Oh man!, I am so freaking nostalgic about Nehalem

Rakanoth

n00b
Joined
Oct 6, 2017
Messages
48
Do you remember when Nehalem came out? It was a huge, incredible performance jump. It was crazy. I had bought a complete new system with this chip i7-920 and combined it with an Radeon HD5870. It was Battlefield BC2 and Dirt 2 times, the games were amazing, the weather was amazing and the amount of Frames per Second was amazing too with this setup.

Maybe, because I was much younger, I felt much more pleasure back then. I am not sure.

That CPU can still hold up at 60Hz monitor at 1080p, depending on the game. I am not sure but I believe it outperforms most of the current mobile CPUs out there.

I read somewhere that Intel put lots of R&D money into developing this CPU and less and less over time after it. Maybe that is a total rubbish information. But, anyway, I just can't stop wishing for another Nehalem.

Is it true that Nehalem was the biggest jump ever on mainstream desktop platform? I might wrong. Correct me please.
Do you think we can see another huge jump like Nehalem?
How did you feel about this architecture back then? :)
 
I'm still running my 920 in an HTPC box that can do nearly anything. Nehelam was great, but Gulftown even better! It does seem Nehelam was the largest jump in performance for Intel and the near future isn't looking too good.
 
I retired all my 920 D0s back in 2013 (switched to Haswell at launch),
but still have a crapton of Westmeres (4x X5690 in 2x HP Z800s, 8x X5670s in 4x HP Z600s)
 
i'd rather see another sandy bridge.

now that a cpu that is still kick ass.

Indeed.

It's amazing how much they still fetch on the used market. Fucking seven year old CPU and the 2600k still fetches $100 easy. And finding a Z68/Z77 board to throw it in costs almost as much....

My own i5-2500k is still kicking ass at 4.6ghz on a Gigabyte Z68XP-UD3 in the kid's gaming / guest use PC.
 
Indeed.

It's amazing how much they still fetch on the used market. Fucking seven year old CPU and the 2600k still fetches $100 easy. And finding a Z68/Z77 board to throw it in costs almost as much....

My own i5-2500k is still kicking ass at 4.6ghz on a Gigabyte Z68XP-UD3 in the kid's gaming / guest use PC.

Because again was a good platform, had the little Nudge over 1366 that was wanted.
 
Do you remember when Nehalem came out? It was a huge, incredible performance jump. It was crazy. I had bought a complete new system with this chip i7-920 and combined it with an Radeon HD5870. It was Battlefield BC2 and Dirt 2 times, the games were amazing, the weather was amazing and the amount of Frames per Second was amazing too with this setup.

Maybe, because I was much younger, I felt much more pleasure back then. I am not sure.

That CPU can still hold up at 60Hz monitor at 1080p, depending on the game. I am not sure but I believe it outperforms most of the current mobile CPUs out there.

I read somewhere that Intel put lots of R&D money into developing this CPU and less and less over time after it. Maybe that is a total rubbish information. But, anyway, I just can't stop wishing for another Nehalem.

Is it true that Nehalem was the biggest jump ever on mainstream desktop platform? I might wrong. Correct me please.
Do you think we can see another huge jump like Nehalem?
How did you feel about this architecture back then? :)
Both 286->386 and 32-bit->64bit were much larger jumps, IMHO
 
Buying off lease Dell or HP Sandy desktops and workstations gives you an incredible bang for the buck.
This++.

I won an off-lease dell for my wife to upgrade her Q8200/2G ram system to an i5-3570 with 4G ram. What a jump in performance.....for $94 shipped.
 
Yes, Sandy has a noticeable IPC increase over Nehalem, but Nehalem brought SMT back from the dead. The difference today is a larger impact than even 2008 timeframe because multithread is much more important today. Nehalem and Westmere have provided now dirt cheap 8 and 12 thread systems that still pack a punch for today's use. Sandy and SB-E only brought an IPC increase but retained the same thread count. So Core 2 --> Nehalem was a bigger jump in the long run than Nehalem--> Sandy Bridge.
 
I'm sure a lot of professionals really liked what happened this year with Ryzen/Threadripper. Definitely not a substantial IPC increase, but AMD's core count increase directly affected the market and forced Intel's hand on it's expensive 6+ core cpus. I myself enjoyed the switch to Conroe from the dreaded Pentium D.
 
I'm sure a lot of professionals really liked what happened this year with Ryzen/Threadripper. Definitely not a substantial IPC increase, but AMD's core count increase directly affected the market and forced Intel's hand on it's expensive 6+ core cpus. I myself enjoyed the switch to Conroe from the dreaded Pentium D.

Uhh, IPC increase is huge with Ryzen vs Vishera/FX CPUs.
 
I meant the total industries IPC, they obviously didn't catch up to Intel on that front.
No they just closed the gap and became competitive again. Intel has had very little increase since Sandy though.
 
I rocked my 920 for quite a while. I replaced it with my 2600k which I still have to this day. I sold my 920 to a buddy and he used it for years. I have to check with him, but I thought I heard him say his system finally gave up the ghost last year, but it may have been a MB issue, not the processor. I had spilled water on that processor about 3 separate times when I was experimenting with different custom cooling setups. After cleaning it and airing it out, it kept chugging on with no issues.
 
My 920 was a total new build from the ground up (in a clear plexi case) and huge performance jump for me. Dual to quad and higher clocks. The Gigabyte x58 UD5 mobo I ran it on was one of the most beautiful boards ever imo. I think I was able to run mine at 3.8ghz sans HT on air which was pretty good. I ran that for a long while.

(my 920 which was reserved the day of release at Central Computer in San Francisco)
IMG_0612.JPG


I think the other huge cpu performance jump was to Conroe E6600 which I got 3.5ghz out of. That was a jump from the Opteron I was running before.
 
Last edited:
Do you remember when Nehalem came out? It was a huge, incredible performance jump. It was crazy. I had bought a complete new system with this chip i7-920 and combined it with an Radeon HD5870. It was Battlefield BC2 and Dirt 2 times, the games were amazing, the weather was amazing and the amount of Frames per Second was amazing too with this setup.

Maybe, because I was much younger, I felt much more pleasure back then. I am not sure.

That CPU can still hold up at 60Hz monitor at 1080p, depending on the game. I am not sure but I believe it outperforms most of the current mobile CPUs out there.

I read somewhere that Intel put lots of R&D money into developing this CPU and less and less over time after it. Maybe that is a total rubbish information. But, anyway, I just can't stop wishing for another Nehalem.

Is it true that Nehalem was the biggest jump ever on mainstream desktop platform? I might wrong. Correct me please.
Do you think we can see another huge jump like Nehalem?
How did you feel about this architecture back then? :)

The old golden times have gone.

x86 performance hits two walls: the frequency wall and the IPC wall. I add next a beautiful illustration of the frequency wall (as you can see Intel is not the only affected, any vendor is, including non-x86)

f7.jpg


The existence of those two walls have been known for a long time. That is the reason why Intel/HP engineers used the transition to 64bit as opportunity to abandon a serial ISA as x86. They developed a scalable 64bit ISA, IA64, but the whole approach failed due to a combination of multiple factors and the industry embraced AMD64. Now we are stuck with a bloated, inefficienct, and nonscalable ISA.

Some researchers are treating to continue the work of Intel and HP. For instance people working in the Mill CPU. They expect to get IPC levels ~4x higher than current top x86 CPUs as Haswell. I can see them getting that performance in some specific code, but I am skeptic for general case (and I am not alone).

Intel AVX ISA can be seen as an continuation of some concepts behind IA64, except that AVX is only an extension to x86 and targets only specific workloads, unlike IA64.

So, unless someone comes tomorrow with some magic solution. I think the epoch of increasing IPC has finished. Of course, this doesn't imply Intel cannot continue adding incremental small IPC updates from further optimizations of the current SS/OOO pipeline, but each update will be more difficult to achieve.

About frequencies, unless silicon is replaced with some other material with different electric properties, the epoch of increasing frequencies has finished as well. Of course, engineers can work hard to optimize and get some few percents here and here. But you will not see a 10GHz CPU in next three years.

With IPC and frequency walls the only remaining option to increase general performance by significant ways is adding moar cores. This is why we move from single to dual cores then quads... The problem here is Amdahl's law and other basic laws of computing that introduce another wall to performance. For each degree of parallelism in the code, there is a performance wall beyond which adding more cores doesn't increase performance.
 
Back
Top