Is Intel in 2019 = AMD in the early 2010s?

[Kisai]

42 minutes ago, leadeater said:

What about the TDP? Intel has similar TDP on the Xeons also so I don't see how this matters much at all, buy a lesser SKU with a lower TDP if you think you don't have the cooling required. Remember TDP is only about cooling so nothing else matters other than that, if you can cool it then what the number happens to be isn't of great relevance. That's why Cascade Lake-AP being 400W TDP doesn't matter, it's designed for direct die water cooling and any system you can buy with them comes like that as it's BGA only.

It matters when your purchase decisions depend on parts you already have (including UPS equipment.) I bought a larger case last time because the Geforce GTX 1080 didn't fit the one I had, which was also a large case with PWM fans I had purchased specifically to replace the three-speed RGB ones that made too much noise. Current case has no LED fans and is too much of a pain in the ass to take apart to replace the fans if I decide to.

 

Existing system has a 84w TDP. (Passmark Single thread: 2227; MT: 9778)

AMD Ryzen 9 3900X has a 105w TDP (Passmark Single thread:  2914; MT: 31782 )

AMD Ryzen 5 3600X has a 95w TDP (Passmark Single thread: 2907 ; MT: 20502)

AMD Ryzen 7 3700X has a 65w TDP (Passmark Single thread: 2916 ; MT:23907)

 

So any of those parts is a 25% theoretical improvement, with all of the Multi-threaded options being at least twice the performance, entirely because they have more cores (6, 8 and 12 respectively), I said at one point in time, that when this system reaches the end of it's useful life, I'd probably end up buying a workstation platform because the desktop boards and CPU's keep heading in the direction of being terrible (where 40 PCIe lanes are available for a 2x16+2x4 configuration instead of nonsense 24 lanes.)

 

https://www.guru3d.com/news-story/amd-ryzen-3000-new-block-diagram-about-pcie-4-on-matisse-and-x570-chipset.html

 

As it is the Geforce GTX 1080 has thus far refused to operate in 16x mode on the existing motherboard, because... there are other cards in the other PCIe slots. Now what happens if I decide to put an m2 PCIe SSD in? The existing way that you lose 8 lanes by plugging in ANY second PCIe card into the CPU's PCIe lanes is terrible.

 

Now I could just remove said 1x PCIe parts and replace them with a USB 3.0 device, thus moving it from the CPU to the PCH's lanes, but now we're just shuffling things around because of enforced design limitations of "desktop's that nobody upgrades"

 

[leadeater]

19 minutes ago, Kisai said:

The existing way that you lose 8 lanes by plugging in ANY second PCIe card into the CPU's PCIe lanes is terrible.

Yea that's always sucked, still a thing on HEDT too just way less of a problem. Without a PLX chip you just can't dynamically allocate PCIe lanes to slots as required so you end up having to design each slot in two static dual mode operations which means any time another slot shares lanes with another one it drops to the other mode no matter how many lanes you are actually using. I get that on my X79 system with my 10Gb NIC, have to put it in a specific slot or my 2nd GPU drops to x8.

 

19 minutes ago, Kisai said:

It matters when your purchase decisions depend on parts you already have.

Unless you have one of the worst motherboards with effectively no VRM cooling and pair it with the 12 or 16 core it's not going to matter, TDP wise. Actual power usage isn't any different to anything in the past or currently from Intel.

 

As for the X570 chipset you can actually run that without a heatsink at all on it, it won't throttle. I have no idea what AMD was thinking but der8auer did testing to actually see if it's needed and the answer is no. You can use the chipset coolers to cool the CPU though, barely but you can lol.

[mr moose]

1 hour ago, Ryan_Vickers said:

and you can bet that the people running them are well aware of this.

absolutely,  which is why I generally only listen to people who work  directly with the gear in the industry. I find it all too often  that most of the comments regarding server-side computing on forums come from domestic end users with zero clues there is even a market difference let alone a complete difference in hardware to end users.

 

It has been a while since I did any work in that industry so I try to restrict my comments to those which I have been directly told and not assumed from some article or forum post.

 

.

 

 

[MMKing]

It's difficult to say, but i would say AMD was in a worse situation. And it will take at least a couple more years for Intel to start losing, but here is the thing. Due to Intel owning all of their own equipment, factories, fabrications, labs and all the other infrastructure. There is no painless way for Intel to scale back, huge parts of the company is integrated. You can't close down 10% of it, not in a week or half a year.

 

Then again, revenue for 2018 is still up compared to revenue for 2017. Intel is not losing money, they are as profitable as ever. But Intel did report a debt of 25.09 billion USD per June 2019, representing about 1/3 the company's equity. Right now, this is a tolerable situation for Intel. They have debt, but they have more than enough money to pay off said debt.

 

Meanwhile, AMD has a total debt of 1.9 billion USD. Representing about 54% of their equity.

 

This whole situation, beginning with Ryzen in the first quarter of 2017. Could be the very first signs of Intel catastrophically collapsing, akin to how giants such Kodak collapsed in the 1980s onward. Where sales fail, parts of the company is sold off or closed while debt continues to mount.

 

More likely, Intel will have a few years of lower profitability and at worst a slight loss in revenue a year or two. But it's unlikely Intel will go in the negative, and be forced to borrow money to maintain their company.

[vanished]

9 hours ago, mr moose said:

absolutely,  which is why I generally only listen to people who work  directly with the gear in the industry. I find it all too often  that most of the comments regarding server-side computing on forums come from domestic end users with zero clues there is even a market difference let alone a complete difference in hardware to end users.

 

It has been a while since I did any work in that industry so I try to restrict my comments to those which I have been directly told and not assumed from some article or forum post.

Do you remember when the first wave of vulnerabilities went around and many major providers including Amazon took a noticeable hit to their available CPU performance?  The impact varied by company and workload but in at least a few high profile cases it was significant enough to be reported on publicly - no need for internal knowledge or getting hands on with the servers.  More recently, Google and Microsoft, along others, have signed up to use EYPC 2 chips.  Clearly the industry is ready to give them a try.  Whether it's the lead in performance, cost, energy, interconnectivity, or security, or some combination of all those and more, AMD has managed to get their foot in the door in a big way.  I don't think it's that insane or speculative to think this wouldn't happen if they were so content with Intel and/or weary of AMD.

[mr moose]

1 hour ago, Ryan_Vickers said:

Do you remember when the first wave of vulnerabilities went around and many major providers including Amazon took a noticeable hit to their available CPU performance?  The impact varied by company and workload but in at least a few high profile cases it was significant enough to be reported on publicly - no need for internal knowledge or getting hands on with the servers.  More recently, Google and Microsoft, along others, have signed up to use EYPC 2 chips.  Clearly the industry is ready to give them a try.  Whether it's the lead in performance, cost, energy, interconnectivity, or security, or some combination of all those and more, AMD has managed to get their foot in the door in a big way.  I don't think it's that insane or speculative to think this wouldn't happen if they were so content with Intel and/or weary of AMD.

No, I have no first hand experience of the performance hits due to those mitigations.   I haven't even heard of people complaining about them outside of the usual forum noise.  But then I suspect most of that was just people saying stuff because they wanted to join in the conversation.  I can ask my mates who manage education networks (reasonable size server system) what they experienced. 

 

 

[vanished]

3 minutes ago, mr moose said:

No, I have no first hand experience of the performance hits due to those mitigations.   I haven't even heard of people complaining about them outside of the usual forum noise.  But then I suspect most of that was just people saying stuff because they wanted to join in the conversation.  I can ask my mates who manage education networks (reasonable size server system) what they experienced.

Well it was talked about a fair bit and it's not like it was all made up.  May as well, would be interesting to know if they were some of the ones affected.

[mr moose]

1 minute ago, Ryan_Vickers said:

Well it was talked about a fair bit and it's not like it was all made up.  May as well, would be interesting to know if they were some of the ones affected.

 I never said the performance hits from meltdown or any of those first wave exploits were made up.  I merely pointed out that there is a reason Intel maintain  a large market lead in the corporate world and that what we hear on forums about it is largely ignorant noise.

[vanished]

Just now, mr moose said:

 I never said the performance hits from meltdown or any of those first wave exploits were made up.  I merely pointed out that there is a reason Intel maintain  a large market lead in the corporate world and that what we hear on forums about it is largely ignorant noise.

That I can agree with