Intel’s Xeon Platinum 8284 CPU: When 300 MHz Cost $5,500
by Anton Shilov on July 18, 2019 12:00 PM EST- Posted in
- CPUs
- Intel
- Xeon
- Enterprise CPUs
- Servers
- Xeon Platinum
Besides Xeon processors that are officially mentioned on its website and price list, Intel has tens of ‘off roadmap’ server CPUs only available to select customers that have special requests. Recently journalists from ComputerBase discovered that Intel has Xeon Platinum 8284, the company’s fastest 28-core chip for multi-socket servers. The CPU runs 300 MHz faster than the ‘official’ Xeon Platinum 8280, but costs considerably more.
Intel’s Xeon Platinum 8284 packs 28 cores with Hyper-Threading that run at 3.0-4.0 GHz, feature a 38.5 MB cache, a six-channel memory controller supporting up to 1 TB of DDR4-2933 with ECC, 48 PCIe 3.0 lanes, and other capabilities found in codenamed Cascade Lake CPUs. Since the chip runs at 300 MHz higher base frequency when compared to the Xeon Platinum 8280, it has a 240 W TDP, up from 205 W. Meanwhile, Tcase of the CPU (the maximum allowed temperature on the IHS of the processor) was reduced to 65°C (down from 84°C), so the CPU requires a very sophisticated cooling system that can take away 240 W at the aforementioned temperature.
Being Intel’s fastest 28-core CPU for multi-socket servers, the Xeon Platinum 8284 processor costs $15,460 (recommended customer price for 1k unit order, RCP), whereas the Xeon Platinum 8280 that runs at a 300 MHz lower frequency, costs $10,009 for 1ku.
| Intel Second Generation Xeon Scalable Family (Cascade Lake) |
||||||||
| Cores | Base Freq |
Turbo Freq |
L3 Cache |
TDP (W) |
Optane | Price (1ku) |
||
| Xeon Platinum 8200 | ||||||||
| 8284 | 28 | 3.0 | 4.0 | 38.50 | 240 | Yes | $15460 | |
| 8280 | L | 28 | 2.7 | 4.0 | 38.50 | 205 | Yes | $17906 |
| 8280 | M | 28 | 2.7 | 4.0 | 38.50 | 205 | Yes | $13012 |
| 8280 | 28 | 2.7 | 4.0 | 38.50 | 205 | Yes | $10009 | |
| 8276 | L | 28 | 2.2 | 4.0 | 38.50 | 165 | Yes | $16616 |
| 8276 | M | 28 | 2.2 | 4.0 | 28.50 | 165 | Yes | $11722 |
| 8276 | 28 | 2.2 | 4.0 | 38.50 | 165 | Yes | $8719 | |
| 8270 | 26 | 2.7 | 4.0 | 25.75 | 205 | Yes | $7405 | |
| 8268 | 24 | 2.9 | 3.9 | 35.75 | 205 | Yes | $6302 | |
| 8260 | L | 24 | 2.4 | 3.9 | 25.75 | 165 | Yes | $12599 |
| 8260 | M | 24 | 2.4 | 3.9 | 25.75 | 165 | Yes | $7705 |
| 8260 | 24 | 2.4 | 3.9 | 25.75 | 165 | Yes | $4702 | |
| 8260 | Y | 24 | 2.4 | 3.9 | 35.75 | 165 | Yes | $5320 |
| 8256 | 4 | 3.8 | 3.9 | 16.50 | 105 | Yes | $7007 | |
| 8253 | L | 16 | 2.2 | 3.0 | 35.75 | 165 | Yes | ? |
| 8253 | M | 16 | 2.2 | 3.0 | 35.75 | 165 | Yes | ? |
| 8253 | 16 | 2.2 | 3.0 | 35.75 | 165 | Yes | $3115 | |
The Xeon Platinum 8284 is not mentioned in Intel’s pricelist, and not under Cascade Lake on Intel's ARK database, but it is searchable if you know the exact number. This typically means that the CPU is only available to select customers or even a customer. That said, it is possible that apart from higher clocks, this 'semi-custom' off-roadmap processor may come with features that go beyond that and this might explain the huge price difference when compared to the model 8280.
Related Reading
- The Intel Second Generation Xeon Scalable: Cascade Lake, Now with Up To 56-Cores and Optane!
- Intel’s Enterprise Extravaganza 2019: Launching Cascade Lake, Optane DCPMM, Agilex FPGAs, 100G Ethernet, and Xeon D-1600
- Intel Architecture Manual Updates: bfloat16 for Cooper Lake Xeon Scalable Only?
- Intel Xeon Update: Ice Lake and Cooper Lake Sampling, Faster Future Updates
Source: Intel’s ARK (via ComputerBase)
Facebook
Twitter
RSS
81 Comments
View All Comments
psychobriggsy - Thursday, July 18, 2019 - link
I guess AMD will have a 32C Rome SKU that destroys this at a significantly lower price and TDP in the near future.It's only hope is clock speed, but TSMC's 7nm seems optimal to around 3.3GHz before power goes up more than linearly with speed, so if AMD were to make a clock-speed targetting 32C SKU they should be able to.
MonkeyPaw - Thursday, July 18, 2019 - link
I think they could go past 4.0ghz at 32C and still match Intel’s power budget. The 3950x is only 105W, so if AMD wants to play with 215W, they could aim pretty high. I bet most of the best chiplets are already heading to this market to begin with due to better margins.edzieba - Thursday, July 18, 2019 - link
I wouldn't bet on it. While there are power saving son the 7nm 'core' dies, there is also power loss on the IF interconnects (all 8 of them) plus the large not-a-Northbridge die that remains on an older process. We saw with Zen that for high core count Epyc more power was consumed by the uncore rather than the cores themselves (https://www.anandtech.com/show/13124/the-amd-threa...eek2121 - Thursday, July 18, 2019 - link
EPYC Rome is not slated to reach 4.0 GHz if leaks are to be believed. However it's important to note that Intel's boost clocks are likely single core boost, with all core boost being a bit lower. Zen 2 has already been shown to be quite the formidable CPU even at lower clocks. Even with a rumored boost of 3.4 GHz, chances are AMD will have a clear win in the data center. After all, when Intel charges 3X more for 10% more performance...well, as the joke goes: "Nobody ever got fired for buying Xeon...until NOW!"eek2121 - Thursday, July 18, 2019 - link
Intel's TDP listings are not the actual TDP, nor are they power consumed. All IO is handled by the IO Die itself, so AMD can have as few or as many chiplets as it wants. Thus, an 8 core, well binned, EPYC ROME CPU could indeed hit a 4.7 GHz Boost if it only has a single chiplet.Gondalf - Thursday, July 18, 2019 - link
Still they can not, the low turbo of Epyc 2 line say something about TSMC low power process huge issues. On Ryzen 3000 line (performance process) the turbo frequence is obtained with a too high voltage applied to the core, moreover on all cores up the cpu overclock is pretty bad and spread an insane power.It is a fact that at least for now the single thread performance of Epyc 2 line is suboptimal, and Intel beat badly the brand new AMD server line in responsiveness.
A good reason to say that it is not all rosy at Sunnyvale and many critical workloads are right now forbidden to Epyc.
Death666Angel - Thursday, July 18, 2019 - link
You seem to have a lot of great information that is nowhere to be found on the sites I frequent. Where can I see the detailed information on Epyc 2 single thread performance? I have not seen anything about clockspeeds so far. What I have seen is a lot of speculation that AMD is saving the best chiplets for TR and Epyc. Considering that TR2 hit the same peak clocks as Ryzen 2000, the conclusion might be to say that Epyc will hit the same peak clocks as Ryzen 3000. But I can only speculate, you seem to have all the answers.eek2121 - Thursday, July 18, 2019 - link
Epyc "ROME" or EPYC 2 as you are calling it is based on the Zen 2 core. Quite frankly, based on what we have seen thus far, it doesn't matter where clock speeds fall (3.4 GHz is the rumored boost clock, and a bit under 10 grand is the rumored top of the line price point for a 64 core, 128 thread "ROME" CPU). Intel is going to have a hard time selling these things when "ROME" launches regardless. After all, would you spend $18,000 on a 28 core CPU just because single core workloads can boost to 4.0 GHz? Or would you instead spend $6000 on a 64 core/128 thread CPU that only boosts to 3.35 GHz? That's 1/3rd the price for more than 2X more cores and a slight clock speed reduction that matters very little for most workloads that these CPUs are going to be handling. Also, as others have mentioned, AMD could easily hit higher clocks than rumored.remosito - Wednesday, July 24, 2019 - link
> would you spend $18,000 on a 28 core CPU just because single core workloads can boost to 4.0 GHz? Or would you instead spend $6000 on a 64 core/128 thread CPU that only boosts to 3.35 GHz?For our database server, I'd pick neither. I'd get dual 8-12 core highest boost cpus I can get. They are usually (last few gens) in the 3000-4000$ range.
mike_bike_kite - Tuesday, August 6, 2019 - link
"For our database server, I'd pick neither. I'd get dual 8-12 core highest boost cpus I can get. They are usually (last few gens) in the 3000-4000$ range."Wouldn't you do better spending the money on maxing out the RAM and getting as much on fast SSD as you could. Obviously it depends on workload but wouldn't the new 3900X, 64GB RAM and the fastest PCIe4 M.2 SSD's offer better performance for less? I guess it depends where the bottleneck is in modern database servers but my "guess" would be on accessing the data rather than processing it. This isn't advice as I've never spec'ed database servers but I would genuinely be interested in any real world tests out there.