I'd love to see a video of AMD explaining how they went so badly wrong with Bulldozer as that first video seems to them celebrating everything that Bulldozer wasn't. Were they simply unable to create a competitive chip at the time?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 2 (Refresh) 3900XT/3800XT/3600XT
I'd love to see a video of AMD explaining how they went so badly wrong with Bulldozer as that first video seems to them celebrating everything that Bulldozer wasn't. Were they simply unable to create a competitive chip at the time?
Soldato
- Joined
- 18 May 2010
- Posts
- 12,812
Ah of course, its because of the BIOS update that I didn't want to go 3000 series plus I want my next build to be ITX so it would be a side grade to B450 ITX board
I got mixed up with an B550 board not support my current CPU
Soldato
My understanding was that they had always banked on programs and tasks becoming threaded and more parallel much earlier than they actually did. In essence Bulldozer was an OK (not great) design, but at a COMPLETELY wrong time (although in more heavily threaded stuff etc they have sometimes aged better for this) as single/low core IPC was to remain more important, especially in the short term; AMD had thrown all their affordable eggs in that basket, and as such, until Zen was ready they had to simply make do with what they had.
AMD were incredibly financially tight and almost went bankrupt during this period, and weren't able to make something new and abandon Bulldozer's design as quickly as they might have wished.
Associate
If Intel were a bit more forward looking with raising core counts then in my opinion AMD would be bankrupt or bought out by now. Even just bringing out 6 cores quicker would have put AMD in a lot of trouble. They stagnated and the rest is history. I'm glad they didn't as at least two players should push both on. Intel really were very close to burying AMD and the management there must be shaking their heads wondering how they survived. They have now given AMD not just the chance to catch up but also lead in the technology battle. Chiplets, Infinity Fabric, 7nm and 5nm processors among others. It's now Intel who are chasing.
Soldato
It wasn't Intel's stagnation or lack of forethought that almost pushed AMD to bankruptcy, it was the illegal manipulation of the OEM market. AMD have proven numerous times that they are the match for and better than Intel in CPU design, it's really only Bulldozer that's the big black spot in recent memory.
Besides, Intel would never push to completely destroy AMD, there would be too many legal issues arising from Intel becoming a monopoly. Also I think there's some really strange legal nuance with AMD's x64 patents which prevent it from being sold and bought up, so if AMD went bust Intel would lose all access to x64, crippling their products.
Besides, Intel would never push to completely destroy AMD, there would be too many legal issues arising from Intel becoming a monopoly. Also I think there's some really strange legal nuance with AMD's x64 patents which prevent it from being sold and bought up, so if AMD went bust Intel would lose all access to x64, crippling their products.
I'm not saying for one moment that this is a "just as planned" situation, but would Intel have wanted to throw AMD a bone for the purposes of competition law?
Ie, if they went bankrupt or got bought out by somebody with a lot of cash, it could have been worse for Intel in the long run?
Or to put it another way: did Intel enjoy having a "barely alive" AMD, more than they would have enjoyed a completely dead AMD?
Associate
Maybe I should have meant maybe "Natural Shrinkage" if you know what I mean. I know full well about what Intel did in the past and are still doing to this day (how the hell can you not get higher than a 2060 in a AMD laptop and they must be giving huge rebates to keep AMD out of the server room). But if Intel had progressed a bit quicker maybe by natural means AMD would have been an irrelevance. There is no way AMD could have kept going with only maybe around 5% marketshare. They would not have got the console contract, that is fairly certain. Without even that, would they have survived? So Intel did give them a chance to survive and get into markets. Intel are defending laptops, OEM and servers ferociously. Can AMD crack those and make the big bucks? I really hope they can, at least 25% marketshare in each will make Intel bleed a lot of money. But it will take a total mindshare change from manufacturers, purchasers and IT managers etc.
Soldato
Before this goes completely off topic, I think your perceptions are way off on this.
Laptops:
You don't get more than 2060s in laptops because OEMs don't have full confidence in AMD's resurgence to properly invest in new chassis designs. You can't just plop and AMD APU into a chassis designed for an Intel CPU and expect it to work properly.
By the time OEMs do have confidence in Ryzen APUs, Nvidia will have mobile Ampere so there's no point in investing in Turing-compatible AMD chassis design.
(it's not outside the realm of possibility that OEMs are waiting for the full run of Ryzen APUs before committing to chassis design given how they're likely to give all but the top Ampere mobile SKUs a run for their money).
Server:
It takes years to test and validate hardware for datacentre, and years to plan migration. You can't just click your fingers and totally change your infrastructure to another hardware provider. Some of these validations and migrations can take longer than EPYC has even existed. However, loss of power and capability is an immediate concern, so server operators have no choice but to buy in more Intel kit to recoup "overnight" performance lost from Intel's security woes.
Some server operations still benefit from Intel's single-threaded performance superiority ("vertical" scaling I think it's called?).
NEW contracts and server operations that require "horizontal" scaling (multi-threaded performance) are jumping on EPYC and have been for years.
Consoles:
AMD was in the Wii and powered PS4 and XBOne before Zen even existed. The revenue from those alone would've kept the company afloat despite their PC CPUs being woeful at the time. So it's not like they landed "the console contract" out of the blue, AMD has been working with the 3 console providers for over a decade already.
All in all, AMD DID keep going with about 5% market share because they are a tiny company compared to Intel. They don't actually need a massive market share to exist.
Now if Intel wanted to they could've have crushed AMD, but as I said before - and FoxEye also points out - there would have been serious legal ramifications if they did. So it's not really a case of Intel gave AMD a bone to keep them ticking over, it's more like they literally could not kill them off even if they were so inclined.
Regardless of how we got here though, you are correct in saying that Intel got fat and lazy, stagnant and complacent, and now it is really biting them in the ass. Mindshare is significantly shifting though, especially in the face of Intel's ongoing security woes, inability to produce competitive and efficient products and massive supply strain by being stuck on 14nm nodes (even after the years of bribes, even Dell are talking about moving away from Intel because they just can't supply enough product).
Laptops:
You don't get more than 2060s in laptops because OEMs don't have full confidence in AMD's resurgence to properly invest in new chassis designs. You can't just plop and AMD APU into a chassis designed for an Intel CPU and expect it to work properly.
By the time OEMs do have confidence in Ryzen APUs, Nvidia will have mobile Ampere so there's no point in investing in Turing-compatible AMD chassis design.
(it's not outside the realm of possibility that OEMs are waiting for the full run of Ryzen APUs before committing to chassis design given how they're likely to give all but the top Ampere mobile SKUs a run for their money).
Server:
It takes years to test and validate hardware for datacentre, and years to plan migration. You can't just click your fingers and totally change your infrastructure to another hardware provider. Some of these validations and migrations can take longer than EPYC has even existed. However, loss of power and capability is an immediate concern, so server operators have no choice but to buy in more Intel kit to recoup "overnight" performance lost from Intel's security woes.
Some server operations still benefit from Intel's single-threaded performance superiority ("vertical" scaling I think it's called?).
NEW contracts and server operations that require "horizontal" scaling (multi-threaded performance) are jumping on EPYC and have been for years.
Consoles:
AMD was in the Wii and powered PS4 and XBOne before Zen even existed. The revenue from those alone would've kept the company afloat despite their PC CPUs being woeful at the time. So it's not like they landed "the console contract" out of the blue, AMD has been working with the 3 console providers for over a decade already.
All in all, AMD DID keep going with about 5% market share because they are a tiny company compared to Intel. They don't actually need a massive market share to exist.
Now if Intel wanted to they could've have crushed AMD, but as I said before - and FoxEye also points out - there would have been serious legal ramifications if they did. So it's not really a case of Intel gave AMD a bone to keep them ticking over, it's more like they literally could not kill them off even if they were so inclined.
Regardless of how we got here though, you are correct in saying that Intel got fat and lazy, stagnant and complacent, and now it is really biting them in the ass. Mindshare is significantly shifting though, especially in the face of Intel's ongoing security woes, inability to produce competitive and efficient products and massive supply strain by being stuck on 14nm nodes (even after the years of bribes, even Dell are talking about moving away from Intel because they just can't supply enough product).
I'm asking about the refreshes.
Does anyone know when/if AMD are going to make a 3700X refresh?
They won't. Every 6-8 core Zen 2 at reasonable base and boost speeds is covered now with these introductions.
You can start to understand why AMD is delaying Zen 3 - because it has no competition.
https://www.techpowerup.com/268046/no-intel-rocket-lake-s-or-ice-lake-x-this-year
Intel will not release rocket lake in 2020
https://www.techpowerup.com/268046/no-intel-rocket-lake-s-or-ice-lake-x-this-year
Intel will not release rocket lake in 2020
Soldato
Which has been the case since last year/early this year. They knew they'd only be able to do a further 14nm++++ release for desktop in 2020, and 10nm reserved only for Xeon/DC in 2020 for obvious reasons.
As for the delay in Zen3, I don't think they'll delay Zen3. I fully expect them to release the Zen3 EPYC parts as promised in 2020 be that late Q3 or Q4.
The 4.2 Ghz base on the 3800XT might be bs after all: https://twitter.com/_rogame/status/1268547560505069569
4.56 Ghz to 4.76 Ghz had an 8% score increase.
These scores were done with DDR4-2933Mhz apparently...
4.56 Ghz to 4.76 Ghz had an 8% score increase.
These scores were done with DDR4-2933Mhz apparently...
Last edited:
I know that line of thinking. Again, I'm talking about the "new" chips (refreshes) only. But it's a minor point obviously.
Soldato
Since the XT's are not on a new node or architecture, they could sell you an 8-core CPU that has a 3.9 base and a 4.5 boost with "3700XT" on the side of the box.
Is that what you're looking for? (a 3800X in a new box that says "3700XT" on the side)
That score is more like a 4.1ghz all core.
It's particularly amusing when people are still trotting out the "You'll be waiting forever!" line and you're only a few weeks away from a new gen... still it happens.
Although Zen3 is likely a 2021 product now, rather than being a few weeks away.
I don't get the impatience either. Clearly people who are prepared to wait aren't exactly suffering with what they have. Don't have an urgent need to replace it. So why people keep getting admonished for having a little patience I have no idea.
Considering the quality of silicon in the 3950 there's not really scope to do this.
The 3900 and below were using lower quality silicon and now yields have improved they can step this up.
The 3800X uses the same chip.