How long before Intel get their act together?

[Rroff]

LGA 1700 with Alder Lake will likely be the moment Intel leapfrogs AMD, especially if they use TSMC's 7nm or 5nm process to built it on.

I'm like 99% sure a lot of the details out there for Alder Lake are wrong.

They tried with me and even then they couldn't make it competitive. Seriously they cant make them free and the only way it's worth an intel build with everything factored in is if they literally give me the kit! If you run the numbers it's increasingly difficult to make intel look even in the same league. So if you literally can't give them away because, power draw, because patching, because so many reasons. I know what your thinking it doesn't matter but it really matters in a DC where you pay by the amp.

Yeah - I'm no AMD fan for *reasons* but if I was building a professional server setup, etc. at the moment there is no way I'd touch Intel. Almost all the systems I've built lately for friends and family I've put Ryzen in unless they are using older software that is optimised for Intel as when spending someone else's money I just can't spend it on Intel in good conscience. (My own builds will likely still be Intel for some time to come however).

 

[Vince]

I'm like 99% sure a lot of the details out there for Alder Lake are wrong.



Yeah - I'm no AMD fan for *reasons* but if I was building a professional server setup, etc. at the moment there is no way I'd touch Intel. Almost all the systems I've built lately for friends and family I've put Ryzen in unless they are using older software that is optimised for Intel as when spending someone else's money I just can't spend it on Intel in good conscience. (My own builds will likely still be Intel for some time to come however).

I think we are pretty much almost at the point where it's not about being a fan one way or the other. I know people in here know I waited for Rome and had budget sat there for 2 years that I refused to spend because I couldn't stomach buying Intel again and was willing to migrate everything. I also know people in here know I have a preference to not buy Intel at the moment (For context I literally finished a 9th gen desktop rollout of approx 100 prodesk 400 G5 (i think) on desktop not long ago and am already replacing those again, again for power/size budget issues). The simple fact of the matter is that if you sit down and be boring putting numbers into a spreadsheet you just can't make the intel estate look compelling unless you have some bias for whatever reason.

The way I see it and this may be unpopular is people are buying Intel in the DC because 1) They are already a big Intel estate and cant be doing with the migration 2) Are tied to Intel because of an app or instruction set or something. Anybody that has actually run the performance numbers, core count, memory, server price and other metrics properly will or should agree.

I love that we are so far on either end of the scale, but you know what I don't think we have ever butt heads over the AMD / Intel argument. You simply have to respect that people will have their reasons one way or another, personally for me there isn't a processor in the Intel stack that I would actually buy right now. In fact they couldn't pay me to take one of their chips at this point in time, even though they are still decent on the desktop. AMD on the other hand id love a 3rd gen TR if you are listening and want to show me some more love! Just saying!

 

[jigger]

They tried with me and even then they couldn't make it competitive. Seriously they cant make them free and the only way it's worth an intel build with everything factored in is if they literally give me the kit! If you run the numbers it's increasingly difficult to make intel look even in the same league. So if you literally can't give them away because, power draw, because patching, because they get spanked silly by the competitions chips, because so many reasons.

I know what your thinking, it doesn't matter! But it really does matters in a DC where you pay by the rack and by the amp, if people are super interested I can flesh out the current costing of my 10 racks including power draw etc for both AMD and Intel to give an idea where the lay of the land is right now in the server space.

Yeah, it’s a sad state of affairs but, to make Intel work for us would require equipment plus cash. Free wouldn’t cover the losses over the lifetime vs Rome (never mind Milan) and as you point out even if you did take free kit you’ll always be operating from behind the curve.

In a couple of meetings I’ve had to step in and stop reps in their tracks. Even with incentives on the table it’s still bad for business.

 

[Vince]

Yeah, it’s a sad state of affairs but, to make Intel work for us would require equipment plus cash. Free wouldn’t cover the losses over the lifetime vs Rome (never mind Milan) and as you point out even if you did take free kit you’ll always be operating from behind the curve.

In a couple of meetings I’ve had to step in and stop reps in their tracks. Even with incentives on the table it’s still bad for business.

You see that is really interesting. On the scale I'm operating at and over a period of 5 years, when I ran the numbers on Rome before it was released it probably could have worked if they gave me everything for free. By everything I mean chassis, cpu's and memory, but even then the other issues such as patching/security and the monster power draw were still a turn off. I was that confident back then in my numbers that you might remember I ended up with pre production EPYC Rome servers and chips? So early in fact nobody else seemed to be able to buy them and one of my servers is still running a pre production HP bios. I also had to be sent a couple of Naples CPU's to get the things running (so if you ever read that this was just people saying it and not true, I can assure you it is) Anyway for me I'm absolutely convinced I made the right decision, they have been in production for about a year and they are literally epic.

I do hope intel come up with something though, for me at least for the forseable it's all about density, I need as much horsepower in the smallest space possible while consuming as little power as possible. Power draw is money to me at the moment as running everything out of a colo isn't cheap so anything that reduces power draw or consumed space is a total win. Intel know what they need to do, they need to meet AMD on density/power draw/cost. If they do that they are back in the running. For me they are no longer the number 1 vendor in the server space and haven't been since the first day Rome hit the market.

I can assure you had I known what was going to happen and that we would move out of our office and run entirely remote like this that decision back then would have been far easier. As it is my decision back then is saving us even more today than it was when I made the call. I think it was you that agreed back in the epyc thread! Anyway just saying we were right back then so are probably still right today

 

[Kelt]

My firm is looking at a DC build right now. Intel would have to pay us to use their equipment.

Isn't that their standard business practise?

 

[jigger]

You see that is really interesting. On the scale I'm operating at and over a period of 5 years, when I ran the numbers on Rome before it was released it probably could have worked if they gave me everything for free. By everything I mean chassis, cpu's and memory, but even then the other issues such as patching/security and the monster power draw were still a turn off. I was that confident back then in my numbers that you might remember I ended up with pre production EPYC Rome servers and chips? So early in fact nobody else seemed to be able to buy them and one of my servers is still running a pre production HP bios. I also had to be sent a couple of Naples CPU's to get the things running (so if you ever read that this was just people saying it and not true, I can assure you it is) Anyway for me I'm absolutely convinced I made the right decision, they have been in production for about a year and they are literally epic.

I do hope intel come up with something though, for me at least for the forseable it's all about density, I need as much horsepower in the smallest space possible while consuming as little power as possible. Power draw is money to me at the moment as running everything out of a colo isn't cheap so anything that reduces power draw or consumed space is a total win. Intel know what they need to do, they need to meet AMD on density/power draw/cost. If they do that they are back in the running. For me they are no longer the number 1 vendor in the server space and haven't been since the first day Rome hit the market.

I do remember, at the time I was considering Gigabyte EPYC boxes.

Looking at AMD and Intel from a business perspective things quickly become very black and white. I think this situation is probably the best thing that could have happened to Intel as it forces them to focus on the hardware and separate it from the marketing fluff.

 

[Steampunk]

I do remember, at the time I was considering Gigabyte EPYC boxes.

Looking at AMD and Intel from a business perspective things quickly become very black and white. I think this situation is probably the best thing that could have happened to Intel as it forces them to focus on the hardware and separate it from the marketing fluff.

That may take some time, as Intel always uses their money and marketing muscle to make up any deficit, but AMD has specifically targeted where Intel is weak. Intel are going to have to fundamentally change their core philosophies around how they do business and the way they design products in order to combat AMD.

AMD have taken a step into the future with what they've done in modular design, interconnects, and process nodes, and Intel are left languishing in the old way, trying to catch up. Yes, Intel has a lot of money, but they are playing catch-up. It may take a lot of time to turn the Intel ship given the marketing, engineering, and leadership inertia they seem to be carrying.

Look at how many years it took AMD to right their ship, sell off the fabs, get a visionary with an engineering background into a leadership role to remake the company into something that can create world beating products. Those are the kinds of major sea changes Intel is going to have to make to compete with a resurgent AMD.

And that's before we start talking about where ARM (with or without Nvidia), Apple, or RISC-V might go during that period of the next five years when Intel has to remake themselves.