• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

9900KS...

technology moves forward and the cpu is how old now? When intel sold the most cpu there completion was the 2700x. Fast forward to now and amd have a much stronger line up vs the 9900k a 12 core cpu and a 16 core (coming) cpu.

now whilst it might seem like value to someone who paid paid a lot more for a binned chip for the main stream it doesn’t and pricing the cpu at over the gauged price of a amd 3900x to me pointless and just another marketing stunt to tide of intel till the 10 series comes out.

to me this is intel need to drop prices and they will a lot more then trying to get people to pay over the ods for the 9900ks move the 9900k to price against ryzen 7 3800x a d drop the price of i7 to just under 3700x and bring the ks in at 450 or 480 pounds and the value to performance perspective of intel changes dramatically

that’s why I think the pricing for i9 doesn’t make sense

I feel like the i9 is not nearly manly enough!! A concept I have shared in other threads which has gained absolutely no following and frankly shouldn't, in fact even typing it in here will upset the masses... hear me out though as it applies to pretty much all Intel hedt and the high end 9900k/s.

- isnt man size, is a small and feminine CPU of diminishing proportions. (- man points)
- Doesnt come with a tool for torquing the bad boy down. (- man points)
- Doesnt have manly name, in fact the new iteration (ks) subjectively sounds even less manly than just 9900k where as the competition *Ryzen (sounds like some sort of ninja) & Threadripper (Say no more, that **** is literally ripping through threads, you can't argue, it's right there in the name along with stuff like heavy metal and ****).

There is more to this I am sure. If you bought only on man maths there is no option other than threadripper. Otherwise the ks is a decent 40 ish ghz cpu (lol who's arguing, if that's relevant then so is my man maths.)

I wonder how many people bought threadripper with the name being the final deciding factor?
 
Last edited:
LOL :rolleyes:



My 3900X is 4.4 x 6 + 4.6 x 2 + 4.625 + 4.57x2 + 4.55 = 53.915Ghz give or take depending boost :D

64 thread epyc shows up In esxi as 75ghz. :eek::cool:

That's also not "man size"... 1st gen TR 1950x, 4.1x16 so 65.6, at which point I should point out my man maths post above... With man maths involved there is only 1 option. Even at a conservative 3.7 it is still 59.2 gigawatts.
 
Last edited:
Depends what epyc and what base clock, although my experience with esxi would indicate it hasnt a clue and you could try and run it on a donut and it would come back with "dual core" or something.
 
Depends what epyc and what base clock, although my experience with esxi would indicate it hasnt a clue and you could try and run it on a donut and it would come back with "dual core" or something.

I have 3x 7452. The esxi screenshots are in the EPYC thread. I won't lie I bought them using real maths.
 
1450 MHz GDDR is quad-data rated 5.8 Gbps.

You can't multiply the clock of a CPU by its number of units and put GHz as the unit of the measurement, you must use some type of theoretical throughput value like the above...

See the above, 1450 MHz GDDR5 can achieve 92 GB/s maximum throughput, with a 128-bit memory bus.
 
1450 MHz GDDR is quad-data rated 5.8 Gbps.

You can't multiply the clock of a CPU by its number of units and put GHz as the unit of the measurement, you must use some type of theoretical throughput value like the above...

See the above, 1450 MHz GDDR5 can achieve 92 GB/s maximum throughput, with a 128-bit memory bus.

Why not? You can do whatever you want! - VMWare do and they know a thing or two about hardware! Here are two screenshots to illustrate the point, ive even taken the opportunity to highlight it for you :) (one intel one amd):




You say you can't just do that... Well they kinda did and you know what i'm not even complaining. :cool:
 
Last edited:
technology moves forward and the cpu is how old now? When intel sold the most cpu there completion was the 2700x. Fast forward to now and amd have a much stronger line up vs the 9900k a 12 core cpu and a 16 core (coming) cpu.


that’s why I think the pricing for i9 doesn’t make sense

it makes sense to me because 6 cores is more than enough for my needs and the quicker i can make those six cores go, the better.

a point has come where i look at the amd lineup and whilst i am very happy for the competition they are giving Intel, i think the state of the art remains the 8086k, simply because i don’t need more than 6 cores and out of the box, the 8086k gives me those six cores faster than anything else.
 
Why not? You can do whatever you want! - VMWare do and they know a thing or two about hardware! Here are two screenshots to illustrate the point, ive even taken the opportunity to highlight it for you :) (one intel one amd):

You say you can't just do that... Well they kinda did and you know what i'm not even complaining. :cool:

Because seriously, what matters is number of transistors and instruction per cycle (IPC).
You can have a single transistor running at 75 GHz, and it will have the same capacity, right?
 
1450 MHz GDDR is quad-data rated 5.8 Gbps.

You can't multiply the clock of a CPU by its number of units and put GHz as the unit of the measurement, you must use some type of theoretical throughput value like the above...

See the above, 1450 MHz GDDR5 can achieve 92 GB/s maximum throughput, with a 128-bit memory bus.

I don't do VMWARE ESX sizing anymore but its an art form as you need to really know your kit. The server manufacturers will give you the total GHZ rating based on their all core boost targets. You really needed to dig into the manuals of the blades you were running to find them. Cisco blades were a good example of this.

I would imagine that with all the newer Intel CPU bins they introduced that this became a nightmare in the past few years as there must be around 40 Xeon SKUs.

Pain in the neck but that is how it is done.

The only time I used to use IPC/throughput modelling, as you have pointed out, is when i was going from one CPU architecture to another, usually from SPARC.

In this case you are both right, the former GHZ total is a blunt instrument for VMware sizing and the more delicate throughput measurements needed for high-throughput, latency sensitive modelling.

It is a shame as much of this art-form has been thrown in the bin with the transition to public cloud, you are given a shopping list to select from and that's your lot.
 
Cascade lake x pricing from today puts the 10 core at 580 - 20 cheaper than the 9900ks pricing.

So now you have options
 
Back
Top Bottom