Greater AMD

AMD is indeed great company. If not AMD now we would buy Intel processors with awful perfomance (like Pentium 4 or Itanium), overpriced, and requiring overpriced platform and memory (Rambus...)

Nevertheless Intel is too powerful and capable to great things if kicked to ass.

In this TL among IT history I listed some events which could help AMD to perform more successfully.

1993 – first Am486’s, joint venture Fujitsu AMD Semiconductor Limited, or Spansion

1994 – AMD aquires Nexgen Nx686 assets (RISC86 ALU core) and team, AMD’s team helps with FPU & P5 bus& production (implementation), Compaq and AMD form long-term alliance, Am486DX5-75/133 16KbL1, Am5x86-100-133 as Pentium analog (Am486DX5 core+32Kb L1 on P5 bus)
1995 – Large-scale production begins at Fab 25, AMD Nx686 (32KbL1 + AMD's FPU), Intel launched PentiumPro

1996 – Dirk Meyer (21264) started Alpha-like K7 CPU design
AMD takes part in NGIO, FIO, creates with DEC и Compaq HyperTransport: PCI programming model (IRQ, DMA), LVDS peer-peer proc-proc[chipset] serial scalable high-speed packet-switched bus, chain topology, DDR. Coherent interprocessor mode – low latency, cash coherency. Based on IEEE 1596.3-1996 SCI-LVDS (Scalable Coherent Interface).
Partially free specifications for chipset/peripheral implementation-only HT1 max – noncoherent 32bit/400 – free HT controller and implementation help

1996end – AMD K6-2 (MMX+64KbL1+pipelined FPU), Cyrix acquired by NatSemi, AMD develops SuperSocket7: AGP, SDRAM PC100, 75/83/100FSB(25/28/33 PCI)

1997 – ATI aquires Tseng Labs's graphics assets, adopted engineer’s team, launches Rage128, Intel aquires Chips&Technologies(DX5,laptop), Nvidia Riva128

1997 – 0.25, K6-3 3DNow+ 166-300 MHz, Intel PentiumMMX, PII, AGP, K7 announce
1998 – 0.25, K6-3 400 128KbL2, PowerNow, AMD-Motorola 0.18 copper alliance, common fabs usage, 3Dfx buys STB, Nvidia Riva128ZX (OpenGL ICD), Intel acquire Real3D: i740, AGP2x, Riva TNT, Rage128MAXX.

1999start – 0.25, K7 Athlon/Opteron 400..700 (SSE, 0.25, 128KbL1, 512-2048Mb L2 in MultiChipPackage (IBM's C4 flip-chip) ), ThermalControl, heatspreader, SocketHT 1 link up to 2*32bit 600 MHz 4.8 GB/s.
Chipsets: for Athlon – 1 link ( 2*16bit 400 MHz 1.6 GB/s) (VIA(AMDtuned) –AGP4x+64bit SDR133),
for Opteron chipset – up to 4 links (AMD+Alpha: 256bit ECC SDRAM).
“Ahead the race – Athlon", "MORE Perfomance for the same price”
1999-2001 - real - Intel not successful with RDRAM, Itanium, 1GHz 0.18 P3, 0.18 Pentium4
1999 – ATI acquired ArtX, Via – Centaur и Cyrix, AMD builds up Fab30 200-мм 0.18, Riva TNT2, Rage128Pro, Rage128MAXX Pro, DX6.1(Win98SE), DX7 (Win2000), alternative Radeon256 (pureDX7->simple->fast, HyperZ, 4*1, SDR&DDR), Geforce256, PIII, AGP4x, i820, ATI chipset for AMD Athlon: HT+64bit133MhzSDRAM+integratedRage

2000 – Creative buys out Aureal (bankrupcy), Nvidia buys out 3Dfx, GeForce2GTS, Nvidia’s and ATI’s 128bit SDRAM SocketHT чипсеты (ATI uses Ali south bridge), Flipper (DX8.1 GameCube, Wii), Intel-Amd x86-extensions crosslicensing (SSE2,x64), AMD lets UMC, TSMC manufacture K5+, K6 to stiffen low-end competition with Intel & Via in socket7 segment.

2000end – 0.18 K8 AthlonXP/Athlon64/Opteron 800-1200, 256Kb L2, x86-64, SocketHT2-800 (backward compatible with SocketHT), 1/2/4 links HT2 ([2*16bit]800MHz 3.2GB/s) (3-other CPUs,1-chipset/memory), heatspreader, MS DX8, ATI Radeon2 (8500)(pureDX8->simple->fast(325MHz), HyperZ, 4*2),
Geforce3, Nvidia’s and ATI’s 128bit DDR SocketHT чипсет, Nvidia in XBOX.

2001 – HP buys Compaq, DX8.1 (WinXP), Athlon 2000+, P4 2000, Sun’s Opteron adoption and DDR chipset.

2002 – Intel acquires 3Dlabs (DX9.0, OpenGL2 P10), AMD merges with ATI, 0.15 R300(R9700), 0.13 K8 3000+, AGP8x, AMD-IBM 0.09 alliance, ATI Imageon, Via buys S3 graphics division with DX9 chip designs

2003 – AMD buys Sensaura и Ali(Uli)-for chipset techs, SIS buys Trident. DX9, Win2003. Geforce 5x00(NV30), October – ATI 9800XT. Freescale, Chartered make 0.13 Athlon64s for AMD. AMD lets UMC, TSMC to make K7s to increase supply, propagation and adoption of AMD HT platform.
ATI quickly advances over Nvidia in manufacturing tech, XBOX360: dualcoreK8+R500(unified shaders architecture), CrossFire over HT2

2004 – AMD starts to build Fab36 (in operation – 2006, 65 nm), May – R420 0.13low-k 16*1, September– X700, 6600, October – 0.11low-k R430 16*1 (DX9+), K8 MCP 1/2/4 cores (../DUAL/QUADRO)

2005 – 0.09 R440 Radeon X900 (24*1), Geforce7800, September – X950, Gf7900, 2-4 ATI chips via HT-Crossfire, “More power at less power”,
Fusion MCP for low-end/mobile: CPU+X300HT-DDR2/DDR3”.
2006 – Intel Conroe ("PentiumPro strikes back";-) )(0.065 dualcore 64bit, FB-DIMM),

AMD Athlon K9 X2 (half-quadcore design ;-), L3 cash, higher IPc), 0.09 256bit GDDR4(2,5GHz) DX10 Radeon1800 (24 ROP + 32 TMU, external interface chip (HDMI/TDMI/TVOut/Audio/Crossfire) and 384bit GDDR3(1600GHz) Geforce8800,
Nvidia buys Transmeta for CPU technologies

2007 – 0.08 Radeon 1850 256bit GDDR4@3200, Gf8900, Radeon 1600, Vista, 0.045 K10, MCP Athlon quadcore
Intel buys Havok, AMD buys PhysX
 
Last edited:
Far too detailed for me to have a clue what's going on! And, are those, Russian characters or something snook in there?

What are the actual effects of this, just AMD as a market leader in processors (electronics), so that we have cheaper and more efficient hardware! :)
 

Fyrwulf

Banned
Probably not market leader, but far more competition with Intel. I think AMD has something like 25% of total microprocessor market share and it still makes huge waves. An AMD with a large and successful graphics division plus a more successful (not to mention rapid) implementation of core technologies would probably land AMD something in the range of 30-40% microprocessor share and at least 50% market share in graphics. In other words, AMD is just as capable in its own way as Intel in this TL with fewer shortcoming (money and infrastructure.)
 
Probably not market leader, but far more competition with Intel. I think AMD has something like 25% of total microprocessor market share and it still makes huge waves. An AMD with a large and successful graphics division plus a more successful (not to mention rapid) implementation of core technologies would probably land AMD something in the range of 30-40% microprocessor share and at least 50% market share in graphics. In other words, AMD is just as capable in its own way as Intel in this TL with fewer shortcoming (money and infrastructure.)
yeah, you are quite right.

So let me underline major changes
1.Buying Nexgen in 1994 not in 1996 could let AMD don't waste time on its own awful K5. Nexgen wouldn't waste time and money on own platform (chipsets) and creating FPU which is no better than AMD's (not pipelined either).
So Intel could be kicked in ass in 1995.

2. NO EV6 bus - Earlier Hypertransport (EV7) even if AMD would launch early Athlons on Socket7 and then faster move to HT.
3. Then for ATI - it should launch Radeon R100 as 4*1 not 2*3. With their superiority in memory contollers (HyperZ) they could compete with Nvidia.

4. ATI - awful architecture with ringbus, little TMU and ROPs. I would fire man who pushed this ideas.


5. AMD or ATI could buy ULi in 2005 before Nvidia to strengthen their position in chipsets.
6. AMD should make SocketAM2 and Socket939 backward compatible (there so small difference)
7. ATI should sell all non-graphics divisions (e.g. DTV for 1 billion $) before merging with AMD. Or at least divide in 2 parts.
8. Radeon2900 (R600) should be 256bit GDDR4 not 512bit GDDR3
9. Barcelona should be native dualcore with 2Mb L2 - not huge native quadcore.
Quads should be like Intel MCP. And taking into account HT - six-core is viable with no losses for bus throughtput.
10. There should be only one 65nm CPU - dualcore Barcelona. No waste of time on Brisbane.

11. In the middle of 90s AMD could lead technology and market by earlier than Intel introducing MMX and 3DNow! instruction sets. 'Cos Nexgen had MMX earlier or same time as Intel.
 
Last edited:
Additionally
1. AMD might not buy Geode and concentrate on CPUs (until merge with ATI)
2. AMD might not shell money for Alchemy Semiconductor - don't go for handhelds.
3. Iа AMD wanted somу graphics - it would buy Trident in 2003 for little money and get some platform and experience. Uli could be sold to AMD in 2003 (last chance - 2005).
In this case AMD could not merge with ATI because of having ex-Trident some graphics(not ATI level but better than Intel) and ULi for chpsets.
 
Far too detailed for me to have a clue what's going on! And, are those, Russian characters or something snook in there?
Yes, note yyz's location in ?nizhny? novgorod. Although why he names himself after Toronto's airport, I'm sure I don't know!

1996 – Dirk Meyer (21264) started Alpha-like K7 CPU design
AMD takes part in NGIO, FIO, creates with DEC и Compaq HyperTransport:
1996end – AMD K6-2 (MMX+64KbL1+pipelined FPU), Cyrix acquired by NatSemi, AMD develops SuperSocket7: AGP, SDRAM PC100, 75/83/100FSB(25/28/33 PCI)

и is 'and' in Russian (here and a couple of places below)

2000 – Creative buys out Aureal (bankrupcy), Nvidia buys out 3Dfx, GeForce2GTS, Nvidia’s and ATI’s 128bit SDRAM SocketHT чипсеты (ATI uses Ali south bridge), Flipper (DX8.1 GameCube, Wii), Intel-Amd x86-extensions crosslicensing (SSE2,x64), AMD lets UMC, TSMC manufacture K5+, K6 to stiffen low-end competition with Intel & Via in socket7 segment.
чипсеты =chipsety directly transliterated. Presumably Russian for chipsets.
 
Yes, note yyz's location in ?nizhny? novgorod. Although why he names himself after Toronto's airport, I'm sure I don't know!



и is 'and' in Russian (here and a couple of places below)


чипсеты =chipsety directly transliterated. Presumably Russian for chipsets.

My excuses. Text was initially in 2 languages and some Russian words left.

About "Nizhny Novgorod"
It's a joke. 'Cos in Russian Nizhny=low, lower
and Novgorod=Noviy+Gorod=New+City
 
I don't think a more competitive AMD would make a whole lot of difference. Performance is constrained by possess technology more than anything else, and it's unlikely to advance much faster even with AMD and Intel throwing more money at the problem.

Intel might be forced to sell with lower margins, but as it is CPUs are actually quite cheap on the low end already.

Now a Russian x86 competitor - now that would be interesting. There were Russian x86 clones, so it's not that much of a stretch.

Would you use a Russian CPU?
 
I don't think a more competitive AMD would make a whole lot of difference. Performance is constrained by possess technology more than anything else, and it's unlikely to advance much faster even with AMD and Intel throwing more money at the problem.

Intel might be forced to sell with lower margins, but as it is CPUs are actually quite cheap on the low end already.

Now a Russian x86 competitor - now that would be interesting. There were Russian x86 clones, so it's not that much of a stretch.

Would you use a Russian CPU?
Intel and IBM-AMD process techs are comparable - AMD is just half-step behind.
AMD's architechture strting from K7 is more advanced.
It's just Intel marketing, supply and money force + market inertia.
Last good Intel CPU was P6 (up to PIII Tualatin). Then inefficient P4.
And 2006 - Core 2 Duo - son of P6.

About Russian x86. Last x86 russian clones were 286. then Russia just imported 386+ and motherboards from Taiwan.
And by the way nobody can blame USSR for cloning - it was usual in 80s-90s practice of reverse-engineering of market best chips. It was chinese, japanese, european and many american companies practice.
And only AMD (and Cyrix) could turn it piracy clones of x86 in something legal.

Russian x86 impossible 'cos Russia in 90s had no money (and no legal right to purchase such fabs due to Jackson-Venick act). And Russia never had own design of x86. If you haв heard about Elbrus - it's virtual CPU - no real chips. And best russian CPUs in military avionics are clones of Sun Sparc architechture.
 
You said Russians had no money, but small companies have developed quite capable CPUs. Had Russia put more importance on CPU technology early on they might have been able to keep up.

I've heard that Russia did have their own CPU arch, but management forced them to clone x86 CPUs due to a "they're western so they're better" mentality. Probably just an urban legend though.
 
You said Russians had no money, but small companies have developed quite capable CPUs. Had Russia put more importance on CPU technology early on they might have been able to keep up.

I've heard that Russia did have their own CPU arch, but management forced them to clone x86 CPUs due to a "they're western so they're better" mentality. Probably just an urban legend though.

Never heard of such companies with x86 clones (may be Elbrus project which was positioned as Itanium competitor, but it's fake) though I am IT specialist and interested in technology history.
I've heard about small companies that develop specific DSP processors - heritage of military experience and great mathematic knowledge.

It's not urban legend - it's real - mainly because large western software base (pirated) - it's started in middle of 60s with IBM System/360/370 clones - Soviet ES EVM series.
then contiued in x86 clones - for PC.
 
Top