AMD Zen 3 Launch Event Discussion

So my 2950X is now trash.

My renders are getting more complicated and taking literal DAYS on my system, and upgrading to a 16-24 core chip using Zen3 will greatly improve render times...

The only issue is that I have 128GB of RAM on 8 sticks. If I upgrade to a 5000 series CPU on a Dual-Channel platform, I'll have to lose half my RAM, OR buy 4 32GB sticks. Or I could buy into Threadripper 5000, which probably has the price right in the name...

Being an enthusiast aint easy...
2990WX?
 
You need a 64 core 5xxx thread ripper my friend. Or learn c4d and use octane render with a 3080
 
When are the benchmarks supposed to be available?

I'm bummed I just bought a 3.0 NVME. Oh well, I guess it's not going to matter all that much.
 
Anyone who was expecting 5.0GHz out of it was out of touch with reality. We may have wanted it, but no one should be surprised it didn't happen. As for overclocking, don't get your hopes up on that front either. Ryzen has never been a good overclocker. To extend its lead over Intel as far as it did, I'd wager AMD had to bin these pretty close to the edge of what the silicon could do using TSMC's 7nm process node. So no, I don't think overclocking to 5.0GHz with good cooling is doable. You can only cool Ryzen so much given how small the physical die is.
It seems AMD wasn't that far away from hitting 5 GHz, and they could have but decided against it in favor of availability:
https://twitter.com/aschilling/status/1314247773584228355

Also Summit Ridge through Picasso could very well be overclocked. That stopped only with Matisse. Renoir is somewhere in between can be overclocked to ~4,6 GHz without much effort.
 
Threadripper 5000 series should be amazing with these IPC gains. It'll require a new motherboard though, won't it?

no it shouldn't due to the changes that were made with the TRX4 socket but i wouldn't expect to hear anything about them til probably February or March. going to depend on yields and demand for the epyc chips most likely.
 
When are the benchmarks supposed to be available?

I'm bummed I just bought a 3.0 NVME. Oh well, I guess it's not going to matter all that much.
It is said to be on the 5th or the night before. Also you won't notice a real world difference between 3.0 and 4.0. You saved yourself some money.
 
It seems AMD wasn't that far away from hitting 5 GHz, and they could have but decided against it in favor of availability:
https://twitter.com/aschilling/status/1314247773584228355

Also Summit Ridge through Picasso could very well be overclocked. That stopped only with Matisse. Renoir is somewhere in between can be overclocked to ~4,6 GHz without much effort.
I'd guess they didn't hit 5ghz due to yeilds not availability. And I'd guess the lowish base clocks are due to TDP constraints, not thst the cores can't all do higher.
 
Did I miss something during the launch, or did they not really discuss any changes to memory support? Does it really make sense to go with ram faster than 3600 even on the current 3000 series?
 
Did I miss something during the launch, or did they not really discuss any changes to memory support? Does it really make sense to go with ram faster than 3600 even on the current 3000 series?
Memory support is still 3200mhz, using the same memory controller as ryzen 3000. Though I think with the xt cpus people had better memory controller stability at 1900+, so maybe they've refined it to perform better.
 
When are the benchmarks supposed to be available?

I'm bummed I just bought a 3.0 NVME. Oh well, I guess it's not going to matter all that much.

Embargo date should be November 5th. I imagine there will be some leaked benchmarks of questionable validity between now and then.
 
So my 2950X is now trash.

My renders are getting more complicated and taking literal DAYS on my system, and upgrading to a 16-24 core chip using Zen3 will greatly improve render times...

The only issue is that I have 128GB of RAM on 8 sticks. If I upgrade to a 5000 series CPU on a Dual-Channel platform, I'll have to lose half my RAM, OR buy 4 32GB sticks. Or I could buy into Threadripper 5000, which probably has the price right in the name...

Being an enthusiast aint easy...
It was already trashed... by the 3950X.;) All kidding aside, it really is tough trying to keep up to date nowadays. Back when Intel was giving us 4-core after 4-core, it was pretty easy to stay put and not lose a lot of ground.
 
Humm interesting. Given that I’ve just retired my i7 920(due to death), would you recommend trx40 for the long term or just go am4?

They will both have similar longevity - new socket coming for ddr5 - so depends on the number of cores or pcie lanes you need. Trx40 is pretty niche, am4 covers most needs. Basically if you don't know for sure you need trx40, you don't need trx40.
 
They will both have similar longevity - new socket coming for ddr5 - so depends on the number of cores or pcie lanes you need. Trx40 is pretty niche, am4 covers most needs. Basically if you don't know for sure you need trx40, you don't need trx40.
Over the years I upgraded hard drives, video cards (don’t always need the top of the line). More ram and Blu-ray drives. Last system lasted almost 10 years!
 
I was all set to go with the 5800X but I think the 5900X might be the better price/performance option...
 
I was all set to go with the 5800X but I think the 5900X might be the better price/performance option...
I just want to know if the 5800X will be a single 8 CCX or 4+4 CCX. If it’s the former it might be better off in latency sensitive tasks such as gaming as I’m assuming the 5900X will be 6+6.
 
It will be single ccx, judging by the size of the cache. Any cpu using two ccxs has 64mb of l3 cache.

That means only the top two skus which also have the highest bins giving them the highest single core perf.

But for 98% of ppl a 5800x should be the sweet spot.

Hmm, a 5950x with PBO running with a 150w+ target should be quite fun.
 
That means only the top two skus which also have the highest bins giving them the highest single core perf.

But for 98% of ppl a 5800x should be the sweet spot.

Hmm, a 5950x with PBO running with a 150w+ target should be quite fun.
Yeah, the $100 price difference between the 5800x and 5900x seems to make a bit more sense now. Going from 8 to 12 cores will not benefit everyone, and I'm sure in some workloads, the 5800x will actually perform better thanks to that single ccx. I'm hopeful for better overclocking than zen 2. Even with IPC gains, boost clocks are up a couple hundred MHz across the board, so I'm guessing all core will be the same case.
 
So my 2950X is now trash.

My renders are getting more complicated and taking literal DAYS on my system, and upgrading to a 16-24 core chip using Zen3 will greatly improve render times...

The only issue is that I have 128GB of RAM on 8 sticks. If I upgrade to a 5000 series CPU on a Dual-Channel platform, I'll have to lose half my RAM, OR buy 4 32GB sticks. Or I could buy into Threadripper 5000, which probably has the price right in the name...

Being an enthusiast aint easy...
Don't feel bad. Just think of yourself as an Intel fanboi for the last decade. Same feeling. Yet yours might be all a bit dramatic even though :ROFLMAO: :ROFLMAO: :ROFLMAO: .
 
I'd guess they didn't hit 5ghz due to yeilds not availability. And I'd guess the lowish base clocks are due to TDP constraints, not thst the cores can't all do higher.
Is it not obvious that everyone is struggling@ or around 5GHz? For how long now? Since P4 for X86?
But you have a philosophy on why no 5 GHz for Zen3?
Good for you buddy. You go make sense of 5 GHz and report back to us.
:ROFLMAO: :ROFLMAO: :ROFLMAO: :ROFLMAO:
 
I'm not impressed, especially with that demo at the end with Borderlands 3 and COD Warzone using the 5900x and Big Navi. I was totally prepared to move to the new 5950x but I'm getting 150 FPS @ 4K in box those games. COD Warzone for sure but I know I am getting a hellva lot more than 60fps in Borderland 3.

Still, it's a nice upgrade for AMD fans.
Rarely have I seen this much nonsense in so little words.


I'm surprised and disappointed with the new lineup. I expected to get better value than with the Zen2 parts. But it seems to me that cost/performance figures are about the same with the price hike. I secretly hoped for a downward shift in the price segmentation. eg: 12 core part replaces the 3800X, and 16 core part replaces the 3900X, while the 5950X would be a monster 20 core.

I thought I'd be upgrading to Zen3 from my 3700x, but now that the cat is out of the bag, I think I'll pass. 5900X doesn't seem like a better deal than a 3900X was all along. The 5950X is outside of what I'm willing to spend.
You had unrealistic expectations. Why would they go for better value when they move from little to no competition? They are dominating the CPU space right now and charging less would only hurt them financially for some unreasonable altruistic cause.
 
I just want to know if the 5800X will be a single 8 CCX or 4+4 CCX. If it’s the former it might be better off in latency sensitive tasks such as gaming as I’m assuming the 5900X will be 6+6.

It will be single ccx, judging by the size of the cache. Any cpu using two ccxs has 64mb of l3 cache.

so for gaming the single CCX design is more important then the higher cache on the 5900X (64 MB)?
 
You had unrealistic expectations. Why would they go for better value when they move from little to no competition? They are dominating the CPU space right now and charging less would only hurt them financially for some unreasonable altruistic cause.
Charging less = Better value = More sales. If they offered better value I'd have immediately jumped on it as soon as they become available. But the case being what it is, I'm holding on to the 3700x for the foreseeable future.

Selling 3 CPUs for 400 is better than selling 2 CPUs for 500. Unless they can't fulfill demand in which case it's another story entirely.
 
Charging less = Better value = More sales. If they offered better value I'd have immediately jumped on it as soon as they become available. But the case being what it is, I'm holding on to the 3700x for the foreseeable future.

Selling 3 CPUs for 400 is better than selling 2 CPUs for 500. Unless they can't fulfill demand in which case it's another story entirely.
But what if they can now sell cpus to the cutting edge single core performance crowd that always went Intel, while also still selling cheaper 3000 cpus to those who don't need the bleeding edge?

Edit - I feel they're keeping some wiggle room for the inevitable Intel response, anyways.
 
Charging less = Better value = More sales. If they offered better value I'd have immediately jumped on it as soon as they become available. But the case being what it is, I'm holding on to the 3700x for the foreseeable future.

Selling 3 CPUs for 400 is better than selling 2 CPUs for 500. Unless they can't fulfill demand in which case it's another story entirely.
They will probably sell all they can produce as is. 5600 and 5700X will probably follow to provide better value choices when production ramps up.
 
Charging less = Better value = More sales. If they offered better value I'd have immediately jumped on it as soon as they become available. But the case being what it is, I'm holding on to the 3700x for the foreseeable future.

Selling 3 CPUs for 400 is better than selling 2 CPUs for 500. Unless they can't fulfill demand in which case it's another story entirely.
Oh if only it was that easy...
AMD sits near 40% profit margin.
$500*0.40 = $200 margin, which puts the cost around $300...
Now you sell for $400 instead, but the cost doesn't change... So you make 20% or $100 instead. So selling 3 @ 400 makes them $300, while selling 2 @ $500 makes them $400... So no, you can't always just drop prices to sell more and magically make more money. In your example they would have sell TWICE as many @ $400 as they do @ $500 just to make the same amount in the end. Producing 2x the CPUs is something they would most likely struggle supply wise. It makes no sense to do this, and your #s don't add up at all. Also, you know AMD chips will drop from MSRP some down the line to sell the remaining inventory, so pricing low now when they can easily sell out initial supplies would be stupid.
 
Oh if only it was that easy...
AMD sits near 40% profit margin.
$500*0.40 = $200 margin, which puts the cost around $300...
Now you sell for $400 instead, but the cost doesn't change... So you make 20% or $100 instead. So selling 3 @ 400 makes them $300, while selling 2 @ $500 makes them $400... So no, you can't always just drop prices to sell more and magically make more money. In your example they would have sell TWICE as many @ $400 as they do @ $500 just to make the same amount in the end. Producing 2x the CPUs is something they would most likely struggle supply wise. It makes no sense to do this, and your #s don't add up at all. Also, you know AMD chips will drop from MSRP some down the line to sell the remaining inventory, so pricing low now when they can easily sell out initial supplies would be stupid.
Any sources for that profit margin? Or you just made it up because it supports your argument? I highly doubt manufacturing costs is 60% of MSRP. There are costs that don't multiple by units, like R&D. It is fixed and the more units are produced / sold the less it is per unit. So yeah it's not that easy to calculate / unit cost.
 
If they are not going to milk it at first launch in a decade when they have a clear top of stack product, when will they do it? I am kind of happy that they have not inflated the price more.

They have 10%+ of the market where this is a drop in upgrade, people like me where this will be an easy sell, not to mention that 5950/5900 are reasonably priced in comparison to last gen, and then there are always those who want the best no matter the cost.

As long as Zen2 remains on the market, customers will have the option of faster/slower cores for more or less money, so they have the price points covered.

Also if next gen Intel is competitive again, price drops are inevitable.
 
Any sources for that profit margin? Or you just made it up because it supports your argument? I highly doubt manufacturing costs is 60% of MSRP. There are costs that don't multiple by units, like R&D. It is fixed and the more units are produced / sold the less it is per unit. So yeah it's not that easy to calculate / unit cost.

Look at their financials... AMD sits about 40%, NVIDIA is around 60%... these are known and disclosed numbers (at least previous numbers, they may be going for slightly higher margins with this release for all we know).

"AMD's 43.8% gross margin represents a strong increase of 3.1 percentage points from a year earlier".. so it's actually gone up 3% this year from last, so it's slightly higher than 40% now... it's not as if this is secret information that you can't verify.
https://www.investopedia.com/amd-earnings-4692801

"NVDA has a Gross Margin % of 58.85% as of today(2020-10-02)"
https://www.gurufocus.com/term/grossmargin/NVDA/Gross%2BMargin/NVIDIA+Corp

Well, now you know, I didn't just pull #'s out to support my argument, sorry to let you down ;). But yes, businesses do run on these types of margins even if it doesn't support your argument. I know everyone thinks AMD should run on lower margins than NVIDIA and Intel so they can have better prices, but it's not a charity and they really need to get their margins higher to actually compete with nvidia and intel. AMD forced Intel into lower profit margins with Zen...

"Intel's gross profit margin hit its five-year low in December 2019 of 58.6%"
https://finbox.com/NASDAQGS:INTC/explorer/gp_margin

It's all the way down to 58%... why people think AMD should offer lower prices is beyond me, they are already running on slimmer margins than their competitors. I do agree that there are some fixed costs and more products = less per product, but you're still asking them to lower their profit margins even further than they already are, which is already 33% less than their competition (or viewed from their perspective, their competition has a 50% higher margin). If they are trying to take market share and have unlimited supply, it would make sense. If they can sell everything they make already, it doesn't make any sense to lower their prices further. It's still a better value than their competition and they have to be because they are viewed as 2nd best in their market. They have gained back a good amount of market share, now they need to start turning that into profits to further their R&D (and allow expansion into other areas). They have limited capacity at TSMC, so they have to match the price to the market and to their supply.
 
Gross margins aren't necessarily on individual products - just saying (is in the middle of an eMBA program and just had to cover this) - AMD and Nvidia sell a lot of stuff other than just what we buy. Your point of the company as a whole running on smaller margins is accurate and well made though.
 
My 5820k has served me well, I've gotten almost 6yrs out of it. I'm wondering if I should even bother upgrading at this point and wait for the next round? Most of my gaming is 4k/60Hz- I imagine upgrading won't help me push any more FPS but will help me maintain more consistant framerates?
 
Back
Top