Intel unveils record-breaking $17,800 price for 128-core Xeon 6980P processor

S

Skye Jacobs

Posts: 204   +7
Staff
What just happened? Intel has announced some pretty bold pricing for its Granite Rapids family of server processors after years of trailing AMD in core count and multi-thread performance. Team Blue is determined to reclaim its leadership in this segment of the market, especially as data centers continue to demand more powerful and efficient processors. The pricing is a shot across the bow from Intel in a battle with AMD that will only intensify.

Intel is raising eyebrows in the server processor market with pricing for its latest flagship Xeon 6980P CPU, part of Intel's new Granite Rapids family. Listed at $17,800, it is the most expensive modern CPU in history. In fact, Intel's pricing for the Granite Rapids lineup is notably higher than previous generations – and, as Tom's Hardware notes, is higher than AMD's EPYC Genoa 9654 offering with 96 cores, which costs $11,805.

The 120-core Xeon 6979P costs $15,750, or $131 per core. The 96-core Xeon 6972P is priced at $14,600, which is $2,795 more than AMD's 96-core EPYC 9654, despite having the same core count. Even Intel's 72-core Xeon 6960P, at $13,750, is more expensive than AMD's 96-core chip. The only Intel processor that undercuts AMD's pricing is the 96-core Xeon 6952P, which has lower base clock speeds.

Intel's pricing history, though, reflected the realities of the market. For several years, Team Blue struggled to compete with AMD in terms of the number of cores and multi-threaded performance. As a result, Intel kept its processor prices relatively low during this period. AMD, on the other hand, was focused on increasing its market share against Intel. While AMD's EPYC processors were priced at a premium, they were not excessively expensive compared to Intel's offerings.

However, the Xeon 6980P marks a significant leap in Intel's product lineup and sets a new benchmark for high-end server processors, boasting 128 high-performance cores and 256 threads. This core count surpasses AMD's top-tier EPYC Genoa 9654, which offers 96 cores at a comparatively modest $11,805.

The premium pricing suggests Intel is highly confident in the overall value proposition of its new processors, considering factors like compute density and operational efficiency in data centers. The Xeon 6980P offers a lower cost per core at $139, compared to $181 per core for the previous generation.

The launch of these high-priced CPUs coincides with AMD's release of its Zen 5 EPYC Turin server chips, which is expected this week. AMD's pricing has not been released yet, so it remains to be seen how competitive Intel's offering really is.

Permalink to story:

Intel unveils record-breaking $17,800 price for 128-core Xeon 6980P processor

 
  • Like
Reactions: lucasthomass
When AMD released threadripper and Epyc it was really cheap, they managed to get in many datacentres and workstations. Now they are becoming INTEL ... with chips costing 7K to 11K. I wonder though if there is a bit of a price war.
 
alexnode said:
When AMD released threadripper and Epyc it was really cheap, they managed to get in many datacentres and workstations. Now they are becoming INTEL ... with chips costing 7K to 11K. I wonder though if there is a bit of a price war.
Click to expand...
Let's not put all the blame on AMD. A large reason for these increasing costs is TSMCs monopoly on high end chip manufacturing.
 
Last edited:
  • Like
Reactions: 0dium, Vayven, mosu and 1 other person
This may all easily be a scam, a CPU that's not even ready for production, to prime the company's stock for puchase by Qualcomm.
 
alexnode said:
When AMD released threadripper and Epyc it was really cheap, they managed to get in many datacentres and workstations. Now they are becoming INTEL ... with chips costing 7K to 11K. I wonder though if there is a bit of a price war.
Click to expand...
Yeah who would have thought a new 128 core cost more than a couple years old 32 core EPYC. Ofc they cost more, the process node is more expensive.

What to keep in mind here, is obviously the performance per watt, and here you get whole lot more today than you did few years back.
 
  • Like
Reactions: Vayven
I wonder if this is just a bluff from Intel to try and get AMD to raise their pricing similarly whilst then offering heavily discounted prices themselves to interested parties.

Afaik these server CPUs are basically never sold without negotiating. If they just put out silly prices whilst selling much lower they might get AMD to announce at high prices as well.
Result: They're both getting a fatter margin (something Intel could really use ATM).
Not exactly price fixing so it's not illegal.

I would say it has worked for Nvidia in the consumer market but I don't think it went like that there. AMD is barely on NVIDIAs radar and AMDs silly launch prices are due to their inept marketing department.
Duopolies with a high barrier of entry are not much better than monopolies for the market.
 
  • Like
Reactions: 0dium and mosu
“it is the most expensive modern CPU in history”. Oh contrary mong bong amee, IBM’s Mainframe CPU’S are very modern indeed and cost a great deal more.
 
HardReset said:
Intel's power consumption will no doubt be much higher so this is no problem for AMD.

It has crap cores. These ones are real deal, at least.
Click to expand...
AMD is going Zen 5C for enterprise very soon too.
You obviously don't know much about enterprise loads.

Even ARM based chips gain enterprise-marketshare and is expected to have 15-20% by 2030.

You know nothing about Intels power consumption on this new stuff and especially Clearwater Forest looks very good.

Intel has leading power consumption with Lunar Lake. Beats both AMD and ARM with ease.

You always ramble about AMD being superior in enterprise, yet Intel has higher enterprise revenue every quarter. Just because AMD is not totally garbage anymore, does not mean they are outselling Intel. Look at their recent financial reports for proof. This stuff is not secret you know.
 
yRaz said:
Let's not put all the blame on AMD. A large reason for these increasing costs is TSMCs monopoly on high end chip manufacturing.
Click to expand...
Yes you are right, let's not forget the AI hype train... now pressure cookers need to be AI. Nvidia is having probably 60% of all TSMC in full production mode. Just think that in a five years all these will need to be sold on ebay.
 
yRaz said:
Let's not put all the blame on AMD. A large reason for these increasing costs is TSMCs monopoly on high end chip manufacturing.
Click to expand...
And the funny part about this is many people on this forum hope Intel will go bankrupt and then TSMC will have close to full monopoly, raising prices even further (they did that many times in the last 5-10 years).

This will affect AMD too. AMD can't even afford to use peak TSMC nodes today, for example. Apple, Intel and Nvidia will use TSMC 3nm before AMD.

Also, TSMC is (mostly) in Taiwan. China could enter Taiwan whenever they want and then chip production is on hold. This is why all tech companies want chip production out of asia. TSMC most used (and most advanced) fabs are still in Taiwan, even tho they built a few fabs outside. TSMC Arizona fab is "only" 5nm (N4) and is not even running yet, for example.

At least Intel 18A looks to be working, and is way more advanced than Samsung 2nm (which is more like a 5-7nm process in terms of density anyway). Samsung always lied about their processes, to make them sound better than they actually are and lure in customers.

Remember RTX 3000 series on Samsung 8nm? Samsung just renamed their 10nm node for this. It was called 10nm in the beginning and it was worse than TSMC 12nm to begin with. Their "8nm" node is 100% identical to their 10nm node. It was cheap and available tho, which is why Nvidia used it. AMD insisted on TSMC, which was heavily overbooked at the time.

TSMC, Intel and Samsung is the only ones that can make advanced chips. Global Foundries is way behind and could not even hit 7nm, still cant and probably never will do better than 12nm. China talks about 5nm but only really did working 28nm AFAIK.

Intel 18A is around TSMC N2 / 2nm level in terms of density.
Intel will also be first with backside power aka PowerVia.

What Intel needs to do, is splitting chip design business with fab business and start taking in customers. I expect this to happen in the next 12 months.
 
Last edited:
  • Like
Reactions: yRaz
Rapob said:
And the funny part about this is many people on this forum hope Intel will go bankrupt and then TSMC will have close to full monopoly, raising prices even further (they did that many times in the last 5-10 years).

This will affect AMD too. AMD can't even afford to use peak TSMC nodes today, for example. Apple, Intel and Nvidia will use TSMC 3nm before AMD.

Also, TSMC is (mostly) in Taiwan. China could enter Taiwan whenever they want and then chip production is on hold. This is why all tech companies want chip production out of asia.

At least Intel 18A looks to be working, and is way more advanced tham Samsung 2nm (which is more like a 5-7nm process in terms of density anyway). Samsung always lied about their processes really.

Remember RTX 3000 series on Samsung 8nm? Samsung just renamed this node to 8nm, it was called 10nm in the beginning and it was worse than TSMC 12nm to begin with.

TSMC, Intel and Samsung is the only ones that can make advanced chips.

Intel 18A is around TSMC N2 / 2nm level in terms of density.
Intel will also be first with backside power aka PowerVia.

What Intel needs to do, is splitting chip design business with fab business and start taking in customers.
Click to expand...
I don't claim to understand how Intel went from a monopoly in CPU chip manufacturing to be in financial ruin, but one thing I can tell for sure is loads of shareholders did very well in the past four decades. Intel didn't push the envelop for at least 10 years, they didn't manage to fight AMD, they didn't invest, and now they are in trouble. They should have been the first to jump onto RISC processors 15 years ago, they should have been the first to jump to 8 and 12 cores first, to make a threadripper for workstations, they should have put billions on their compute Xeon Phi card. But when you need to put all your profits in investment, or even borrow, then shareholders get jittery share price goes down etc.
 
Rapob said:
And the funny part about this is many people on this forum hope Intel will go bankrupt and then TSMC will have close to full monopoly, raising prices even further (they did that many times in the last 5-10 years).

This will affect AMD too. AMD can't even afford to use peak TSMC nodes today, for example. Apple, Intel and Nvidia will use TSMC 3nm before AMD.

Also, TSMC is (mostly) in Taiwan. China could enter Taiwan whenever they want and then chip production is on hold. This is why all tech companies want chip production out of asia. TSMC most used (and most advanced) fabs are still in Taiwan, even tho they built a few fabs outside. TSMC Arizona fab is "only" 5nm (N4) and is not even running yet, for example.

At least Intel 18A looks to be working, and is way more advanced than Samsung 2nm (which is more like a 5-7nm process in terms of density anyway). Samsung always lied about their processes, to make them sound better than they actually are and lure in customers.

Remember RTX 3000 series on Samsung 8nm? Samsung just renamed their 10nm node for this. It was called 10nm in the beginning and it was worse than TSMC 12nm to begin with. Their "8nm" node is 100% identical to their 10nm node. It was cheap and available tho, which is why Nvidia used it. AMD insisted on TSMC, which was heavily overbooked at the time.

TSMC, Intel and Samsung is the only ones that can make advanced chips. Global Foundries is way behind and could not even hit 7nm, still cant and probably never will do better than 12nm. China talks about 5nm but only really did working 28nm AFAIK.

Intel 18A is around TSMC N2 / 2nm level in terms of density.
Intel will also be first with backside power aka PowerVia.

What Intel needs to do, is splitting chip design business with fab business and start taking in customers. I expect this to happen in the next 12 months.
Click to expand...
Well, so much of the nodes these day are less about the actual feature size and more about transistor density and FIN height while balancing yields. We're in the middle of "FIN depopulation" where we try to make them taller so they can handle more current while also making then smaller. If we go from the typical 6-8 FINs a transistor down to 4, we can drastically increase densities. TSMC even has plans to take it as low as 3 with their FINFLEX nodes. We probably aren't going to hit a hardwall in Moores Law until GAA comes around in ~2030.

The worst part about GAA is that that is the next big step and Intel has been the only one investing it it heavily since the early 2010s but nothing has made it to production yet.

I also don't see why everyone wants Intel to fail. It's like the nVidia fanboys who want to see AMD fail. How does that benefit anyone?

Frankly, I haven't seen much of a need for cutting ead performance parts for years now. Everyone talks about how nVidia is the best because they have the 4090. Meanwhile, only about 0.3% of gamers actually have 4090s. To put that in perspective, that's 3 per 1000.

I want to see more compelling products in the $3-500 range
 
yRaz said:
Well, so much of the nodes these day are less about the actual feature size and more about transistor density and FIN height while balancing yields. We're in the middle of "FIN depopulation" where we try to make them taller so they can handle more current while also making then smaller. If we go from the typical 6-8 FINs a transistor down to 4, we can drastically increase densities. TSMC even has plans to take it as low as 3 with their FINFLEX nodes. We probably aren't going to hit a hardwall in Moores Law until GAA comes around in ~2030.

The worst part about GAA is that that is the next big step and Intel has been the only one investing it it heavily since the early 2010s but nothing has made it to production yet.

I also don't see why everyone wants Intel to fail. It's like the nVidia fanboys who want to see AMD fail. How does that benefit anyone?

Frankly, I haven't seen much of a need for cutting ead performance parts for years now. Everyone talks about how nVidia is the best because they have the 4090. Meanwhile, only about 0.3% of gamers actually have 4090s. To put that in perspective, that's 3 per 1000.

I want to see more compelling products in the $3-500 range
Click to expand...
With TSMC driving up prices, it will be harder and harder to deliver something good in the 300-500 range.

Lets wait and see how RDNA4 does tho. Performance-wise, not much will happen. We will be looking at 7900GRE / 4070 SUPER performance levels tops but if they can keep the price below 500 dollars and improve FSR, then it might be good enough.

I personally don't just look at raster performance in soon to be 2025. I look at overall package, drivers, features like upscaling, downsampling and even RT performance as it will matter more and more going forward, and AMD is pretty much behind on all areas here.
 
Rapob said:
With TSMC driving up prices, it will be harder and harder to deliver something good in the 300-500 range.

Lets wait and see how RDNA4 does tho. Performance-wise, not much will happen. We will be looking at 7900GRE / 4070 SUPER performance levels tops but if they can keep the price below 500 dollars and improve FSR, then it might be good enough.

I personally don't just look at raster performance in soon to be 2025. I look at overall package, drivers, features like upscaling, downsampling and even RT performance as it will matter more and more going forward, and AMD is pretty much behind on all areas here.
Click to expand...
I'd be happy with 7800xt performance for $300. I still see RT performance being another full generation away. I'm mostly concerned 4k120raster performance in the games I play, which is mostly EvE and ESO. I also don't like how raytracing is all or nothing these days. There is still lots of room for optimization on the developer side for raytracing.

And I'm not against raytracing, I just don't see the value in paying $1000+ for it.
 
yRaz said:
I'd be happy with 7800xt performance for $300. I still see RT performance being another full generation away. I'm mostly concerned 4k120raster performance in the games I play, which is mostly EvE and ESO. I also don't like how raytracing is all or nothing these days. There is still lots of room for optimization on the developer side for raytracing.

And I'm not against raytracing, I just don't see the value in paying $1000+ for it.
Click to expand...

The problem for AMD is that more and more games start including RT elements, forced RT that is and performance hit on AMD cards are substancial.

Example, Avatar which was even AMD sponsored:

www.techpowerup.com

Avatar: Frontiers of Pandora Performance Benchmark Review - 30+ GPUs Tested

Avatar: Frontiers of Pandora features stunning visuals that recreate the movie franchise's unique universe. There's also support for AMD FSR 3 Frame Generation and NVIDIA DLSS. In our performance review, we're taking a closer look at image quality, VRAM usage, and performance on a wide selection...
www.techpowerup.com www.techpowerup.com

Avartar is confirmed to use raytraced reflections and shadow.

AMD has officially said that RT has been a focus on RDNA4. RT will be improved with 8000 series. AMD can't just ignore RT because some games don't use it yet. More and more games will use it, because developers can't wait to not being forced to do fake lighting in their games. RT and Path Tracing will do perfect lighting, how it is actually supposed to look and will work dynamicly with the environment. Pre-baked lighting don't.

In 5-10 years, RT will be used in most games. PS5 Pro had a massive RT uplift too, some people think it's got some RDNA4 tech in it. Still behind Nvidia, but at least an improvement.

RDNA4 won't deliver 7800XT performance for 300 dollars. You will get 7900GRE performance or so, with the top SKU which is predicted to be 400-500 dollars.
 
Rapob said:
AMD is going Zen 5C for enterprise very soon too.
You obviously don't know much about enterprise loads.
Click to expand...

Zen5c has same IPC as Zen5. Intel Crap cores IPC is much worse.

Rapob said:
Even ARM based chips gain enterprise-marketshare and is expected to have 15-20% by 2030.

You know nothing about Intels power consumption on this new stuff and especially Clearwater Forest looks very good.

Intel has leading power consumption with Lunar Lake. Beats both AMD and ARM with ease.
Click to expand...

FYI, Lunar Lake has basically Nothing to do with server chips Well, same core architecture but that's it.

In other words, no matter what Lunar Lake is ir isn't, won't matter here.

Rapob said:
You always ramble about AMD being superior in enterprise, yet Intel has higher enterprise revenue every quarter. Just because AMD is not totally garbage anymore, does not mean they are outselling Intel. Look at their recent financial reports for proof. This stuff is not secret you know.
Click to expand...

Who cares about outSelling? AMD server CPUs outperform Intel parts by wide margin. No wonder AMD gains market share and Intel loses.
 
HardReset said:
Zen5c has same IPC as Zen5. Intel Crap cores IPC is much worse.



FYI, Lunar Lake has basically Nothing to do with server chips Well, same core architecture but that's it.

In other words, no matter what Lunar Lake is ir isn't, won't matter here.



Who cares about outSelling? AMD server CPUs outperform Intel parts by wide margin. No wonder AMD gains market share and Intel loses.
Click to expand...

Zen 5C is only 25% smaller than a regular Zen 5 core and performance will be lower due to less/no cache.

Intel can put 4 e-cores for every 1 p-core in comparison. That approach will win every time. 4-e cores beats 1-p core in 99% of enterprise tasks. Pretty much every workload is multithreaded to perfection.

Both Intel and AMD cares about selling, it is what they do and how they make money. Intel sold more enterprise products than AMD in the last quarter. Even in its current state, Intel outsells AMD.

Once again, even ARM is gaining traction in server/enterprise, around 10-12% marketshare now, looking to hit 20% in 2023. Tons of workloads here needs lots of cores, not speed.
 
Rapob said:
Zen 5C is only 25% smaller than a regular Zen 5 core and performance will be lower due to less/no cache.

Intel can put 4 e-cores for every 1 p-core in comparison. That approach will win every time. 4-e cores beats 1-p core in 99% of enterprise tasks. Pretty much every workload is multithreaded to perfection.
Click to expand...
Zen5c performance is identical against Zen5 with same cache and clock speed.

Intel's approach loses because Crap core IPC is much lower than Zen5 IPC. Also most of that "area reduction" comes from fact that E-cores have much slower memory access and have less cache also. Because Zen5 CCD only takes around 70 mm2 and around half of that is L3 cache, die area is not an issue. Heck, single Zen5 core takes less than 5 mm2 without L3 cache. Without L2 cache, even less.

99% enterprise tasks, again you have no idea what you are talking about. If you were right, Intel would have flooded market with CPUs that have 500+ tiny cores. Unsuprisingly we haven't really seen that.
Rapob said:
Both Intel and AMD cares about selling, it is what they do and how they make money. Intel sold more enterprise products than AMD in the last quarter. Even in its current state, Intel outsells AMD.

Once again, even ARM is gaining traction in server/enterprise, around 10-12% marketshare now, looking to hit 20% in 2023. Tons of workloads here needs lots of cores, not speed.
Click to expand...
Intel sells because some stupid customers are stuck with Intel. But Intel already lost much of market share and will lose more. Soon Intel will have less share than AMD.

Intel got server market share over 90% mostly with high performance cores. Not tiny ones. You are wrong as always. ARM is getting market share because some companies make their own cores, not because ARM is actually better.
 
HardReset said:
Zen5c performance is identical against Zen5 with same cache and clock speed.

Intel's approach loses because Crap core IPC is much lower than Zen5 IPC. Also most of that "area reduction" comes from fact that E-cores have much slower memory access and have less cache also. Because Zen5 CCD only takes around 70 mm2 and around half of that is L3 cache, die area is not an issue. Heck, single Zen5 core takes less than 5 mm2 without L3 cache. Without L2 cache, even less.

99% enterprise tasks, again you have no idea what you are talking about. If you were right, Intel would have flooded market with CPUs that have 500+ tiny cores. Unsuprisingly we haven't really seen that.

Intel sells because some stupid customers are stuck with Intel. But Intel already lost much of market share and will lose more. Soon Intel will have less share than AMD.

Intel got server market share over 90% mostly with high performance cores. Not tiny ones. You are wrong as always. ARM is getting market share because some companies make their own cores, not because ARM is actually better.
Click to expand...
Haha no its not.
Zen 5C is not performing the same as Zen 5 when its 25% smaller. Obvious really.

AMD never said Zen 5C will get the performance of regular Zen 5. They pretty much say it won't if you read their statements and understand them. They will deliver "enough power" meaning less.

I sell enterprise stuff b2b, I think I know what most companies want but you can keep dreaming, sure.

Also, you never responed on ARM gaining server marketshare, and did for years. You are out of touch with what actual companies want. ARM is soon at 15% and rising, will be around 20% in 2030. Way more efficient than AMD and Intel and does optimized tasks just as well.

If AMD was better for everything, they would dominate enterprise, but they don't. Intel has had higher data center revenue than AMD in both Q1 and Q2 this year. Go read financial reports instead of making up stuff.

Also, most companies spend way more buying actual laptops than servers and Intel completely dominate here, with 90% marketshare or so.
 
Last edited:
Rapob said:
Haha no its not.
Zen 5C is not performing the same as Zen 5 when its 25% smaller. Obvious really.

AMD never said Zen 5C will get the performance of regular Zen 5. They pretty much say it won't if you read their statements and understand them. They will deliver "enough power" meaning less.
Click to expand...
With same cache, same clock speed and same IPC Zen5c is identical to Zen5.

Of course c-version will not achieve as high clock speeds as regular Zen5 but on servers that hardly matters too much. Obvious. Really.
Rapob said:
I sell enterprise stuff b2b, I think I know what most companies want but you can keep dreaming, sure.

Also, you never responed on ARM gaining server marketshare, and did for years. You are out of touch with what actual companies want. ARM is soon at 15% and rising, will be around 20% in 2030. Way more efficient than AMD and Intel and does optimized tasks just as well.

If AMD was better for everything, they would dominate enterprise, but they don't. Intel has had higher data center revenue than AMD in both Q1 and Q2 this year. Go read financial reports instead of making up stuff.

Also, most companies spend way more buying actual laptops than servers and Intel completely dominate here, with 90% marketshare or so.
Click to expand...
So you say enterprises want many small cores and still Intel dominated market with not so many high performance cores. Yeaaah.

ARM is not "more efficient" than Intel or AMD. It's just that ARM does not need x86 license, nothing else.

As you are selling b2b stuff, you also know that better product basically means nothing. Because some companies only sell Intel and buyers do not want to change to another seller, so they buy whatever crap Intel has to offer.

Oh, so Intel losing SERVER market share is not reason why Intel is in trouble 🤦‍♂️
 
HardReset said:
With same cache, same clock speed and same IPC Zen5c is identical to Zen5.

Of course c-version will not achieve as high clock speeds as regular Zen5 but on servers that hardly matters too much. Obvious. Really.

So you say enterprises want many small cores and still Intel dominated market with not so many high performance cores. Yeaaah.

ARM is not "more efficient" than Intel or AMD. It's just that ARM does not need x86 license, nothing else.

As you are selling b2b stuff, you also know that better product basically means nothing. Because some companies only sell Intel and buyers do not want to change to another seller, so they buy whatever crap Intel has to offer.

Oh, so Intel losing SERVER market share is not reason why Intel is in trouble 🤦‍♂️
Click to expand...
Yes ARM is more efficient. That is the whole reason why ARM is used really. You again know nothing about ARM servers. They are 15-25% more efficienct and gain marketshare as we speak, did for years and is expected to have 20% marketshare by 2030.

Zen 5C won't have same performance as Zen 5. Core is 25% smaller, do you think they magically made it 25% smaller without loosing performance? AMD never said performance would be the same. Not once.

I know exactly what companies want, you don't.

Intel outsold AMD in data center both quarters this year. They will do it again in Q3. Look up financials if in doubt.

Intel outsold AMD like 9 to 1 in Laptops and Mobile.

Intel even outsold AMD in desktop. Intel gains marketshare: https://store.steampowered.com/hwsurvey/processormfg/

So yeah, you can dream, reality is what I look at.
 
You must log in or register to reply here.

Similar threads

zohaibahd
Benchmarks tease upcoming Intel Granite Rapids 120-core, 240-thread monster CPU
Replies
0
Views
89
zohaibahd
zohaibahd
zohaibahd
Intel accidentally leaks monstrous 9,324-pin socket for "Diamond Rapids" Xeon CPUs
Replies
19
Views
311
HardReset
H
D
Intel refutes AMD's claims, says its 5th-gen Xeon is faster than Epyc Turin in AI workloads
Replies
5
Views
235
redgarl
R

Latest posts

Back