From ATI to AMD back to ATI? A Journey in Futility @ [H]

Well since I pointed out Anandtech again, this is the second time, to you , why not just drop it?

There is a post in this or the Vega Thread where I told everyone how to program around this problem too........ at least for single databases without any interaction with other databases, but that is a rarity in today's world.
 
razor1 said:
Look at Anandtech's preview, they showed it right there. LOL forgot about that?

https://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/18

I called it even before Eypc was released, its common sense if you know anything about database programming and how NUMA works.

Ryzen has huge issues with data locality. Eypc even worse. I know exactly how to program around it. But can't do that with older programs. Need updates, and bigger the database the worse/harder, performance and difficulty in fixing its going to be.

This is not BS, its well know where the problem is!
Click to expand...
So you are focusing on one negative. You should look at the summation:

All in all, it must be said that AMD executed very well and delivered a new server CPU that can offer competitive performance for a lower price point in some key markets. Server customers with non-scalar sparse matrix HPC and Big Data applications should especially take notice.

As for Intel, the company has delivered a very attractive and well scaling product. But some of the technological advances in Skylake-SP are overshadowed by the heavy price tags and somewhat "over the top" market segmentation.
Click to expand...

The whole of the conclusion read like that, far from being as catastrophic as you point out. That is my point. No one believes EPYC will work for everyone, but at its price and in those sectors they do work well in it is a win. It will take time to get anywhere, and likely EPYCs impact wont be truly felt for a year or 2 more.
 
Please watch the name calling, flame-baiting, and trolling posts, please.
 
JustReason said:
So you are focusing on one negative. You should look at the summation:



The whole of the conclusion read like that, far from being as catastrophic as you point out. That is my point. No one believes EPYC will work for everyone, but at its price and in those sectors they do work well in it is a win. It will take time to get anywhere, and likely EPYCs impact wont be truly felt for a year or 2 more.
Click to expand...


That is HUGE negative, do you realize any program that needs to cross communicate data needs a database! Think about all the programs that have search features? Amazon, Newegg, any online store, any web search. Any forms with auto fill, any forms period, all are database dependent!

Any server that connects to the web, web services, any server that is needed for CRM, CSM, etc all of these a database driven. Its like 75% of server sales are for databases! Actually its probably higher than 75%, the only place Eypc can compete with is in some HPC spaces, which Intel's offerings do just fine too, so you are left with what? 5% of the effective server market that might be interested in Eypc? What about the other 95%?

This is HUGE! The one thing server CPU's should be good at no matter what, Eypc comes up woefully short!
 
razor1 said:
That is HUGE negative, do you realize any program that needs to cross communicate data needs a database! Think about all the programs that have search features? Amazon, Newegg, any online store, any web search. Any forms with auto fill, any forms period, all are database dependent!

Any server that connects to the web, web services, any server that is needed for CRM, CSM, etc all of these a database driven. Its like 75% of server sales are for databases! Actually its probably higher than 75%, the only place Eypc can compete with is in some HPC spaces, which Intel's offerings do just fine too, so you are left with what? 5% of the effective server market that might be interested in Eypc? What about the other 95%?

This is HUGE! The one thing server CPU's should be good at no matter what, Eypc comes up woefully short!
Click to expand...

The world doesn't revolve around databases and there are even ways around that weakness. Does not change the fact that several companies have signed on for Epyc, will see if others will also join in, will take time before you see it make a serious dent in the server world. We used to do tons of simulations in house at Chrysler and Epyc would work just fine for that and anything that could do 90% of peak speed for half the cost we would buy. You got it cranked to 11 on AMD lately unless your holding a ton of Intel stock and are worried I dont think it matters a whole ton to us what Epyc does, would just be more income for AMD and hopefully better competition with Intel. Google and Amazon you may not be into Epyc but that is not the whole market but obviously two companies AMD would like to get, but who knows.
 
  • Like
Reactions: N4CR
like this
Sure, we all know how to program around it. The question is, would anyone bother?

It's not just programmer time cost, which is huge, but also higher risks. Changing the existing programming paradigm for a well-known program can introduce huge headaches, performance issues, all of which leads to much, much higher support costs.

I'm sorry, but no sane person with a mission critical server will choose Epyc. If he/she does, they deserve to be fired on the spot.
 
Meeho said:
How realistic is a L3 fitted database vs much larger ones?
Click to expand...
That was reflected in the comments. They claim to have a second review coming, though not sure when. They said AMD claimed setting memory to interleaving would boost performance 10-15%. They were however unable to test as they were cutoff from their connection, out of time.
 
razor1 said:
Yeah its not just RTG, its all of AMD, we will see that in the next 3 Q's lots of red.

They aren't competitive that is the problem, that is why they aren't showing on the bottom line what AMD has been saying in the past Q's and its got Investors spooked, cause if AMD is doing well, it has to be hurting Intel or nV, but those two guys are doing better then usual.
Click to expand...

When mindfactory.de data was spread among newsites, everyone focused in raw final data, but I analyzed the incremental sales and found that the increase in sales of AMD products was not coming at the expense of Intel sales, in general.

I advanced then the hypothesis that most people was purchasing RyZen was people wasn't going to purchase Intel, and that only some few RyZen sales were from people that chose AMD instead Intel.

Then days latter Lisa Su posted a tweet thanking "AMD fans" for sales in mindfactory. As I mentioned then, she using the term "fans" instead the term "customers" just added weight to my hypothesis that most AMD sales were coming from people would purchase only AMD.

Certain people invented weird arguments to try to dismiss my claims (just search that past discussion), but Q3 numbers have confirmed what I said. AMD had a good Q3, but Intel also had a good Q3. As analysts are now admitting, RyZen has minimum impact on Intel sales.
 
Last edited:
Meeho said:
How realistic is a L3 fitted database vs much larger ones?
Click to expand...

Hmm depends, the levels are different for different databases, usually its just user level control bases on what ever the company that makes the database sees fit to call it lol.

The size of the database, as long as it can fit on one node (core) it won't have a problem, and this is a where the problem is, Eypc has to be coding per node, a systems architect must know which Eypc processor the software is targeting, because he needs to target where the database resides, because there is no NUMA control so the only way to stop the CCX cross talk is if the programmer specifically states which core is doing what. So what the difference between a 12 core Eypc vs a 16 core Eypc or other sizes. The arrangement of cores is different across the modules. If it was a standard different cpu's NUMA takes care of this problem on its own.
 
Gideon said:
The world doesn't revolve around databases and there are even ways around that weakness. Does not change the fact that several companies have signed on for Epyc, will see if others will also join in, will take time before you see it make a serious dent in the server world. We used to do tons of simulations in house at Chrysler and Epyc would work just fine for that and anything that could do 90% of peak speed for half the cost we would buy. You got it cranked to 11 on AMD lately unless your holding a ton of Intel stock and are worried I dont think it matters a whole ton to us what Epyc does, would just be more income for AMD and hopefully better competition with Intel. Google and Amazon you may not be into Epyc but that is not the whole market but obviously two companies AMD would like to get, but who knows.
Click to expand...


Everything in computers revolves around databases, even games man. The smallest of programs that have to save data and retrieve them all of it is databases.

You love playing MMO's or online games, all those use giant size databases.

You don't think they matter, but every damn software has some sort of DB in them to make them run. Its the size of that database or multiple databases with aggregate data that hurts Ryzen/Eypc.
 
JustReason said:
Actually TR has been getting the nod over Intel in many a review. EPYC too has gotten extremely positive reviews. Ryzen originally had positives even on an immature platform, when considering price of CPUs or even adding platform costs Ryzen is a huge win. Discrepancies in performance between a 6 to 8 core mean little to the common consumer the Ryzen platform is geared toward.
Click to expand...

Those three-four biased reviews of TR comparing 10-core Intel to 16-core threadripper on rendering, rendering, rendering, rendering, rendering,... and that ignored everything else? Those biased sites and new sites that forced an angry answer from der8auer? Those biased reviews that crippled SKL-X chips in any imaginable way?

This is a honest and complete characterization of the i9-7900X:

In the end Intel’s i9-7900X appears to offer the best combination of singlethreaded performance, multithreaded performance, and efficiency at the $1000 price point. It’s not as fast as AMD’s Ryzen Threadripper in well threaded tasks, but it offers significantly stronger performance in single and lightly threaded workloads while remaining more efficient than the competition. More to the point its performance in multithreaded workloads is really quite good. Given the massive disadvantage it has in core count, the gap in performance is smaller than one would expect. As time goes on I wouldn’t be surprised to see the performance needle move further in favor of Threadripper as these apps are updated to make better use of the high core counts now at their disposal. But for now Intel’s i9-7900X offers extremely stiff resistance to AMD’s Ryzen Threadripper assault.
Click to expand...

And once Intel released the higher core count models, ThreadRipper momentum has gone. Why do you believe AMD just introduced a huge price discount in the 1950X? Because sales are being stellar or because it isn't selling well? Be serious.

Do you mean that EPYC review from ServeTHehome showing that intrasocket latency in 1P EPYC was so bad that was at same level than intersocket latency in 4P Broadwell Xeon?

Do you mean that biased EPYC review from Anandtech where the performance of Skylake and Broadwell Xeons was crippled by 40--60% in order to get EPYC competitive?

RyZen got momentum only because 8-core Zen was compared to 4-core Kabylake. Once Intel released 6-core CoffeLake, the RyZen multithread advantage has gone, and Intel continues to win in everything else.
 
Last edited:
razor1 said:
Everything in computers revolves around databases, even games man. The smallest of programs that have to save data and retrieve them all of it is databases.

You love playing MMO's or online games, all those use giant size databases.

You don't think they matter, but every damn software has some sort of DB in them to make them run. Its the size of that database or multiple databases with aggregate data that hurts Ryzen/Eypc.
Click to expand...

Look who liked your post, go figure they drivel on about nothing as well, yet AMD keeps on selling chips. Funny you guys dont like it yet the reviewers all liked it, even serve the home recommend it. I will take their reviews and thoughts over yours.
 
Gideon said:
Look who liked your post, go figure they drivel on about nothing as well, yet AMD keeps on selling chips. Funny you guys dont like it yet the reviewers all liked it, even serve the home recommend it. I will take their reviews and thoughts over yours.
Click to expand...


The only people that say AMD is selling chips, is AMD in presentations, their bottom line isn't showing any change from before outside of margin changes, and even those increased margins aren't anywhere near what AMD forecasted they will be at.

Do you know the entire financial industry, health care industry, government is run off of databases too, and all of these three use very old databases, based off of cobol, fortan, legacy languages that will probably never be updated to newer languages because its too costly for the turn around? Without databases we couldn't really do much with computers in the professional world.

Yeah go figure, all of us actually do work in sectors that use computers daily based on these things. That is why we know. You call yourself an engineer, but still. Stick with mechanical and you will be fine.

Even the software you use to safe your mechanical engineering files, uses a database, just open your files up in notepad or what ever text editor, you will see those files are comprised of multiple files and the save file acts as a header to point to the different parts.

Shit these stuff is easy to know. A database is a just a program that has fields and you can save to or get info from.

All programs that are currently used have databases to some degree.
 
Last edited:
razor1 said:
The only people that say AMD is selling chips, is AMD in presentations, their bottom line isn't showing any change from before outside of margin changes, and even those increased margins aren't anywhere near what AMD forecasted they will be at.

Do you know the entire financial industry, health care industry, government is run off of databases too, and all of these three use very old databases, based off of cobol, fortan, legacy languages that will probably never be updated to newer languages because its too costly for the turn around? Without databases we couldn't really do much with computers in the professional world.

Yeah go figure, all of us actually do work in sectors that use computers daily based on these things. That is why we know. You call yourself an engineer, but still. Stick with mechanical and you will be fine.

Even the software you use to safe your mechanical engineering files, uses a database, just open your files up in notepad or what ever text editor, you will see those files are comprised of multiple files and the save file acts as a header to point to the different parts.
Click to expand...

*yawn* Dont you have some stuff to draw.
 
Gideon said:
*yawn* Dont you have some stuff to draw.
Click to expand...


Yeah I do, but I also have a lunch break ya know ;).

Basic shit man. The entire cloud is based off of DB's, yet AMD tries to market Eypc for cloud services, its Eypc, Eypc Fail.

Google search is your friend

https://www.techwalla.com/articles/how-are-databases-used-in-the-real-world

Databases are used just about everywhere including banks, retail, websites and warehouses. Banks use databases to keep track of customer accounts, balances and deposits. Retail stores can use databases to store prices, customer information, sales information and quantity on hand. Websites use databases to store content, customer login information and preferences and may also store saved user input. Warehouses use databases to manage inventory levels and storage location. Databases are used anywhere that data needs to be stored and easily retrieved. The filing cabinet has all but been replaced by databases.
Click to expand...

Yawn is right, I'm yawning, because having a discussion with you, I can turn off my cerebrum and have my cerebellum do all the work.
 
Last edited:
ecmaster76 said:
I imagine the main holdup is Intel's Marketing department trying to figure out how to spin the huge performance gain whilst avoiding bringing any positive light to AMD. Parts have been ready for months if leaks are to be believed.
Click to expand...

Canard PC claims we'll see Kaby Lake G with AMD GPU soon: "almost here." I guess if it was for Apple, might coincide with the December iMac Pro release. Mac mini hasn't been refreshed for a while...would that small form factor work well with MCM?

 
razor1 said:
Yeah I do, but I also have a lunch break ya know ;).

Basic shit man. The entire cloud is based off of DB's, yet AMD tries to market Eypc for cloud services, its Eypc, Eypc Fail.

Google search is your friend

https://www.techwalla.com/articles/how-are-databases-used-in-the-real-world



Yawn is right, I'm yawning, because having a discussion with you, I can turn off my cerebrum and have my cerebellum do all the work.
Click to expand...

Well I know when it comes to financials you turn your brain off, also I love these edits as your mind must work to slow to not get it all the first time.
 
botgreet said:
Canard PC claims we'll see Kaby Lake G with AMD GPU soon: "almost here." I guess if it was for Apple, might coincide with the December iMac Pro release. Mac mini hasn't been refreshed for a while...would that small form factor work well with MCM?

Click to expand...

Will be interesting to see it.
 
botgreet said:
Canard PC claims we'll see Kaby Lake G with AMD GPU soon: "almost here." I guess if it was for Apple, might coincide with the December iMac Pro release. Mac mini hasn't been refreshed for a while...would that small form factor work well with MCM?

Click to expand...

I am still leaning that Kaby Lake G will just use its own Intel brand gpu rather than AMD gpu.
 
Gideon said:
Well I know when it comes to financials you turn your brain off, also I love these edits as your mind must work to slow to not get it all the first time.
Click to expand...


Took a while for a rebuttal ;) btw, both of those are parts of a brain, not turning it off, its just not using higher level functions ;).

As being an engineer I guess those bio classes weren't something to remember.......
 
Gideon said:
Will be interesting to see it.
Click to expand...


Why you see it in Ryzen already, not that great...... Apple needs it for Open CL work, you think AMD gets a lot on their bottom line from Apple? I doubt it, their GPU's get very little, think it was around 150 million a year..... APU, less margins, if not a flat rate to use their IP, which explains why margins increased this Quarter, "an unmentionable source from IP sales".

What would be damn hilarious to see, now we can actually break out the power consumption of Ryzen APU if the specs are close to Ryzen's APU.......
 
razor1 said:
Took a while for a rebuttal ;) btw, both of those are parts of a brain, not turning it off, its just not using higher level functions ;).

As being an engineer I guess those bio classes weren't something to remember.......
Click to expand...

What can I say I have a life outside the forum and the only bio matter I had to worry about was whether or not the airbag would crush your skull or not, it's simple math but perhaps beyond your understanding. As for the APU profits, some is better then none and look at that it's simple math again.
 
Gideon said:
What can I say I have a life outside the forum and the only bio matter I had to worry about was whether or not the airbag would crush your skull or not, it's simple math but perhaps beyond your understanding. As for the APU profits, some is better then none and look at that it's simple math again.
Click to expand...

Odd you posted in other threads after that post ;) lol. Guess its life yeah.

Wow physics, force acceleration based on gas and pressure ;) not complex, pretty easy shit.

Yeah, some is better then none, simple math for you, but that doesn't mean its anything good for AMD ;)

Short term gains don't mean shit in the world of finance, now this is the problem with AMD and why Wall Street is finally seeing all the BS AMD has been stating about Zen, Vega, Polaris, they haven't made any ground.
 
razor1 said:
Short term gains don't mean shit in the world of finance, now this is the problem with AMD and why Wall Street is finally seeing all the BS AMD has been stating about Zen, Vega, Polaris, they haven't made any ground.
Click to expand...

I don't plan to change my signature soon, but if I was to do it, I would add the above quote verbatim, because it reflects reality very well. I have been fighting HYPE and BS during years. It is heartwarming how the truth is seeing the light finally.
 
juanrga said:
I don't plan to change my signature soon, but if I was to do it, I would add the above quote verbatim, because it reflects reality very well. I have been fighting HYPE and BS during years. It is heartwarming how the truth is seeing the light finally.
Click to expand...

Short term drives speculation, which drives stock price, means very little for company health. AMD is healthy than before Zen, before Zen it was close to bankruptcy, by no means is it out of the water yet, and tell ya what, till both divisions GPU and CPU's are much more healthy than now (40+ margins overall 400-500 million a quarter profit) AMD is not going to be healthy. 400 to 500 million per quarter, that means they will have to eat away at Intel and nV, both, in ever segment they are in competition in. Server, HPC, DL, Gaming, Professional.

What are the prospects of that? Minimal.

Gaming, its not even a question mark if they can do it. Right now they can't.

Server, pretty much the same thing.

HPC, they might be able to get something there but its limited.

DL, Nada there too.

Desktop and Notebook, penetration is time line driven, which is based on what Intel does, with CFL, don't see much here either for AMD. I stated this before right around when Zen was released, its up to what Intel does/responds to see how much time AMD has to make any inroads into these markets. At first it seemed like they would at least get some notebook marketshare, but that changed fairly quickly. It looked like they had a solid 1 year before we saw Intel being able to do any meaningful response, because of the 10nm delay, but looks like they had back up plans, Oh well, AMD just got a tough battle to fight.

One analyst can't remember which financial company he works for, but he was gong ho on AMD, started back when AMD was 2 or 3, finally now he says AMD uses too many buzz words and he doesn't see them getting much farther than they already have gone. He was also shorting Intel this entire time, which didn't happen to his favor either.

Speculation is good if a company is on a rebound, investors should take advantage of it, nothing wrong with that, but understanding what the growth potential and when it will peak, is what separates an average investor vs a good one. The only way to understand that is, to know the company and its competition inside and out. Many analysts don't know much about the companies themselves, and the information doesn't come from financial calls either, financial calls always show the company in the best light possible within SEC guidelines. So when AMD says 12-18% drop, expect it to be on the high end of the scale.

Why are nV and Intel still going up in Q4 (predicting) or staying level, well reason is they are growing into markets that are new. So even with the season down turn of the traditional computer markets (consumer) where AMD is also in, they have other markets that aren't affected by seasonality yet, since those markets are growing. AMD hasn't been able to make in roads into those markets and won't for a long time, because they don't have the resources to do it, they barely have the resources to keep what they have in the traditional markets.

Now lets talk about cyrotomining for a sec. Speculation on bitcoin, yeah its awesome right now, made over 200% in the past month and a half. Why is that, its because of the forks of Bitcoin (2 new ones, one of them is GPU mine-able) Bit coin gold, which can be mined on GPU's, is based on Equihash, which is Z cash. Great of nV cards, not as good for AMD cards, AMD cards can still do it but not as good. Why is Wall Street thinking cryptomining is going down? Because AMD says so, its not, if anything there are more prominent coins coming to play with GPU mining than ever before. They don't understand the cryptomarkets and how EVERYTHING is inter related, all coins have certain stigmas from the forks that drive prices up or down. There are analysts saying BTC will go up to 100k by end of next year. That is not realistic, because what are the driving forces to make BTC do that, the only way I can see that happening is if BTC is adopted as a universal payment method by everyone, that is not going to happen in one year, it will take many years for that to happen.

Does anyone really remember why the cryptomarket crashed a couple/few of years ago? It wasn't solely because of not being able to mine with GPU's anymore. It was because of the 200 million dollar theft that occurred in the most used exchanges. It killed cryptocurrancy confidence. Prices dropped and mining because not so profitable.

That will should not happen this time around, hopefully, I'm a strong proponent of what can be made can be unmade, so in my view its always possible. It can happen.

So why is AMD saying what they are when cryptomining is in a boom right now? Well because they aren't as competitive as they were before, that is why they are saying what they are saying. Its not the fault of cryptomarkets, its AMD's products that don't seem as good.
 
Last edited:
Connect the dots?


Apple suppliers upping production of AMD Radeon Pro Vega GPUs for iMac Pro launch
https://9to5mac.com/2017/10/23/apple-suppliers-imac-pro/

The iMac Pro uses high-end GPUs from AMD including the Radeon Pro Vega 56 and 64 graphics cards; these parts are packaged using a newer 2.5D technology ...


Intel Custom Foundry EMIB
https://www.intel.com/content/www/us/en/foundry/emib.html

Embedded Multi-die Interconnect Bridge (EMIB) is an elegant and cost-effective approach to in-package high density interconnect of heterogeneous chips. ... Instead of using a large silicon interposer typically found in other 2.5D approaches, EMIB ...
 
  • Like
Reactions: noko
like this
P Penguin said:
Connect the dots?


Apple suppliers upping production of AMD Radeon Pro Vega GPUs for iMac Pro launch
https://9to5mac.com/2017/10/23/apple-suppliers-imac-pro/

The iMac Pro uses high-end GPUs from AMD including the Radeon Pro Vega 56 and 64 graphics cards; these parts are packaged using a newer 2.5D technology ...


Intel Custom Foundry EMIB
https://www.intel.com/content/www/us/en/foundry/emib.html

Embedded Multi-die Interconnect Bridge (EMIB) is an elegant and cost-effective approach to in-package high density interconnect of heterogeneous chips. ... Instead of using a large silicon interposer typically found in other 2.5D approaches, EMIB ...
Click to expand...

This Intel-AMD hybrid CPU uses MCM not EMIB.

EDIT: I was wrong.
 
Last edited:
No company will tell you until they have to, especially when working with a rival.
 
ecmaster76 said:
Might not be a lie. If these parts are MCM packages with separate Radeon dies then technically Intel is not licensing the IP but is acting more like an AIB and integrating the part into a board of their own design. Its possible these will use stock Radeon drivers if that's the case.
Click to expand...


If its a separate Radeon dies, I really don't see it fitting into thin and light notebooks ;) Its gotta be a very small Vega ;)

Put it this way Polaris uses as much power as a gtx 1080, and we haven't seen the gtx 1080 in any thin and light notebooks. The highest Pascal chip in thin and light is then 1060? Maybe the 1070. That means if its a separate die, they need to be able to get the power consumption below that of the 1070 in notebooks. And added to the thermal density of the MCM package it has to be much lower. Very small Vega, we are talking about rx 460 level or lower performance.

Yeah the report takes a wild leap of faith to go up against nV, its not that, unless they aren't targeting thin and light notebooks.
 
Last edited:
razor1 said:
If its a separate Radeon dies, I really don't see it fitting into thin and light notebooks ;) Its gotta be a very small Vega ;)

Put it this way Polaris uses as much power as a gtx 1080, and we haven't seen the gtx 1080 in any thin and light notebooks. The highest Pascal chip in thin and light is then 1060? Maybe the 1070. That means if its a separate die, they need to be able to get the power consumption below that of the 1070 in notebooks. And added to the thermal density of the MCM package it has to be much lower. Very small Vega, we are talking about rx 460 level or lower performance.

Yeah the report takes a wild leap of faith to go up against nV, its not that, unless they aren't targeting thin and light notebooks.
Click to expand...

Well, Nvidia have Max-Q 1080s that are in thin-ish light-ish notebooks like the Asus Zephyrus.
 
You must log in or register to reply here.
Back
Top