Intel is using heavy discounts on Xeon CPUs to stop AMD from eating its server lunch

HardReset

Posts: 1,265   +926
So you've bought one because they are so cheap? That is still speculation.
I have no need for those. Speculation?
And your subjective observations are still speculation.
Claim was that AMD would do same as Intel if AMD had like 90+ market share. So far nobody has shown any evidence about that.
Hmm. Did I say that? If so, where??? Or are you still speculating at what I've said.
Said it here:
With the prices AMD slapped on their latest Threadripper offerings, I do not see the speculative nature of the assessment.
You say prices are problem, OK. If AMD didn't slap so high price tag for TR CPU's (read: no 64-core model = top end CPU costs less), there would be lower prices and less problems. Simple.
EPYC and Threadripper are two different classes of processors with differing specs, such as PCI-e lanes. Perhaps you should figure out something yourself?
EPYC and Threadripper CPU's use exactly same CPU cores and IO-chip. About only differences aside core count are memory channels and PCIe lanes (TR has only half) BUT TR CPU's also have higher clock speeds. On some scenarios (other than heavy memory/IO), Threadrippers are much faster. And definitely they are not expensive considering core count and raw speed vs server versions.
Charity. :facepalm: šŸ¤£ More speculation?
It feels like AMD should give CPU's away for free.
https://www.cpu-world.com/CPUs/K8/AMD-Athlon 64 FX-62 - ADAFX62IAA6CS (ADAFX62CSBOX).html In case you cannot find it, look for the "price at introduction" field in page. Of course, that only lasted until sIntel came out with Core 2 and topped AMD in price.
At that time (and many years before) both Intel and AMD priced "best desktop CPU" around 999 dollars or more. AMD couldn't price it much lower than Intel's offering, otherwise people would have bought Intel because "more expensive = better". Btw, that CPU was faster than Intel's offering at same time.

I just don't understand what that proves, since "top notch" Athlon 1 GHz on 2000 cost even several hundred dollars more. It proves that AMD asked around 1000 dollars for top notch CPU at that time but that was common practice for around 6-7 years.
I think that you should study CPU pricing history. The practice of continually pushing prices higher and higher between sIntel/AMD slugs is, as I see it, actually enabling each other to continually price CPUs at introduction well beyond the competition and it is doing nothing to keep CPU prices affordable. sIntel pushing their previous gen parts to $2K without any corresponding performance increase above the next lower binned parts made that $2K part not worth it in any sense of the word unless you wanted to count bragging rights as part of performance.
Of course prices get higher. New manufacturing processes are much more expensive than previous ones and will get increasingly more expensive. It means products will get more expensive too. Customers just need to get used to it.

Intel's pricing is another thing but this discussion was about AMD's pricing.
If you don't like my opinion on the price wars, I cannot help that. These price wars do nothing to keep enthusiast CPU prices within reach of many enthusiasts. The only enthusiasts that can afford those prices are those with deep pockets who purchase for bragging rights only. As I see it, they are the market run wild.

If sIntel manages to come out with a new line of CPUs that top AMD by a wide margin, watch the prices skyrocket again.
Like I said, prices will get higher as it's more and more expensive to manufacture top end products. That alone is enough to raise prices. Add pandemic to equation and were f*'d.
By far, I am not against competition in the marketplace - AMD finally put a leash around sIntel's neck that stopped them from fleecing their customers and releasing piddling performance increases from generation to generation. Poor sIntel will have to innovate again to regain that lead, otherwise, AMD will take it all away from them.

I remember people P&Ming about the fact that the first generation of Ryzen's were still not good for gaming. Personally, I think AMD did the right thing because they were targeting the enterprise market - which is where the money is. The gaming market is a niche market, IMO, in comparison to the enterprise market.
Agreed.
And now that AMD is finally producing CPUs that perform well in games compared with sIntel, there's this whole cheering section chanting "Yeah AMD. Keep raising CPU prices - even though we cannot afford them. We'll glady march to our banks and borrow the money so we can buy them."
AMD does raise prices but also offer huge performance gains vs what Intel did in past. Fact that manufacturing costs raise mean more expensive products. Those who are not willing to pay top dollar, should stick with products manufactured on older (and less expensive) processes...

Those are clueless executives doing that.

They're not as smart as the engineers but they're definitely wiser than the average consumer looking to buy a brand-in-a-box PC. They have performance per watt drilled into their heads along with long-term cost analyses. They'll know that the Xeons still aren't worth it.
Whoever they are, they are not any wiser than average PC enthusiast. And totally unable to do long term analysis (or just doesn't care). Like this:

Server dudes: Intel asks too much for server CPU's
AMD: We have superior server CPU available (vs Intel)
Server dudes: Sorry but it's not Intel, we have contracts with Dell and they won't supply AMD, Intel is more reliable (heavier server stays better in place) and much more effective (on heat output at least) etc etc blah blah excuse nr 9999999
AMD: FU then, we're out of server business
Server dudes: Intel asks too much for server CPU's

That wasn't too hard to predict, wasn't it?
 

NeoMorpheus

Posts: 883   +1,678
Those are clueless executives doing that.

They're not as smart as the engineers but they're definitely wiser than the average consumer looking to buy a brand-in-a-box PC. They have performance per watt drilled into their heads along with long-term cost analyses. They'll know that the Xeons still aren't worth it.
Sadly, in my case, the company I work for (multibilions, international, etc) have some old school techs that dont know or care about that, they only care that is a Dell server and has a Xeon.

The only reason why my division has some Ryzen devices is because its a bit independent, money wise, compared to the mother ship.

If not, reports would not show one AMD device.
 

HardReset

Posts: 1,265   +926
Sadly, in my case, the company I work for (multibilions, international, etc) have some old school techs that dont know or care about that, they only care that is a Dell server and has a Xeon.

The only reason why my division has some Ryzen devices is because its a bit independent, money wise, compared to the mother ship.

If not, reports would not show one AMD device.
This is exactly what I meant when I said about not so wise server buyers...
 

Avro Arrow

Posts: 1,800   +2,155
TechSpot Elite
I have no need for those. Speculation?

Claim was that AMD would do same as Intel if AMD had like 90+ market share. So far nobody has shown any evidence about that.

Said it here:

You say prices are problem, OK. If AMD didn't slap so high price tag for TR CPU's (read: no 64-core model = top end CPU costs less), there would be lower prices and less problems. Simple.

EPYC and Threadripper CPU's use exactly same CPU cores and IO-chip. About only differences aside core count are memory channels and PCIe lanes (TR has only half) BUT TR CPU's also have higher clock speeds. On some scenarios (other than heavy memory/IO), Threadrippers are much faster. And definitely they are not expensive considering core count and raw speed vs server versions.

It feels like AMD should give CPU's away for free.

At that time (and many years before) both Intel and AMD priced "best desktop CPU" around 999 dollars or more. AMD couldn't price it much lower than Intel's offering, otherwise people would have bought Intel because "more expensive = better". Btw, that CPU was faster than Intel's offering at same time.

I just don't understand what that proves, since "top notch" Athlon 1 GHz on 2000 cost even several hundred dollars more. It proves that AMD asked around 1000 dollars for top notch CPU at that time but that was common practice for around 6-7 years.

Of course prices get higher. New manufacturing processes are much more expensive than previous ones and will get increasingly more expensive. It means products will get more expensive too. Customers just need to get used to it.

Intel's pricing is another thing but this discussion was about AMD's pricing.

Like I said, prices will get higher as it's more and more expensive to manufacture top end products. That alone is enough to raise prices. Add pandemic to equation and were f*'d.

Agreed.

AMD does raise prices but also offer huge performance gains vs what Intel did in past. Fact that manufacturing costs raise mean more expensive products. Those who are not willing to pay top dollar, should stick with products manufactured on older (and less expensive) processes...


Whoever they are, they are not any wiser than average PC enthusiast. And totally unable to do long term analysis (or just doesn't care). Like this:

Server dudes: Intel asks too much for server CPU's
AMD: We have superior server CPU available (vs Intel)
Server dudes: Sorry but it's not Intel, we have contracts with Dell and they won't supply AMD, Intel is more reliable (heavier server stays better in place) and much more effective (on heat output at least) etc etc blah blah excuse nr 9999999
AMD: FU then, we're out of server business
Server dudes: Intel asks too much for server CPU's

That wasn't too hard to predict, wasn't it?
I agree that it used to be like that but that was before Opteron. Opteron's performance, economy and efficiency was so great that it's what made Michael Dell himself break down and stop using Intel's illegal "Loyalty Programme".

If Opteron was good enough to crack Michael Dell, the man whom Intel referred to as "The best friend that money could buy", then AMD's reputation in the server space isn't nearly as weak as you think it is.

Zen has been out and proving itself for four years now, an eternity in the tech world. The fact that it was designed by Jim Keller gave it immediate credibility. This is because Keller is considered to be the greatest IC architect on the planet in many high-up circles. Let's face it, this man is the father of AMD64 upon which ALL 64-bit x86 CPUs are based. There are no greater achievements in the PC world today than the credentials sported by Jim Keller and that's no exaggeration. If Keller did it, it must be good because he wouldn't accept anything less of himself and people know it.

Sure, there probably are some ignorant fools out there because you'll always find people doing a job they suck at but I think that the norm has become "I don't care what it says on the chip as long as it delivers the results" and, thus far, EPYC has delivered in spades.
 

HardReset

Posts: 1,265   +926
I agree that it used to be like that but that was before Opteron. Opteron's performance, economy and efficiency was so great that it's what made Michael Dell himself break down and stop using Intel's illegal "Loyalty Programme".

If Opteron was good enough to crack Michael Dell, the man whom Intel referred to as "The best friend that money could buy", then AMD's reputation in the server space isn't nearly as weak as you think it is.
Problem is that Michael Dell does not make purchases, just enables people to do them. Better than nothing but still.
Zen has been out and proving itself for four years now, an eternity in the tech world. The fact that it was designed by Jim Keller gave it immediate credibility. This is because Keller is considered to be the greatest IC architect on the planet in many high-up circles. Let's face it, this man is the father of AMD64 upon which ALL 64-bit x86 CPUs are based. There are no greater achievements in the PC world today than the credentials sported by Jim Keller and that's no exaggeration. If Keller did it, it must be good because he wouldn't accept anything less of himself and people know it.

Sure, there probably are some ignorant fools out there because you'll always find people doing a job they suck at but I think that the norm has become "I don't care what it says on the chip as long as it delivers the results" and, thus far, EPYC has delivered in spades.
Keller was not lead architect for Zen and even didn't have any effect on Zen2 or later ones. Also Zen was not as clean sheet as some think (branch predictor and AVX unit were both very close to predecessors). Of course Keller name gives Some credibility but end product is what matters most, Keller or no Keller.

There are still too many "we need to use Intel because (add some lame excuses here) people but yes, good things are happening.