I'm getting a strong feeling of lust for this CPU. 150W ain't that much. Imagine what this CPU could have been if it had been developed further, with modern branch predictors and huge caches.
Question Intel had a 7 GHz CPU years ago
- Thread starter igor_kavinski
- Start date
Thunder 57
Platinum Member
- Aug 19, 2007
- 2,673
- 3,794
- 136
What kind of nonsense are you babbling about? a 50 stage pipeline? Ha! And it was AMD's fault? I may just watch this and get back to you.
EDIT, yeah I'm not watching that. I think I saw it or something like it before. There's a reason Intel abandoned Netburst. It was a dud.
Well, AMD is the reason Intel couldn't force their nonsense architectures on us 
It was obviously a good thing but a part of me wishes that Intel had continued their experiments in parallel with their business successes. If they can put out an anemic Atom CPU for the masses, why not a single core ultra high frequency CPU for the extra giggles? People would buy it, just to brag about having the highest GHz CPU in existence. So, in a sense, AMD's success with the Athlon 64 forced Intel to abandon their vanity project.
It was obviously a good thing but a part of me wishes that Intel had continued their experiments in parallel with their business successes. If they can put out an anemic Atom CPU for the masses, why not a single core ultra high frequency CPU for the extra giggles? People would buy it, just to brag about having the highest GHz CPU in existence. So, in a sense, AMD's success with the Athlon 64 forced Intel to abandon their vanity project.
- Aug 22, 2001
- 28,437
- 20,428
- 146
He's a cool guy, glad you found his channel. Even if it took a filthy redditor to do it. And about your thread title; Thanks, I hate it.
And about your thread title; Thanks, I hate it.
Thunder 57
Platinum Member
- Aug 19, 2007
- 2,673
- 3,794
- 136
Yup, that's the video I saw awhile back. If anything, Prescott should have been the Pentium 5. Either way, Netburst was going nowhere. A single core narrow CPU at 10GHz would be weaksuace against a modern multicore in anything that matters.
It would make a great retro gaming rig for games based on single core engines. I will try to beam my wish into Elon's mind tonight
TheELF
Diamond Member
- Dec 22, 2012
- 3,973
- 730
- 126
No it wouldn't, because those games are made for a specific range of FPS, if it gets faster the faster the core then it will be broken and unplayable because you wouldn't have a chance of controlling anything.It would make a great retro gaming rig for games based on single core engines. I will try to beam my wish into Elon's mind tonight
At least intel had the sense to not release this thing in the first place, unlike AMD that just went ahead and released the whole dozer gen that was a step back in cores.
But, but, but Intel promised 10GHz eventually on the P4...
Fun fact; P4s run the ALUs double pumped, so a hypothetical 10GHz P4 would have ALUs running at an effective 20GHz.
It would make a great retro gaming rig for games based on single core engines. I will try to beam my wish into Elon's mind tonight
Fun fact; P4s run the ALUs double pumped, so a hypothetical 10GHz P4 would have ALUs running at an effective 20GHz.
Prescott was the >7 GHz one, Tejas was the >9 GHz one.
Prescott:
ieeexplore.ieee.org
Tejas:
ieeexplore.ieee.org
The Pentium 5 (Nehalem) Netburst Pro (>15 GHz) required the:
Prescott:
Low-voltage-swing logic circuits for a 7GHz x86 integer core
Pentium/spl reg/4 processor architecture uses a 2x core clock to implement low latency integer operations. Low-voltage-swing logic circuits in 90nm technology meet the frequency demands of a 3rd generation integer core, with operation demonstrated for frequencies in excess of 7GHz.
Tejas:
A 9-GHz 65-nm Intel® Pentium 4 Processor Integer Execution Unit
<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> This paper describes a fourth generation Intel Pentium 4 processor integer execution core operating at 9 GHz in a 1.3-V, 65-nm CMOS technology at 70<formula> <tex>$\,^{\circ} {\hbox...
The Pentium 5 (Nehalem) Netburst Pro (>15 GHz) required the:
Last edited:
At this point, with power consumption approaching 400W, maybe they should handover the Tejas blueprints to some interns and get a few thousand limited edition dies out at their most under-utilized fab.
30 GHz! Did they run their simulations on a megafarm of Pentium chips with the FDIV bug???
AdamK47
Lifer
- Oct 9, 1999
- 15,204
- 2,837
- 126
This is an AMD heavy forum, so of course this is going to be another Intel vs AMD thread with a title to bait the majority.
No, it was a material/design push...
IBM, AMD unveil terahertz transistor breakthroughs
My chip is smaller than your chip
A Renaissance for Chip Technologies
New chip and transistor breakthroughs are arriving from AMD, Intel and IBM.
www.eweek.com
The main reason why K9 (5 GHz) -> K10 (10 GHz) -> K11 (20 GHz) and the Netburst successors didn't become a thing. Was because that neither AMD nor Intel implemented TeraHertz.
Last edited:
I'm almost sure than on Prescott and later they ran at same clock speed than the rest of the core.
Prescott was such a garbage.
If only they did a 64 bit Northwood. Nah. Too smart. Northwood was an OK CPU (considering).
moinmoin
Diamond Member
- Jun 1, 2017
- 4,944
- 7,656
- 136
For one core it actually is. You successfully make Unlimited RTL look efficient.
Was heat the only issue that prevented its commercialization?
Or was it so fast that it made them very, very afraid?
Boss: So howz the progress going on the Terahertz transistor?
Scientists: It's wayyyyy too fast. If we implement this, it will be the fastest transistor ever and then we'll be stuck. It's theoretically impossible to be any faster.
Boss: (to himself): That's bad. We can't have everyone thinking that there won't be any further progress ever. It will be a disaster for the sustainability of our business model (sell new processors every year with incremental improvements).
Scientists: What should we do?
Boss: Cease any further development on it. There's no future in it. Technological progress must never halt, no matter what!
Found the following links for Terahertz:
Researchers Create Terahertz Transistor-like Device
Why is graphene special, what did the researchers demonstrate, and how could it be used in future electronics?
www.electropages.com
Trying for a Terahertz Transistor
A new transistor design aims to smash speed records.
Quad socket P5 would have consumed only 600W. Who needs more than 4 threads?For one core it actually is. You successfully make Unlimited RTL look efficient.
Even octasocket P5 could be possible. 1600W PSUs are not prohibitively expensive.
The CPU doesn't get much love from us. We are feeding 600W to the current reigning GPU champ!
From the pdf above they state
1. Reduces leakage current by 10,000X for the same capacitance 2. Reduces unwanted current flow by 100X 3. Increased electron mobility 4. Increased reliability 5. High speed 6. Ease of circuit design 7. No leakage path through substrate 8. Lowest junction capacitance 9. Less voltage required to turn ON transistor 10. Eliminates subsurface leakage 11. Solves high resistance 12. Eliminates floating body effect 13. Minimizes soft error rates 14. 50% lower junction capacitance than that of Partially Depleted SOI
1. Reduces leakage current by 10,000X for the same capacitance 2. Reduces unwanted current flow by 100X 3. Increased electron mobility 4. Increased reliability 5. High speed 6. Ease of circuit design 7. No leakage path through substrate 8. Lowest junction capacitance 9. Less voltage required to turn ON transistor 10. Eliminates subsurface leakage 11. Solves high resistance 12. Eliminates floating body effect 13. Minimizes soft error rates 14. 50% lower junction capacitance than that of Partially Depleted SOI
Intel's Netburst/AMD's Hi-Frequency future designs didn't have heat/power issues on the new processes. It was only a problem if Intel or AMD stuck with Bulk or PDSOI. The main reason why they didn't launch terahertz specifically Intel's Ultra-thin FDSOI in the 2000s. Came from a scare that FDSOI didn't scale between nodes. It was only noticed by Leti/IBM/STMicroelectronics that silicon height and buried oxide also need to shrink for scaling.
45nm FDSOI -> 32nm FDSOI => Shrink the Lg, but Intel kept the same silicon height and buried oxide depth.
FDSOI group in 2007+ (this might be on the web-archive, but STMicroelectronics announced that FDSOI had scaling thus were rolling out ST-E processors on FDSOI):
The whole structure/substrate needs to shrink to scale.
Hence;
SOITECs:
UTBOX25
UTBOX20
UTBOX15
UTBOX15-High Mobility
etc.
Foundry:
28nm = tSi -> 8.5nm
22nm = tSi -> 7 nm
14nm = tSi -> 5.5 nm
etc.
Last edited:
That reminds me. Around 2005, I read an interview of some Global Foundries scientist who said that SOI had a serious drawback. One night, he was having a glass of wine. Suddenly, he was hit by a moment of clarity and realized that the drawback could be used to their advantage. I can't find that interview. Forgot the scientist's name.
Thunder 57
Platinum Member
- Aug 19, 2007
- 2,673
- 3,794
- 136
That's interesting, considering GloFo didn't exist in 2005. And FDSOI is Nosta's fetish.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
guys.
Guys!
GUYS!!!
7GHz was the HOPED for frequency for Tejas. There was no chance of that product ever reaching 7GHz. They HOPED for Prescott to reach 6+GHz.
How would a 10GHz be even POSSIBLE if the highest ever frequency reached using exotic, liquid nitrogen cooling fails to reach 9GHz? And that's a Celeron Prescott chip.
Also here's what the linked article about the "terahertz transistor" says
Really? They were that much of idiots back then or something? Sometimes I don't understand how the outsiders see the trend coming before the people that actually work on these things.
You know that the Tri-Gate transistor(now known as FinFET) introduced with 22nm Intel process pretty much achieves all that right? Even the FDSOI effect. The High-K dielectric mentioned above to reduce gate leakage was addressed in 2007 with 45nm!
The so-called Terahertz transistor was just a stepping stone on the way to commercialize the real thing, which was the Tri-Gate transistor.
The whole Netburst architecture didn't make sense. Only the process side saved Northwood. That's why in mobile Pentium 4-M(which was Northwood) was quickly abandoned and Pentium III-M and Pentium M took over.
Last edited:
Prescott's Integer core operated at 2x core frequency => https://ieeexplore.ieee.org/document/1332640
Thus, 3.4 GHz => 6.8 GHz and 3.6 GHz => 7.2 GHz
Tejas's Integer core operated at 2x core frequency => https://ieeexplore.ieee.org/document/4039605
~4.5 GHz -> 9 GHz
I don't think anyone is going to say Tejas was going for >7 GHz core clock. Since, they never stated that for Tejas only Nehalem's Netburst. Where it would have been on first-gen TeraHertz.
Actually, the whole TeraHertz Transistor was specifically FDSOI. Since, Intel dropped TeraHertz early on, they pulled across the features meant for FDSOI to Bulk while preparing for FinFETs. This is the actual cause of the Netburst getting killed off; no process -> no product.
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 9K
-
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
-
Facebook
Twitter