FPGA DEVICES · ENGINEERING · SOURCING Source-linked information. Clear project requirements.

Hard-to-find

70T3319S133BC: Check the Read Mode Before Buying a 133 MHz RAM

Published · Updated

70T3319S133BC is an obsolete synchronous dual-port RAM whose speed label does not tell you which read mode your board uses. The exact manufacturer ordering row identifies a 256K × 18 device, standard power, a BC256 CABGA package and commercial grade. This purchasing dossier helps a maintenance buyer preserve that identity and ask the engineer the questions that can change an order: output mode, port-voltage selection, package finish and collision handling. It provides no stock, price, delivery promise or approved substitute.

Start with the released BOM, board revision, the supplier’s complete proposed code and your required quantity. The output is an evidence-backed sourcing request with unresolved engineering questions attached. This is an AI-assisted document review, not a hardware test, authenticity assessment or operating-temperature qualification. An inconsistency in the source’s temperature terminology is kept visible rather than resolved by assumption.

Read the official electrical datasheet · Open the revised original EOL notice · Request the exact MPN, quantity 25

On small screens, each diagram scrolls horizontally to keep its labels readable. The captions and section text provide the same decision boundaries.

BC256 and a tray are part of the purchasing identity

Keep the whole 70T3319S133BC purchase identityThe exact ordering row links 70T3319S133BC to 133 MHz, BC256, CABGA and commercial C. Blank packing means tray. Different finish, package or carrier proposals need separate review.Keep the whole 70T3319S133BC purchase identity70T3319 | S | 133 | BC | no G / I / 8Device and power256K × 18; Sstandard powerPackage identityBC256 / CABGA256 connectionsPurchase boundaryTray; commercialnot a green-code claimOriginal document-review diagram · not a pinout, footprint or measured result
The exact ordering row links 70T3319S133BC to 133 MHz, BC256, CABGA and commercial C. Blank packing means tray. Different finish, package or carrier proposals need separate review.

On page 26 of DSC-5652/11, November 2019, the literal orderable row reads 70T3319S133BC, 133 MHz, BC256, CABGA, C. The code explanation gives S as standard power, BC as the 256-connection BGA option and a blank packing field as tray. A final 8 indicates tape and reel in this specific ordering table. Keep that last point tied to this family: carrier letters and numbers have different meanings in other manufacturers’ ordering systems.

The absence of G is also relevant. The same document defines a separate green-finish position, while the exact-part web record lists this code as Pb-free No and category e0. Do not rewrite the requested code with G because a supplier calls its offer “equivalent.” A proposed finish change needs its own material, assembly and identity review. The ordering table’s separate commercial C and industrial I rows likewise do not make one grade a silent substitute for another.

The source identifies the 70T3319 organization as 256K by 18 bits. That is 262,144 addressable words and 4,718,592 storage bits; the two ports access one shared array, not two separate arrays with twice that capacity. The density field expressed as 4608 Kb in the catalogue uses binary multiples. Record organization and bus width alongside the exact code so a plain “4M memory” description does not conceal the 18-bit data path.

Use the complete-device-code lesson when a supplier response shortens the string or changes its suffix. The useful comparison is the required code against the entire proposed code. A matching string still establishes neither provenance nor fit on the actual board. A different string is a question to investigate, not an automatically rejected technology and not an automatically accepted replacement. Preserve the supplier’s original wording in the private purchase record.

Read the March addendum without moving the July last-buy date

A revised cover does not reset the last-buy clockPLC230018-A1 was issued March 29, 2024; the exact affected row retains its January 3 notice date,July 3 last buy andJanuary 3, 2025 last shipment. Its replacement field is None.A revised cover does not reset the last-buy clock2024-01-032024-03-292024-07-032025-01-03Original rowA1 coverLast buyLast shipmentA1 removes one other code; these two deadlines stay unchanged.Original document-review diagram · not a pinout, footprint or measured result
PLC230018-A1 was issued March 29, 2024; the exact affected row retains its January 3 notice date,July 3 last buy andJanuary 3, 2025 last shipment. Its replacement field is None.

The source used here is Renesas notification PLC230018-A1. Its cover is dated March 29, 2024. It explains that Addendum 1 removes RH9N0611000GNB#BC0 from the discontinued list and leaves last-buy and last-shipment dates unchanged. Page 2 still explicitly lists 70T3319S133BC. The row retains January 3, 2024 (2024-01-03) as the notice date, July 3, 2024 (2024-07-03) as last buy, January 3, 2025 (2025-01-03) as last shipment and None in the replacement field. These are separate document-revision and affected-row facts.

As of this November 2026 review, both historical deadlines are past. They describe the manufacturer’s stated EOL process; they do not prove that any particular reseller has a genuine lot, can deliver it or has permission to reopen manufacturer ordering. The cover also describes NCNR shipments in the stated interval and retains discretion over accepting or cancelling orders. A date in a notice is not a present-day supply commitment.

“None” means the affected-row replacement column provides no candidate. It is not proof that no redesign could ever work. A buyer should separate two paths: ask about the exact original code with traceability evidence, or ask engineering to evaluate an explicitly identified alternative. The second path needs an accountable change process and a record of what the source does and does not establish. Merely selecting another row from the same document is not that process.

Keep the A1 cover and the exact affected row together in the sourcing file. The original January notification remains a separate source record; use the addendum to understand the later removal, not to erase the original chronology. This prevents the revised cover date being mistaken for a new six-month ordering window. The Hard-to-find evidence directory provides the broader notice context, but this dossier’s later revision check does not silently change every other directory record. No complete search for every possible later rescission is claimed; any newer official revision must be reconciled before relying on its changed terms.

The 2.5 V core and the port’s logic level answer different questions

Separate the core rail from each port’s I/O choiceThe core is 2.4–2.6 V. Each port can use 2.4–2.6 V I/O with OPT at ground or 3.15–3.45 V I/O with OPT tied to the 2.5 V core supply. Temperature measurement basis remains unresolved.Separate the core rail from each port’s I/O choiceCore VDD2.4–2.6 Vall modesFor EACH port: match VDDQ and OPT2.4–2.6 V I/O → OPT = VSS3.15–3.45 V I/O → OPT = VDDTemperature: table heading and note use different bases.Keep this question open; no thermal suitability decision.Original document-review diagram · not a pinout, footprint or measured result
The core is 2.4–2.6 V. Each port can use 2.4–2.6 V I/O with OPT at ground or 3.15–3.45 V I/O with OPT tied to the 2.5 V core supply. Temperature measurement basis remains unresolved.

Page 7 gives recommended core VDD from 2.4 to 2.6 V. Each port has a separate VDDQ choice. For 2.5 V I/O operation, its VDDQ range is 2.4–2.6 V and its OPT input is tied to VSS. For 3.3 V I/O operation, VDDQ is 3.15–3.45 V and OPT is tied to VDD, the 2.5 V core supply. Pages 5 and 7 require the appropriate supply and option levels before signals are applied to that port. “3.3V interface” does not mean the core can be powered at 3.3 V.

The two OPT choices are independent, so the two ports can use different documented I/O levels. The option and pipeline-mode inputs are treated as static configuration signals during normal operation. Do not copy ordinary address-input thresholds onto every pin: page 7 provides distinct limits for JTAG and for ZZ, OPT and the mode-selection inputs. A schematic review should identify the actual source of each strap and verify the relevant pin class, power sequence and receiver limits.

The commercial temperature row has an unresolved definition issue. Its column heading refers to ambient temperature, but note 1 defines TA as the instant-on case temperature. The printed numeric range is 0–70 °C. This dossier does not choose a measurement basis or turn that range into an ambient, junction or thermal-design approval. Obtain an applicable manufacturer clarification or corrected source before making an environmental suitability decision. Increasing a procurement quantity cannot close this engineering question.

Package identity is similarly narrower than a footprint approval. The exact-part page lists BC256, CABGA, 256 connections and a 17 × 17 mm package entry with 1 mm pitch. Those catalogue dimensions are not a complete mechanical tolerance drawing or a land pattern. Preserve the original package and ball-map sources with the released board files. The manufacturing-file package check can organize missing inputs; its output cannot verify pad geometry, assembly yield or the ambiguous temperature basis.

Test the read-mode assumption before repeating “133 MHz”

The S133 column has two different read-mode limitsPipelined mode has a minimum 7.5 ns clock period and maximum 4.2 ns clock-to-data; flow-through has 25 ns and 15 ns. A hypothetical 10 ns clock clears only the pipeline period gate, not full timing.The S133 column has two different read-mode limitsPIPELINEDtCYC2 ≥ 7.5 nstCD2 ≤ 4.2 nsFLOW-THROUGHtCYC1 ≥ 25 nstCD1 ≤ 15 nsExample: T = 10 ns → period gate clears / period gate failsClock period alone does not establish latency or board closure.Original document-review diagram · not a pinout, footprint or measured result
Pipelined mode has a minimum 7.5 ns clock period and maximum 4.2 ns clock-to-data; flow-through has 25 ns and 15 ns. A hypothetical 10 ns clock clears only the pipeline period gate, not full timing.

The commercial S133 column on page 11 has two timing sets. Pipelined operation specifies a minimum 7.5 ns clock period and maximum 4.2 ns clock-to-data interval. Flow-through specifies a minimum 25 ns period and maximum 15 ns clock-to-data interval. These are not interchangeable descriptions of one unconditional 133 MHz interface. The mode strap selects the relevant source waveform, and the controller must account for the corresponding latency and enable behavior.

Here is a deliberately limited worksheet. Suppose a proposed controller clock has a 10 ns period, equivalent to 100 MHz. Subtracting the source’s pipeline minimum gives 10 − 7.5 = 2.5 ns of clock-period headroom for that single check. Applying the flow-through minimum gives 10 − 25 = −15 ns, so that proposed period fails the flow-through period requirement. A negative result is a stop condition. A positive pipeline result is only a cleared period gate; it is not a finished interface budget.

Keep maximum clock-to-data delay separate from the minimum clock period. The source also lists setup, hold, clock high and low times and output-enable limits. Page 12 shows the mode-specific read waveforms. Pipeline chip and byte enables have double-buffered behavior; a deselection request does not make the bus available immediately. Work through the actual edge sequence with the controller’s sampling convention before changing a mode strap or shortening a wait state.

For a second small calculation, the trace propagation estimate can keep assumed flight time visible. Four inches at an explicitly assumed 100 ps per inch produce 400 ps, or 0.4 ns. Enter length 101.6 mm and velocity 254 mm/ns in the tool for that same stated assumption. This is an arithmetic example, not a measured board velocity. It cannot be subtracted from a headline 133 MHz number to prove timing closure. Include the real stackup, package and controller delays, loading, skew, setup and hold requirements, mode latency and the tool’s own limitations in the complete analysis.

Collision flags report trouble; they do not grant ownership

A collision alert is not ownership permissionAt a matched registered address, read/read is not a collision; read/write alerts the reader; write/write alerts both ports. Flag latency and coverage limits mean external ownership control is still necessary.A collision alert is not ownership permissionPort L / Port RFlag consequence at address matchREAD / READNeither port reports a valid collisionREAD / WRITEReader reports collisionWRITE / WRITEBoth report collisionCOL is an alert path; it does not arbitrate or repair stored data.Original document-review diagram · not a pinout, footprint or measured result
At a matched registered address, read/read is not a collision; read/write alerts the reader; write/write alerts both ports. Flag latency and coverage limits mean external ownership control is still necessary.

Two independent clock interfaces are useful when separate devices exchange bursts, but they do not make simultaneous access safe by themselves. Page 20’s collision truth table uses matched internal registered addresses. Both ports reading is not a valid collision. When one reads and the other writes, the reader reports a collision. When both write, both ports report one. The table is a description of alert behavior under its conditions, not a transactional memory protocol.

The collision-detection discussion on page 22 warns that the flags can appear after the second or third relevant clock edge and can miss alerts for consecutive collision events. A flag is therefore not a BUSY signal that has already prevented the harmful write. Do not transplant the master/slave ownership behavior of an asynchronous dual-port family onto this synchronous device. The controller or system protocol must decide who owns a shared location and how a consumer knows a complete record is ready.

A document-level exercise is to draw two sequences for one shared address. First, let the left side write a payload while the right side reads that same registered location. Mark the read result as requiring the source’s collision and port-to-port validity analysis; do not label it “old” or “new” by guess. Second, give the producer exclusive ownership, complete its payload transfer and hand ownership over using a reviewed protocol. The diagram is a planning aid, not implemented or simulated HDL.

The 6116LA20SOG8 sourcing example is a useful architectural contrast: it concerns a single-port asynchronous SRAM, not this synchronous 18-bit dual-port interface, and is not a substitute recommendation. A replacement proposal must preserve or deliberately redesign the interface, capacity, power and ownership behavior as well as package connections. No alternative model is approved here, even if its nominal bit count or package connection count happens to match.

Send an exact-code enquiry with a separate engineering question list

Make the procurement handoff answerableKeep exact-code sourcing separate from engineering approval. Send the original MPN and quantity; retain mode, rail, package, temperature and lot-evidence questions in controlled project records.Make the procurement handoff answerable1. IdentityOriginal full MPNBoard revisionQuantity / carrierSource revision2. EngineeringRead-mode settingPort rail + OPTCollision protocolOpen temp basis3. SourcingLot traceabilityFinish / packageCondition evidenceNo assumed stockOriginal document-review diagram · not a pinout, footprint or measured result
Keep exact-code sourcing separate from engineering approval. Send the original MPN and quantity; retain mode, rail, package, temperature and lot-evidence questions in controlled project records.

Create a compact handoff with three parts. The identity part contains the original full MPN, board revision, quantity, required carrier and the supplier’s unedited proposed code. The engineering part records each port’s voltage choice and OPT strap, output mode, controller timing constraints, collision/ownership protocol and unresolved temperature definition. The sourcing part contains the exact EOL row and addendum revision, requested lot and condition evidence, finish and package requirements and the change-approval owner.

Use the full-MPN catalogue check to inspect how a code is represented on this site. A catalogue miss for this legacy memory is an absence in that tool’s covered catalogue, not proof of a counterfeit or nonexistent manufacturer code. The original ordering table and affected-row evidence remain necessary. Likewise, a catalogue match cannot certify the offered lot, resolve thermal measurement terminology or replace the board’s controlled bill of materials.

For lot questions, the legacy IC evidence guide explains how to ask for missing history without converting a date code into a pass/fail authenticity score. Separate what is documented from what is merely reported by the seller. Quantity 25 in this worked handoff is an example demand, not a manufacturer minimum, a tray multiple or available inventory. Use the actual needed quantity when preparing your request; do not round it to a package-carrier count without agreement.

The useful stopping rules are explicit. A changed finish or package needs review; an unknown read mode prevents a timing conclusion; an unresolved operating-temperature basis prevents environmental qualification; and missing lot history remains a sourcing question. Evidence for one gate cannot be borrowed to close another. This page supports an exact-code enquiry and a better engineering question list. It does not authorize board fabrication, a substitute part, an operating environment or an unverified supplier claim.

Prepare a homepage enquiry for 25 × 70T3319S133BC. Change the quantity to your actual need. For a board or assembly project, use the homepage enquiry and provide the controlled BOM, schematic and relevant constraints through the project process; do not place private designs in URL parameters.

Open the original order-code review GIF (close to hide motion)
Shared original concept animation showing exact code comparison, source lookup and retained open questions; not 70T3319 electrical behavior.
Shared original order-code concept only. Any other example codes in this animation are not memory parameters or replacement recommendations.

Static walkthrough: preserve 70T3319S133BC and the proposed code; find the literal ordering and affected-notice rows; retain the pipeline/flow-through distinction; record the unresolved temperature measurement basis; prepare an exact-code enquiry with actual quantity. The animation is not a read-cycle simulation, pinout or measured collision waveform.

Sources reviewed November 10, 2026: DSC-5652/11, November 2019, especially pp. 5, 7, 11–12, 20, 22 and 26; PLC230018-A1, March 29, 2024, cover and exact p.2 affected row; official exact-part catalogue snapshot. AI-assisted document review; no human engineering certification, HDL simulation, physical measurement or supplier-lot inspection.

Back to Hard-to-find →

Source addresses and citation trail

Full source addresses are retained below. Read each citation with its document revision, page, table and applicability conditions above.

  1. https://www.renesas.com/en/document/dst/70t33391999-datasheet
  2. https://www.renesas.com/en/document/eln/plc-230018-a1-end-life-eol-process-select-part-numbers
  3. https://www.renesas.com/en/document/eln/plc-230018-end-life-process-select-parts?r=13431
  4. https://www.renesas.com/en/products/70t3319/part-details/70t3319s133bc