadripper but if I had spent $1000 on a CPU then I would not be sad if a $2000 CPU outperforms it.

I hope but am pessimistic that AMD can get Epyc 7551P systems into reviewers' hands before the Core i9 7980XE launches, so we can have a comparison of $2000 CPUs.
 
Last edited:
It's twice the price man... and its so damn efficient. It idles 30w more than my skylake under the same config which is kind of unexpected.
 
tangoseal said:
Huh? Your statement is confusing.
Click to expand...

Hmm, the Intel route is twice the price so I don't even think of it in the way you're asking. TR is also very power efficient, only drawing 30 more watts at idle versus my skylake setup in the same case/rig. Background, my TR setup is out of the case/rig waiting on board swap, etc. That's a big plus for me.
 
  • Like
Reactions: N4CR
like this
thesmokingman said:
Hmm, the Intel route is twice the price. TR is very power efficient, only drawing 30 more watts at idle versus my skylake setup in the same case/rig. Background, my TR setup is out of the case/rig waiting on board swap, etc.
Click to expand...

Oh ok that makes sense then. I couldnt understand what you were saying earlier.

Honestly my 1950x at stock speeds hits about 140 watts top. No where near tdp. So yeah they're not bad.
 
Gideon said:
Not going to change voltage needs and the thermals it will create, also the 10c created VRM overheating. There is no way the 18c will be able to run 24/7 at the same speed as the 10c. If it was coming out on the 10nm process then I would say maybe but no matter how many pluses you slap on to that 14nm node wont change the power draw.
Click to expand...

Better binning can reduce voltage requirements to get a given OC level. And the larger die can improve thermals via better dissipation (lower thermal density). The VRM overheating was a result of bad designed mobos with stupid fancy plastic covers that worked as thermal insultators. The solution to the problem was so simple as retiring the fancy plastic and giving cooling to the VRM zone. Precisely motherboards engineers did learn from the mistake and the new X399 mobos come with active cooling in the VRM zone

http://www.legitreviews.com/asus-rog-zenith-extreme-x399-motherboard-pictures_196357
 
W
juanrga said:
Better binning can reduce voltage requirements to get a given OC level. And the larger die can improve thermals via better dissipation (lower thermal density). The VRM overheating was a result of bad designed mobos with stupid fancy plastic covers that worked as thermal insultators. The solution to the problem was so simple as retiring the fancy plastic and giving cooling to the VRM zone. Precisely motherboards engineers did learn from the mistake and the new X399 mobos come with active cooling in the VRM zone

http://www.legitreviews.com/asus-rog-zenith-extreme-x399-motherboard-pictures_196357
Click to expand...
Well now bigger isnt always better fpr thermals.

In physics and biology do you remember that the higher the mass an object or animal the higher the thermal inertia is? I would consider physics in supporting a bigger chip is goong to maintain more thermal inertia. Opposite of how you describe it.
 
More than twice the price, two extra cores and how much faster on a glory bench?
This thing must be spewing more heat than even Vega at those numbers, the revenge of Prescott!

Very interested what the latency will be. Probably higher than an 8 core Ryzen CCX-CCX. Thing is, most workloads that latency are an issue on are not scaling anyway, but for anything with more than a few cores, it could be noticeable.
 
x3sphere said:
I expected Intel would one up AMD when buying my 1950x, but at double the cost its a non starter to me.

Besides that, I think as a workstation platform TR is better due to more pci-e lanes and ECC support. For 2K I'd really want the best of everything, and it doesn't seem like you get that with the Intel - there are still compromises.
Click to expand...

It's a shame, if Intel hadn't gimped ECC support and pci-e lanes, it would justify the price. Currently, its expensive with ECC support fused off, just so they could justify creating the 1s xeons, with ECC for more money. And unless it has changed for this generation, Intel doesn't have unlocked multipliers on xeons.

So, TR is still a pretty sweet deal.
 
zerogg said:
It's a shame, if Intel hadn't gimped ECC support and pci-e lanes, it would justify the price. Currently, its expensive with ECC support fused off, just so they could justify creating the 1s xeons, with ECC for more money. And unless it has changed for this generation, Intel doesn't have unlocked multipliers on xeons.

So, TR is still a pretty sweet deal.
Click to expand...

That is a wonderful point and I appreciate it. I didn't think of it that way.
 
tangoseal said:
W

Well now bigger isnt always better fpr thermals.

In physics and biology do you remember that the higher the mass an object or animal the higher the thermal inertia is? I would consider physics in supporting a bigger chip is goong to maintain more thermal inertia. Opposite of how you describe it.
Click to expand...

I think he's referring to thermal conductivity and/or radiation, which are both functions of surface area. Yes, it can be at odds with increased mass, but the biggest aspect there is potentially increased resistance due to thickness rather than some inertia effect which doesn't affect steady-state.
 
Yeah I am sure TR owners are perfectly happy having their mortgage paid the next month.

I am curious to see what the 7920x can do for a 24/7 overclock compared to the 7900x. SKLx really seem to degrade in clocks when adding cores. We will see if the bigger die helps.
 
tangoseal said:
W

Well now bigger isnt always better fpr thermals.

In physics and biology do you remember that the higher the mass an object or animal the higher the thermal inertia is? I would consider physics in supporting a bigger chip is goong to maintain more thermal inertia. Opposite of how you describe it.
Click to expand...

This is not only higher mass. It is higher mass and higher surface, because the die is larger, so the inertia is similar.

Also for the same heat the temperature delta is inferior, because the capacity C is higher

Q = C ∆T
 
juanrga said:
This is not only higher mass. It is higher mass and higher surface, because the die is larger, so the inertia is similar.

Also for the same heat the temperature delta is inferior, because the capacity C is higher

Q = C ∆T
Click to expand...

It will make a tiny difference and not nearly enough to make up for feeding that many cores at high speeds. All cores running at the same speed will be lucky to run 4.2 without overheating at 10 cores it struggles to stay at 4.5 thermally. Will see higher speeds posted but they will just be suicide benchmark runs.
 
Any links on how long the TR socket will last?
As in Zen, Zen 2, Zen 3, etc.
 
Hagrid said:
Any links on how long the TR socket will last?
As in Zen, Zen 2, Zen 3, etc.
Click to expand...
It was mentioned as 2020 for AM4. Being TR is a sort of after thought, not sure if it is the same time frame, but cant see why it wouldn't be.
 
juanrga said:
Better binning can reduce voltage requirements to get a given OC level. And the larger die can improve thermals via better dissipation (lower thermal density). The VRM overheating was a result of bad designed mobos with stupid fancy plastic covers that worked as thermal insultators. The solution to the problem was so simple as retiring the fancy plastic and giving cooling to the VRM zone. Precisely motherboards engineers did learn from the mistake and the new X399 mobos come with active cooling in the VRM zone

http://www.legitreviews.com/asus-rog-zenith-extreme-x399-motherboard-pictures_196357
Click to expand...
Yeah, I so miss the noisy little fans on MBOs from the past times.
 
Shintai said:
Reviews are out and its not faster than 7980XE. ;)
Click to expand...

Looking at reviews it's either neck and neck or the AMD parts take the lead. The lower-core count Intel chips and AMD chips win in the gaming benchmarks. Where the 7980XE does lead it's by 10% or less... for double the cost. As Master_shake_ said, you can literally buy an entire Threadripper platform for the cost of a 7980XE.

These Xtremely Expensive chips simply make no sense. The cost is too high, multi-threaded performance is compromised by the low base clock speeds, and single threaded performance/efficiency is compromised by the suboptimal 'oh shit we need more cores' design and large cache.
 
OMG look at this, the power draw, everyone jump ship now !!

1Power.png
 
Last edited:
Stock for stock and o/c for o/c, the 7980xe and 7960x are neck and neck. The die can only handle so much.

I am really curious on the 7940x.
 
You must log in or register to reply here.
Back
Top